Move docs

Tested on Fedora 40 with Podman 5.1.1 and ruff "0.5.0" and "latest".
source: https://unix.stackexchange.com/q/651198

.cargo/config.toml

@@ -1,3 +1,10 @@
 [alias]
 dev = "run --package ruff_dev --bin ruff_dev"
 benchmark = "bench -p ruff_benchmark --bench linter --bench formatter --"
+
+# statically link the C runtime so the executable does not depend on
+# that shared/dynamic library.
+#
+# See: https://github.com/astral-sh/ruff/issues/11503
+[target.'cfg(all(target_env="msvc", target_os = "windows"))']
+rustflags = ["-C", "target-feature=+crt-static"]

.gitattributes

@@ -8,6 +8,10 @@ crates/ruff_linter/resources/test/fixtures/pycodestyle/W391_3.py text eol=crlf
 crates/ruff_python_formatter/resources/test/fixtures/ruff/docstring_code_examples_crlf.py text eol=crlf
 crates/ruff_python_formatter/tests/snapshots/format@docstring_code_examples_crlf.py.snap text eol=crlf
 
+crates/ruff_python_parser/resources/invalid/re_lexing/line_continuation_windows_eol.py text eol=crlf
+crates/ruff_python_parser/resources/invalid/re_lex_logical_token_windows_eol.py text eol=crlf
+crates/ruff_python_parser/resources/invalid/re_lex_logical_token_mac_eol.py text eol=cr
+
 crates/ruff_python_parser/resources/inline linguist-generated=true
 
 ruff.schema.json linguist-generated=true text=auto eol=lf

.github/CODEOWNERS

@@ -15,3 +15,6 @@
 
 # Script for fuzzing the parser
 /scripts/fuzz-parser/ @AlexWaygood
+
+# red-knot
+/crates/red_knot/ @carljm @MichaReiser

.github/renovate.json5

@@ -27,9 +27,27 @@
       // Group upload/download artifact updates, the versions are dependent
       groupName: "Artifact GitHub Actions dependencies",
       matchManagers: ["github-actions"],
+      matchDatasources: ["gitea-tags", "github-tags"],
       matchPackagePatterns: ["actions/.*-artifact"],
       description: "Weekly update of artifact-related GitHub Actions dependencies",
     },
+    {
+      // This package rule disables updates for GitHub runners:
+      // we'd only pin them to a specific version
+      // if there was a deliberate reason to do so
+      groupName: "GitHub runners",
+      matchManagers: ["github-actions"],
+      matchDatasources: ["github-runners"],
+      description: "Disable PRs updating GitHub runners (e.g. 'runs-on: macos-14')",
+      enabled: false,
+    },
+    {
+      // Disable updates of `zip-rs`; intentionally pinned for now due to ownership change
+      // See: https://github.com/astral-sh/uv/issues/3642
+      matchPackagePatterns: ["zip"],
+      matchManagers: ["cargo"],
+      enabled: false,
+    },
     {
       groupName: "pre-commit dependencies",
       matchManagers: ["pre-commit"],

.github/workflows/build-binaries.yml

@@ -1,21 +1,23 @@
-name: "[ruff] Release"
+# Build ruff on all platforms.
+#
+# Generates both wheels (for PyPI) and archived binaries (for GitHub releases).
+#
+# Assumed to run as a subworkflow of .github/workflows/release.yml; specifically, as a local
+# artifacts job within `cargo-dist`.
+name: "Build binaries"
 
 on:
-  workflow_dispatch:
+  workflow_call:
     inputs:
-      tag:
-        description: "The version to tag, without the leading 'v'. If omitted, will initiate a dry run (no uploads)."
-        type: string
-      sha:
-        description: "The full sha of the commit to be released. If omitted, the latest commit on the default branch will be used."
-        default: ""
+      plan:
+        required: true
         type: string
   pull_request:
     paths:
-      # When we change pyproject.toml, we want to ensure that the maturin builds still work
+      # When we change pyproject.toml, we want to ensure that the maturin builds still work.
       - pyproject.toml
       # And when we change this workflow itself...
-      - .github/workflows/release.yaml
+      - .github/workflows/build-binaries.yml
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.ref }}
@@ -23,6 +25,7 @@ concurrency:
 
 env:
   PACKAGE_NAME: ruff
+  MODULE_NAME: ruff
   PYTHON_VERSION: "3.11"
   CARGO_INCREMENTAL: 0
   CARGO_NET_RETRY: 10
@@ -31,11 +34,12 @@ env:
 
 jobs:
   sdist:
+    if: ${{ !contains(github.event.pull_request.labels.*.name, 'no-build') }}
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
         with:
-          ref: ${{ inputs.sha }}
+          submodules: recursive
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ env.PYTHON_VERSION }}
@@ -49,8 +53,8 @@ jobs:
       - name: "Test sdist"
         run: |
           pip install dist/${{ env.PACKAGE_NAME }}-*.tar.gz --force-reinstall
-          ruff --help
-          python -m ruff --help
+          ${{ env.MODULE_NAME }} --help
+          python -m ${{ env.MODULE_NAME }} --help
       - name: "Upload sdist"
         uses: actions/upload-artifact@v4
         with:
@@ -58,11 +62,12 @@ jobs:
           path: dist
 
   macos-x86_64:
+    if: ${{ !contains(github.event.pull_request.labels.*.name, 'no-build') }}
     runs-on: macos-12
     steps:
       - uses: actions/checkout@v4
         with:
-          ref: ${{ inputs.sha }}
+          submodules: recursive
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ env.PYTHON_VERSION }}
@@ -74,11 +79,6 @@ jobs:
         with:
           target: x86_64
           args: --release --locked --out dist
-      - name: "Test wheel - x86_64"
-        run: |
-          pip install dist/${{ env.PACKAGE_NAME }}-*.whl --force-reinstall
-          ruff --help
-          python -m ruff --help
       - name: "Upload wheels"
         uses: actions/upload-artifact@v4
         with:
@@ -86,23 +86,29 @@ jobs:
           path: dist
       - name: "Archive binary"
         run: |
-          ARCHIVE_FILE=ruff-${{ inputs.tag }}-x86_64-apple-darwin.tar.gz
-          tar czvf $ARCHIVE_FILE -C target/x86_64-apple-darwin/release ruff
+          TARGET=x86_64-apple-darwin
+          ARCHIVE_NAME=ruff-$TARGET
+          ARCHIVE_FILE=$ARCHIVE_NAME.tar.gz
+
+          mkdir -p $ARCHIVE_NAME
+          cp target/$TARGET/release/ruff $ARCHIVE_NAME/ruff
+          tar czvf $ARCHIVE_FILE $ARCHIVE_NAME
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
         uses: actions/upload-artifact@v4
         with:
-          name: binaries-macos-x86_64
+          name: artifacts-macos-x86_64
           path: |
             *.tar.gz
             *.sha256
 
   macos-aarch64:
+    if: ${{ !contains(github.event.pull_request.labels.*.name, 'no-build') }}
     runs-on: macos-14
     steps:
       - uses: actions/checkout@v4
         with:
-          ref: ${{ inputs.sha }}
+          submodules: recursive
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ env.PYTHON_VERSION }}
@@ -126,18 +132,24 @@ jobs:
           path: dist
       - name: "Archive binary"
         run: |
-          ARCHIVE_FILE=ruff-${{ inputs.tag }}-aarch64-apple-darwin.tar.gz
-          tar czvf $ARCHIVE_FILE -C target/aarch64-apple-darwin/release ruff
+          TARGET=aarch64-apple-darwin
+          ARCHIVE_NAME=ruff-$TARGET
+          ARCHIVE_FILE=$ARCHIVE_NAME.tar.gz
+
+          mkdir -p $ARCHIVE_NAME
+          cp target/$TARGET/release/ruff $ARCHIVE_NAME/ruff
+          tar czvf $ARCHIVE_FILE $ARCHIVE_NAME
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
         uses: actions/upload-artifact@v4
         with:
-          name: binaries-aarch64-apple-darwin
+          name: artifacts-aarch64-apple-darwin
           path: |
             *.tar.gz
             *.sha256
 
   windows:
+    if: ${{ !contains(github.event.pull_request.labels.*.name, 'no-build') }}
     runs-on: windows-latest
     strategy:
       matrix:
@@ -151,7 +163,7 @@ jobs:
     steps:
       - uses: actions/checkout@v4
         with:
-          ref: ${{ inputs.sha }}
+          submodules: recursive
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ env.PYTHON_VERSION }}
@@ -163,13 +175,16 @@ jobs:
         with:
           target: ${{ matrix.platform.target }}
           args: --release --locked --out dist
+        env:
+          # aarch64 build fails, see https://github.com/PyO3/maturin/issues/2110
+          XWIN_VERSION: 16
       - name: "Test wheel"
         if: ${{ !startsWith(matrix.platform.target, 'aarch64') }}
         shell: bash
         run: |
           python -m pip install dist/${{ env.PACKAGE_NAME }}-*.whl --force-reinstall
-          ruff --help
-          python -m ruff --help
+          ${{ env.MODULE_NAME }} --help
+          python -m ${{ env.MODULE_NAME }} --help
       - name: "Upload wheels"
         uses: actions/upload-artifact@v4
         with:
@@ -178,18 +193,19 @@ jobs:
       - name: "Archive binary"
         shell: bash
         run: |
-          ARCHIVE_FILE=ruff-${{ inputs.tag }}-${{ matrix.platform.target }}.zip
+          ARCHIVE_FILE=ruff-${{ matrix.platform.target }}.zip
           7z a $ARCHIVE_FILE ./target/${{ matrix.platform.target }}/release/ruff.exe
           sha256sum $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
         uses: actions/upload-artifact@v4
         with:
-          name: binaries-${{ matrix.platform.target }}
+          name: artifacts-${{ matrix.platform.target }}
           path: |
             *.zip
             *.sha256
 
   linux:
+    if: ${{ !contains(github.event.pull_request.labels.*.name, 'no-build') }}
     runs-on: ubuntu-latest
     strategy:
       matrix:
@@ -199,7 +215,7 @@ jobs:
     steps:
       - uses: actions/checkout@v4
         with:
-          ref: ${{ inputs.sha }}
+          submodules: recursive
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ env.PYTHON_VERSION }}
@@ -216,27 +232,36 @@ jobs:
         if: ${{ startsWith(matrix.target, 'x86_64') }}
         run: |
           pip install dist/${{ env.PACKAGE_NAME }}-*.whl --force-reinstall
-          ruff --help
-          python -m ruff --help
+          ${{ env.MODULE_NAME }} --help
+          python -m ${{ env.MODULE_NAME }} --help
       - name: "Upload wheels"
         uses: actions/upload-artifact@v4
         with:
           name: wheels-${{ matrix.target }}
           path: dist
       - name: "Archive binary"
+        shell: bash
         run: |
-          ARCHIVE_FILE=ruff-${{ inputs.tag }}-${{ matrix.target }}.tar.gz
-          tar czvf $ARCHIVE_FILE -C target/${{ matrix.target }}/release ruff
+          set -euo pipefail
+
+          TARGET=${{ matrix.target }}
+          ARCHIVE_NAME=ruff-$TARGET
+          ARCHIVE_FILE=$ARCHIVE_NAME.tar.gz
+
+          mkdir -p $ARCHIVE_NAME
+          cp target/$TARGET/release/ruff $ARCHIVE_NAME/ruff
+          tar czvf $ARCHIVE_FILE $ARCHIVE_NAME
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
         uses: actions/upload-artifact@v4
         with:
-          name: binaries-${{ matrix.target }}
+          name: artifacts-${{ matrix.target }}
           path: |
             *.tar.gz
             *.sha256
 
   linux-cross:
+    if: ${{ !contains(github.event.pull_request.labels.*.name, 'no-build') }}
     runs-on: ubuntu-latest
     strategy:
       matrix:
@@ -258,11 +283,13 @@ jobs:
             arch: ppc64
             # see https://github.com/astral-sh/ruff/issues/10073
             maturin_docker_options: -e JEMALLOC_SYS_WITH_LG_PAGE=16
+          - target: arm-unknown-linux-musleabihf
+            arch: arm
 
     steps:
       - uses: actions/checkout@v4
         with:
-          ref: ${{ inputs.sha }}
+          submodules: recursive
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ env.PYTHON_VERSION }}
@@ -279,8 +306,8 @@ jobs:
         if: matrix.platform.arch != 'ppc64'
         name: Test wheel
         with:
-          arch: ${{ matrix.platform.arch }}
-          distro: ubuntu20.04
+          arch: ${{ matrix.platform.arch == 'arm' && 'armv6' || matrix.platform.arch }}
+          distro: ${{ matrix.platform.arch == 'arm' && 'bullseye' || 'ubuntu20.04' }}
           githubToken: ${{ github.token }}
           install: |
             apt-get update
@@ -295,19 +322,28 @@ jobs:
           name: wheels-${{ matrix.platform.target }}
           path: dist
       - name: "Archive binary"
+        shell: bash
         run: |
-          ARCHIVE_FILE=ruff-${{ inputs.tag }}-${{ matrix.platform.target }}.tar.gz
-          tar czvf $ARCHIVE_FILE -C target/${{ matrix.platform.target }}/release ruff
+          set -euo pipefail
+
+          TARGET=${{ matrix.platform.target }}
+          ARCHIVE_NAME=ruff-$TARGET
+          ARCHIVE_FILE=$ARCHIVE_NAME.tar.gz
+
+          mkdir -p $ARCHIVE_NAME
+          cp target/$TARGET/release/ruff $ARCHIVE_NAME/ruff
+          tar czvf $ARCHIVE_FILE $ARCHIVE_NAME
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
         uses: actions/upload-artifact@v4
         with:
-          name: binaries-${{ matrix.platform.target }}
+          name: artifacts-${{ matrix.platform.target }}
           path: |
             *.tar.gz
             *.sha256
 
   musllinux:
+    if: ${{ !contains(github.event.pull_request.labels.*.name, 'no-build') }}
     runs-on: ubuntu-latest
     strategy:
       matrix:
@@ -317,7 +353,7 @@ jobs:
     steps:
       - uses: actions/checkout@v4
         with:
-          ref: ${{ inputs.sha }}
+          submodules: recursive
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ env.PYTHON_VERSION }}
@@ -340,26 +376,35 @@ jobs:
             apk add python3
             python -m venv .venv
             .venv/bin/pip3 install ${{ env.PACKAGE_NAME }} --no-index --find-links dist/ --force-reinstall
-            .venv/bin/ruff check --help
+            .venv/bin/${{ env.MODULE_NAME }} --help
       - name: "Upload wheels"
         uses: actions/upload-artifact@v4
         with:
           name: wheels-${{ matrix.target }}
           path: dist
       - name: "Archive binary"
+        shell: bash
         run: |
-          ARCHIVE_FILE=ruff-${{ inputs.tag }}-${{ matrix.target }}.tar.gz
-          tar czvf $ARCHIVE_FILE -C target/${{ matrix.target }}/release ruff
+          set -euo pipefail
+
+          TARGET=${{ matrix.target }}
+          ARCHIVE_NAME=ruff-$TARGET
+          ARCHIVE_FILE=$ARCHIVE_NAME.tar.gz
+
+          mkdir -p $ARCHIVE_NAME
+          cp target/$TARGET/release/ruff $ARCHIVE_NAME/ruff
+          tar czvf $ARCHIVE_FILE $ARCHIVE_NAME
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
         uses: actions/upload-artifact@v4
         with:
-          name: binaries-${{ matrix.target }}
+          name: artifacts-${{ matrix.target }}
           path: |
             *.tar.gz
             *.sha256
 
   musllinux-cross:
+    if: ${{ !contains(github.event.pull_request.labels.*.name, 'no-build') }}
     runs-on: ubuntu-latest
     strategy:
       matrix:
@@ -373,7 +418,7 @@ jobs:
     steps:
       - uses: actions/checkout@v4
         with:
-          ref: ${{ inputs.sha }}
+          submodules: recursive
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ env.PYTHON_VERSION }}
@@ -397,204 +442,29 @@ jobs:
           run: |
             python -m venv .venv
             .venv/bin/pip3 install ${{ env.PACKAGE_NAME }} --no-index --find-links dist/ --force-reinstall
-            .venv/bin/ruff check --help
+            .venv/bin/${{ env.MODULE_NAME }} --help
       - name: "Upload wheels"
         uses: actions/upload-artifact@v4
         with:
           name: wheels-${{ matrix.platform.target }}
           path: dist
       - name: "Archive binary"
+        shell: bash
         run: |
-          ARCHIVE_FILE=ruff-${{ inputs.tag }}-${{ matrix.platform.target }}.tar.gz
-          tar czvf $ARCHIVE_FILE -C target/${{ matrix.platform.target }}/release ruff
+          set -euo pipefail
+
+          TARGET=${{ matrix.platform.target }}
+          ARCHIVE_NAME=ruff-$TARGET
+          ARCHIVE_FILE=$ARCHIVE_NAME.tar.gz
+
+          mkdir -p $ARCHIVE_NAME
+          cp target/$TARGET/release/ruff $ARCHIVE_NAME/ruff
+          tar czvf $ARCHIVE_FILE $ARCHIVE_NAME
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
         uses: actions/upload-artifact@v4
         with:
-          name: binaries-${{ matrix.platform.target }}
+          name: artifacts-${{ matrix.platform.target }}
           path: |
             *.tar.gz
             *.sha256
-
-  validate-tag:
-    name: Validate tag
-    runs-on: ubuntu-latest
-    # If you don't set an input tag, it's a dry run (no uploads).
-    if: ${{ inputs.tag }}
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          ref: main # We checkout the main branch to check for the commit
-      - name: Check main branch
-        if: ${{ inputs.sha }}
-        run: |
-          # Fetch the main branch since a shallow checkout is used by default
-          git fetch origin main --unshallow
-          if ! git branch --contains ${{ inputs.sha }} | grep -E '(^|\s)main$'; then
-            echo "The specified sha is not on the main branch" >&2
-            exit 1
-          fi
-      - name: Check tag consistency
-        run: |
-          # Switch to the commit we want to release
-          git checkout ${{ inputs.sha }}
-          version=$(grep "version = " pyproject.toml | sed -e 's/version = "\(.*\)"/\1/g')
-          if [ "${{ inputs.tag }}" != "${version}" ]; then
-            echo "The input tag does not match the version from pyproject.toml:" >&2
-            echo "${{ inputs.tag }}" >&2
-            echo "${version}" >&2
-            exit 1
-          else
-            echo "Releasing ${version}"
-          fi
-
-  upload-release:
-    name: Upload to PyPI
-    runs-on: ubuntu-latest
-    needs:
-      - macos-aarch64
-      - macos-x86_64
-      - windows
-      - linux
-      - linux-cross
-      - musllinux
-      - musllinux-cross
-      - validate-tag
-    # If you don't set an input tag, it's a dry run (no uploads).
-    if: ${{ inputs.tag }}
-    environment:
-      name: release
-    permissions:
-      # For pypi trusted publishing
-      id-token: write
-    steps:
-      - uses: actions/download-artifact@v4
-        with:
-          pattern: wheels-*
-          path: wheels
-          merge-multiple: true
-      - name: Publish to PyPi
-        uses: pypa/gh-action-pypi-publish@release/v1
-        with:
-          skip-existing: true
-          packages-dir: wheels
-          verbose: true
-
-  tag-release:
-    name: Tag release
-    runs-on: ubuntu-latest
-    needs: upload-release
-    # If you don't set an input tag, it's a dry run (no uploads).
-    if: ${{ inputs.tag }}
-    permissions:
-      # For git tag
-      contents: write
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          ref: ${{ inputs.sha }}
-      - name: git tag
-        run: |
-          git config user.email "hey@astral.sh"
-          git config user.name "Ruff Release CI"
-          git tag -m "v${{ inputs.tag }}" "v${{ inputs.tag }}"
-          # If there is duplicate tag, this will fail. The publish to pypi action will have been a noop (due to skip
-          # existing), so we make a non-destructive exit here
-          git push --tags
-
-  publish-release:
-    name: Publish to GitHub
-    runs-on: ubuntu-latest
-    needs: tag-release
-    # If you don't set an input tag, it's a dry run (no uploads).
-    if: ${{ inputs.tag }}
-    permissions:
-      # For GitHub release publishing
-      contents: write
-    steps:
-      - uses: actions/download-artifact@v4
-        with:
-          pattern: binaries-*
-          path: binaries
-          merge-multiple: true
-      - name: "Publish to GitHub"
-        uses: softprops/action-gh-release@v2
-        with:
-          draft: true
-          files: binaries/*
-          tag_name: v${{ inputs.tag }}
-
-  docker-publish:
-    # This action doesn't need to wait on any other task, it's easy to re-tag if something failed and we're validating
-    # the tag here also
-    name: Push Docker image ghcr.io/astral-sh/ruff
-    runs-on: ubuntu-latest
-    environment:
-      name: release
-    permissions:
-      # For the docker push
-      packages: write
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          ref: ${{ inputs.sha }}
-
-      - uses: docker/setup-buildx-action@v3
-
-      - uses: docker/login-action@v3
-        with:
-          registry: ghcr.io
-          username: ${{ github.repository_owner }}
-          password: ${{ secrets.GITHUB_TOKEN }}
-
-      - name: Extract metadata (tags, labels) for Docker
-        id: meta
-        uses: docker/metadata-action@v5
-        with:
-          images: ghcr.io/astral-sh/ruff
-
-      - name: Check tag consistency
-        # Unlike validate-tag we don't check if the commit is on the main branch, but it seems good enough since we can
-        # change docker tags
-        if: ${{ inputs.tag }}
-        run: |
-          version=$(grep "version = " pyproject.toml | sed -e 's/version = "\(.*\)"/\1/g')
-          if [ "${{ inputs.tag }}" != "${version}" ]; then
-            echo "The input tag does not match the version from pyproject.toml:" >&2
-            echo "${{ inputs.tag }}" >&2
-            echo "${version}" >&2
-            exit 1
-          else
-            echo "Releasing ${version}"
-          fi
-
-      - name: "Build and push Docker image"
-        uses: docker/build-push-action@v5
-        with:
-          context: .
-          platforms: linux/amd64,linux/arm64
-          # Reuse the builder
-          cache-from: type=gha
-          cache-to: type=gha,mode=max
-          push: ${{ inputs.tag != '' }}
-          tags: ghcr.io/astral-sh/ruff:latest,ghcr.io/astral-sh/ruff:${{ inputs.tag || 'dry-run' }}
-          labels: ${{ steps.meta.outputs.labels }}
-
-  # After the release has been published, we update downstream repositories
-  # This is separate because if this fails the release is still fine, we just need to do some manual workflow triggers
-  update-dependents:
-    name: Update dependents
-    runs-on: ubuntu-latest
-    needs: publish-release
-    steps:
-      - name: "Update pre-commit mirror"
-        uses: actions/github-script@v7
-        with:
-          github-token: ${{ secrets.RUFF_PRE_COMMIT_PAT }}
-          script: |
-            github.rest.actions.createWorkflowDispatch({
-              owner: 'astral-sh',
-              repo: 'ruff-pre-commit',
-              workflow_id: 'main.yml',
-              ref: 'main',
-            })

.github/workflows/build-docker.yml

@@ -0,0 +1,68 @@
+# Build and publish a Docker image.
+#
+# Assumed to run as a subworkflow of .github/workflows/release.yml; specifically, as a local
+# artifacts job within `cargo-dist`.
+#
+# TODO(charlie): Ideally, the publish step would happen as a publish job within `cargo-dist`, but
+# sharing the built image as an artifact between jobs is challenging.
+name: "[ruff] Build Docker image"
+
+on:
+  workflow_call:
+    inputs:
+      plan:
+        required: true
+        type: string
+  pull_request:
+    paths:
+      - .github/workflows/build-docker.yml
+
+jobs:
+  docker-publish:
+    name: Build Docker image (ghcr.io/astral-sh/ruff)
+    runs-on: ubuntu-latest
+    environment:
+      name: release
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+
+      - uses: docker/setup-buildx-action@v3
+
+      - uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract metadata (tags, labels) for Docker
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: ghcr.io/astral-sh/ruff
+
+      - name: Check tag consistency
+        if: ${{ inputs.plan != '' && !fromJson(inputs.plan).announcement_tag_is_implicit }}
+        run: |
+          version=$(grep "version = " pyproject.toml | sed -e 's/version = "\(.*\)"/\1/g')
+          if [ "${{ fromJson(inputs.plan).announcement_tag }}" != "${version}" ]; then
+            echo "The input tag does not match the version from pyproject.toml:" >&2
+            echo "${{ fromJson(inputs.plan).announcement_tag }}" >&2
+            echo "${version}" >&2
+            exit 1
+          else
+            echo "Releasing ${version}"
+          fi
+
+      - name: "Build and push Docker image"
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          platforms: linux/amd64,linux/arm64
+          # Reuse the builder
+          cache-from: type=gha
+          cache-to: type=gha,mode=max
+          push: ${{ inputs.plan != '' && !fromJson(inputs.plan).announcement_tag_is_implicit }}
+          tags: ghcr.io/astral-sh/ruff:latest,ghcr.io/astral-sh/ruff:${{ (inputs.plan != '' && fromJson(inputs.plan).announcement_tag) || 'dry-run' }}
+          labels: ${{ steps.meta.outputs.labels }}

.github/workflows/ci.yaml

@@ -59,7 +59,6 @@ jobs:
               - "!crates/ruff_python_formatter/**"
               - "!crates/ruff_formatter/**"
               - "!crates/ruff_dev/**"
-              - "!crates/ruff_shrinking/**"
               - scripts/*
               - python/**
               - .github/workflows/ci.yaml
@@ -168,6 +167,9 @@ jobs:
       - uses: Swatinem/rust-cache@v2
       - name: "Run tests"
         shell: bash
+        env:
+          # Workaround for <https://github.com/nextest-rs/nextest/issues/1493>.
+          RUSTUP_WINDOWS_PATH_ADD_BIN: 1
         run: |
           cargo nextest run --all-features --profile ci
           cargo test --all-features --doc
@@ -194,6 +196,54 @@ jobs:
           cd crates/ruff_wasm
           wasm-pack test --node
 
+  cargo-build-release:
+    name: "cargo build (release)"
+    runs-on: macos-latest
+    needs: determine_changes
+    if: ${{ needs.determine_changes.outputs.code == 'true' || github.ref == 'refs/heads/main' }}
+    timeout-minutes: 20
+    steps:
+      - uses: actions/checkout@v4
+      - name: "Install Rust toolchain"
+        run: rustup show
+      - name: "Install mold"
+        uses: rui314/setup-mold@v1
+      - uses: Swatinem/rust-cache@v2
+      - name: "Build"
+        run: cargo build --release --locked
+
+  cargo-build-msrv:
+    name: "cargo build (msrv)"
+    runs-on: ubuntu-latest
+    needs: determine_changes
+    if: ${{ needs.determine_changes.outputs.code == 'true' || github.ref == 'refs/heads/main' }}
+    timeout-minutes: 20
+    steps:
+      - uses: actions/checkout@v4
+      - uses: SebRollen/toml-action@v1.2.0
+        id: msrv
+        with:
+          file: "Cargo.toml"
+          field: "workspace.package.rust-version"
+      - name: "Install Rust toolchain"
+        run: rustup default ${{ steps.msrv.outputs.value }}
+      - name: "Install mold"
+        uses: rui314/setup-mold@v1
+      - name: "Install cargo nextest"
+        uses: taiki-e/install-action@v2
+        with:
+          tool: cargo-nextest
+      - name: "Install cargo insta"
+        uses: taiki-e/install-action@v2
+        with:
+          tool: cargo-insta
+      - uses: Swatinem/rust-cache@v2
+      - name: "Run tests"
+        shell: bash
+        env:
+          NEXTEST_PROFILE: "ci"
+        run: cargo +${{ steps.msrv.outputs.value }} insta test --all-features --unreferenced reject --test-runner nextest
+
   cargo-fuzz:
     name: "cargo fuzz"
     runs-on: ubuntu-latest
@@ -207,10 +257,13 @@ jobs:
       - uses: Swatinem/rust-cache@v2
         with:
           workspaces: "fuzz -> target"
-      - name: "Install cargo-fuzz"
-        uses: taiki-e/install-action@v2
+      - name: "Install cargo-binstall"
+        uses: cargo-bins/cargo-binstall@main
         with:
           tool: cargo-fuzz@0.11.2
+      - name: "Install cargo-fuzz"
+        # Download the latest version from quick install and not the github releases because github releases only has MUSL targets.
+        run: cargo binstall cargo-fuzz --force  --disable-strategies crate-meta-data --no-confirm
       - run: cargo fuzz build -s none
 
   fuzz-parser:
@@ -288,7 +341,7 @@ jobs:
           name: ruff
           path: target/debug
 
-      - uses: dawidd6/action-download-artifact@v3
+      - uses: dawidd6/action-download-artifact@v6
         name: Download baseline Ruff binary
         with:
           name: ruff
@@ -380,22 +433,16 @@ jobs:
           name: ecosystem-result
           path: ecosystem-result
 
-  cargo-udeps:
-    name: "cargo udeps"
+  cargo-shear:
+    name: "cargo shear"
     runs-on: ubuntu-latest
     needs: determine_changes
     if: ${{ needs.determine_changes.outputs.code == 'true' || github.ref == 'refs/heads/main' }}
-    timeout-minutes: 20
     steps:
       - uses: actions/checkout@v4
-      - name: "Install nightly Rust toolchain"
-        # Only pinned to make caching work, update freely
-        run: rustup toolchain install nightly-2023-10-15
-      - uses: Swatinem/rust-cache@v2
-      - name: "Install cargo-udeps"
-        uses: taiki-e/install-action@cargo-udeps
-      - name: "Run cargo-udeps"
-        run: cargo +nightly-2023-10-15 udeps
+      - uses: cargo-bins/cargo-binstall@main
+      - run: cargo binstall --no-confirm cargo-shear
+      - run: cargo shear
 
   python-package:
     name: "python package"

.github/workflows/daily_fuzz.yaml

@@ -0,0 +1,72 @@
+name: Daily parser fuzz
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: "0 0 * * *"
+  pull_request:
+    paths:
+      - ".github/workflows/daily_fuzz.yaml"
+
+permissions:
+  contents: read
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+env:
+  CARGO_INCREMENTAL: 0
+  CARGO_NET_RETRY: 10
+  CARGO_TERM_COLOR: always
+  RUSTUP_MAX_RETRIES: 10
+  PACKAGE_NAME: ruff
+  FORCE_COLOR: 1
+
+jobs:
+  fuzz:
+    name: Fuzz
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    # Don't run the cron job on forks:
+    if: ${{ github.repository == 'astral-sh/ruff' || github.event_name != 'schedule' }}
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - name: Install uv
+        run: curl -LsSf https://astral.sh/uv/install.sh | sh
+      - name: Install Python requirements
+        run: uv pip install -r scripts/fuzz-parser/requirements.txt --system
+      - name: "Install Rust toolchain"
+        run: rustup show
+      - name: "Install mold"
+        uses: rui314/setup-mold@v1
+      - uses: Swatinem/rust-cache@v2
+      - name: Build ruff
+        # A debug build means the script runs slower once it gets started,
+        # but this is outweighed by the fact that a release build takes *much* longer to compile in CI
+        run: cargo build --locked
+      - name: Fuzz
+        run: python scripts/fuzz-parser/fuzz.py $(shuf -i 0-9999999999999999999 -n 1000) --test-executable target/debug/ruff
+
+  create-issue-on-failure:
+    name: Create an issue if the daily fuzz surfaced any bugs
+    runs-on: ubuntu-latest
+    needs: fuzz
+    if: ${{ github.repository == 'astral-sh/ruff' && always() && github.event_name == 'schedule' && needs.fuzz.result == 'failure' }}
+    permissions:
+      issues: write
+    steps:
+      - uses: actions/github-script@v7
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          script: |
+            await github.rest.issues.create({
+              owner: "astral-sh",
+              repo: "ruff",
+              title: `Daily parser fuzz failed on ${new Date().toDateString()}`,
+              body: "Runs listed here: https://github.com/astral-sh/ruff/actions/workflows/daily_fuzz.yml",
+              labels: ["bug", "parser", "fuzzer"],
+            })

.github/workflows/notify-dependents.yml

@@ -0,0 +1,29 @@
+# Notify downstream repositories of a new release.
+#
+# Assumed to run as a subworkflow of .github/workflows/release.yml; specifically, as a post-announce
+# job within `cargo-dist`.
+name: "[ruff] Notify dependents"
+
+on:
+  workflow_call:
+    inputs:
+      plan:
+        required: true
+        type: string
+
+jobs:
+  update-dependents:
+    name: Notify dependents
+    runs-on: ubuntu-latest
+    steps:
+      - name: "Update pre-commit mirror"
+        uses: actions/github-script@v7
+        with:
+          github-token: ${{ secrets.RUFF_PRE_COMMIT_PAT }}
+          script: |
+            github.rest.actions.createWorkflowDispatch({
+              owner: 'astral-sh',
+              repo: 'ruff-pre-commit',
+              workflow_id: 'main.yml',
+              ref: 'main',
+            })

.github/workflows/pr-comment.yaml

@@ -17,7 +17,7 @@ jobs:
   comment:
     runs-on: ubuntu-latest
     steps:
-      - uses: dawidd6/action-download-artifact@v3
+      - uses: dawidd6/action-download-artifact@v6
         name: Download pull request number
         with:
           name: pr-number
@@ -32,7 +32,7 @@ jobs:
             echo "pr-number=$(<pr-number)" >> $GITHUB_OUTPUT
           fi
 
-      - uses: dawidd6/action-download-artifact@v3
+      - uses: dawidd6/action-download-artifact@v6
         name: "Download ecosystem results"
         id: download-ecosystem-result
         if: steps.pr-number.outputs.pr-number
@@ -48,6 +48,14 @@ jobs:
         id: generate-comment
         if: steps.download-ecosystem-result.outputs.found_artifact == 'true'
         run: |
+          # Guard against malicious ecosystem results that symlink to a secret
+          # file on this runner
+          if [[ -L pr/ecosystem/ecosystem-result ]]
+          then
+              echo "Error: ecosystem-result cannot be a symlink"
+              exit 1
+          fi
+
           # Note this identifier is used to find the comment to update on
           # subsequent runs
           echo '<!-- generated-comment ecosystem -->' >> comment.txt

.github/workflows/publish-docs.yaml

@@ -1,3 +1,7 @@
+# Publish the Ruff documentation.
+#
+# Assumed to run as a subworkflow of .github/workflows/release.yml; specifically, as a post-announce
+# job within `cargo-dist`.
 name: mkdocs
 
 on:
@@ -7,8 +11,11 @@ on:
         description: "The commit SHA, tag, or branch to publish. Uses the default branch if not specified."
         default: ""
         type: string
-  release:
-    types: [published]
+  workflow_call:
+    inputs:
+      plan:
+        required: true
+        type: string
 
 jobs:
   mkdocs:
@@ -47,7 +54,7 @@ jobs:
         run: mkdocs build --strict -f mkdocs.public.yml
       - name: "Deploy to Cloudflare Pages"
         if: ${{ env.CF_API_TOKEN_EXISTS == 'true' }}
-        uses: cloudflare/wrangler-action@v3.4.1
+        uses: cloudflare/wrangler-action@v3.6.1
         with:
           apiToken: ${{ secrets.CF_API_TOKEN }}
           accountId: ${{ secrets.CF_ACCOUNT_ID }}

.github/workflows/publish-playground.yaml

@@ -1,9 +1,16 @@
+# Publish the Ruff playground.
+#
+# Assumed to run as a subworkflow of .github/workflows/release.yml; specifically, as a post-announce
+# job within `cargo-dist`.
 name: "[Playground] Release"
 
 on:
   workflow_dispatch:
-  release:
-    types: [published]
+  workflow_call:
+    inputs:
+      plan:
+        required: true
+        type: string
 
 env:
   CARGO_INCREMENTAL: 0
@@ -40,7 +47,7 @@ jobs:
         working-directory: playground
       - name: "Deploy to Cloudflare Pages"
         if: ${{ env.CF_API_TOKEN_EXISTS == 'true' }}
-        uses: cloudflare/wrangler-action@v3.4.1
+        uses: cloudflare/wrangler-action@v3.6.1
         with:
           apiToken: ${{ secrets.CF_API_TOKEN }}
           accountId: ${{ secrets.CF_ACCOUNT_ID }}

.github/workflows/publish-pypi.yml

@@ -0,0 +1,34 @@
+# Publish a release to PyPI.
+#
+# Assumed to run as a subworkflow of .github/workflows/release.yml; specifically, as a publish job
+# within `cargo-dist`.
+name: "[ruff] Publish to PyPI"
+
+on:
+  workflow_call:
+    inputs:
+      plan:
+        required: true
+        type: string
+
+jobs:
+  pypi-publish:
+    name: Upload to PyPI
+    runs-on: ubuntu-latest
+    environment:
+      name: release
+    permissions:
+      # For PyPI's trusted publishing.
+      id-token: write
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          pattern: wheels-*
+          path: wheels
+          merge-multiple: true
+      - name: Publish to PyPi
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          skip-existing: true
+          packages-dir: wheels
+          verbose: true

.github/workflows/release.yml

@@ -0,0 +1,267 @@
+# Copyright 2022-2024, axodotdev
+# SPDX-License-Identifier: MIT or Apache-2.0
+#
+# CI that:
+#
+# * checks for a Git Tag that looks like a release
+# * builds artifacts with cargo-dist (archives, installers, hashes)
+# * uploads those artifacts to temporary workflow zip
+# * on success, uploads the artifacts to a GitHub Release
+#
+# Note that the GitHub Release will be created with a generated
+# title/body based on your changelogs.
+
+name: Release
+
+permissions:
+  contents: write
+
+# This task will run whenever you workflow_dispatch with a tag that looks like a version
+# like "1.0.0", "v0.1.0-prerelease.1", "my-app/0.1.0", "releases/v1.0.0", etc.
+# Various formats will be parsed into a VERSION and an optional PACKAGE_NAME, where
+# PACKAGE_NAME must be the name of a Cargo package in your workspace, and VERSION
+# must be a Cargo-style SemVer Version (must have at least major.minor.patch).
+#
+# If PACKAGE_NAME is specified, then the announcement will be for that
+# package (erroring out if it doesn't have the given version or isn't cargo-dist-able).
+#
+# If PACKAGE_NAME isn't specified, then the announcement will be for all
+# (cargo-dist-able) packages in the workspace with that version (this mode is
+# intended for workspaces with only one dist-able package, or with all dist-able
+# packages versioned/released in lockstep).
+#
+# If you push multiple tags at once, separate instances of this workflow will
+# spin up, creating an independent announcement for each one. However, GitHub
+# will hard limit this to 3 tags per commit, as it will assume more tags is a
+# mistake.
+#
+# If there's a prerelease-style suffix to the version, then the release(s)
+# will be marked as a prerelease.
+on:
+  workflow_dispatch:
+    inputs:
+      tag:
+        description: Release Tag
+        required: true
+        default: dry-run
+        type: string
+
+jobs:
+  # Run 'cargo dist plan' (or host) to determine what tasks we need to do
+  plan:
+    runs-on: ubuntu-latest
+    outputs:
+      val: ${{ steps.plan.outputs.manifest }}
+      tag: ${{ (inputs.tag != 'dry-run' && inputs.tag) || '' }}
+      tag-flag: ${{ inputs.tag && inputs.tag != 'dry-run' && format('--tag={0}', inputs.tag) || '' }}
+      publishing: ${{ inputs.tag && inputs.tag != 'dry-run' }}
+    env:
+      GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+      - name: Install cargo-dist
+        # we specify bash to get pipefail; it guards against the `curl` command
+        # failing. otherwise `sh` won't catch that `curl` returned non-0
+        shell: bash
+        run: "curl --proto '=https' --tlsv1.2 -LsSf https://github.com/axodotdev/cargo-dist/releases/download/v0.14.0/cargo-dist-installer.sh | sh"
+      # sure would be cool if github gave us proper conditionals...
+      # so here's a doubly-nested ternary-via-truthiness to try to provide the best possible
+      # functionality based on whether this is a pull_request, and whether it's from a fork.
+      # (PRs run on the *source* but secrets are usually on the *target* -- that's *good*
+      # but also really annoying to build CI around when it needs secrets to work right.)
+      - id: plan
+        run: |
+          cargo dist ${{ (inputs.tag && inputs.tag != 'dry-run' && format('host --steps=create --tag={0}', inputs.tag)) || 'plan' }} --output-format=json > plan-dist-manifest.json
+          echo "cargo dist ran successfully"
+          cat plan-dist-manifest.json
+          echo "manifest=$(jq -c "." plan-dist-manifest.json)" >> "$GITHUB_OUTPUT"
+      - name: "Upload dist-manifest.json"
+        uses: actions/upload-artifact@v4
+        with:
+          name: artifacts-plan-dist-manifest
+          path: plan-dist-manifest.json
+
+  custom-build-binaries:
+    needs:
+      - plan
+    if: ${{ needs.plan.outputs.publishing == 'true' || fromJson(needs.plan.outputs.val).ci.github.pr_run_mode == 'upload' || inputs.tag == 'dry-run' }}
+    uses: ./.github/workflows/build-binaries.yml
+    with:
+      plan: ${{ needs.plan.outputs.val }}
+    secrets: inherit
+
+  custom-build-docker:
+    needs:
+      - plan
+    if: ${{ needs.plan.outputs.publishing == 'true' || fromJson(needs.plan.outputs.val).ci.github.pr_run_mode == 'upload' || inputs.tag == 'dry-run' }}
+    uses: ./.github/workflows/build-docker.yml
+    with:
+      plan: ${{ needs.plan.outputs.val }}
+    secrets: inherit
+    permissions:
+      packages: write
+      contents: read
+
+  # Build and package all the platform-agnostic(ish) things
+  build-global-artifacts:
+    needs:
+      - plan
+      - custom-build-binaries
+      - custom-build-docker
+    runs-on: "ubuntu-20.04"
+    env:
+      GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      BUILD_MANIFEST_NAME: target/distrib/global-dist-manifest.json
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+      - name: Install cargo-dist
+        shell: bash
+        run: "curl --proto '=https' --tlsv1.2 -LsSf https://github.com/axodotdev/cargo-dist/releases/download/v0.14.0/cargo-dist-installer.sh | sh"
+      # Get all the local artifacts for the global tasks to use (for e.g. checksums)
+      - name: Fetch local artifacts
+        uses: actions/download-artifact@v4
+        with:
+          pattern: artifacts-*
+          path: target/distrib/
+          merge-multiple: true
+      - id: cargo-dist
+        shell: bash
+        run: |
+          cargo dist build ${{ needs.plan.outputs.tag-flag }} --output-format=json "--artifacts=global" > dist-manifest.json
+          echo "cargo dist ran successfully"
+
+          # Parse out what we just built and upload it to scratch storage
+          echo "paths<<EOF" >> "$GITHUB_OUTPUT"
+          jq --raw-output ".upload_files[]" dist-manifest.json >> "$GITHUB_OUTPUT"
+          echo "EOF" >> "$GITHUB_OUTPUT"
+
+          cp dist-manifest.json "$BUILD_MANIFEST_NAME"
+      - name: "Upload artifacts"
+        uses: actions/upload-artifact@v4
+        with:
+          name: artifacts-build-global
+          path: |
+            ${{ steps.cargo-dist.outputs.paths }}
+            ${{ env.BUILD_MANIFEST_NAME }}
+  # Determines if we should publish/announce
+  host:
+    needs:
+      - plan
+      - custom-build-binaries
+      - custom-build-docker
+      - build-global-artifacts
+    # Only run if we're "publishing", and only if local and global didn't fail (skipped is fine)
+    if: ${{ always() && needs.plan.outputs.publishing == 'true' && (needs.build-global-artifacts.result == 'skipped' || needs.build-global-artifacts.result == 'success') && (needs.custom-build-binaries.result == 'skipped' || needs.custom-build-binaries.result == 'success') && (needs.custom-build-docker.result == 'skipped' || needs.custom-build-docker.result == 'success') }}
+    env:
+      GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+    runs-on: "ubuntu-20.04"
+    outputs:
+      val: ${{ steps.host.outputs.manifest }}
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+      - name: Install cargo-dist
+        run: "curl --proto '=https' --tlsv1.2 -LsSf https://github.com/axodotdev/cargo-dist/releases/download/v0.14.0/cargo-dist-installer.sh | sh"
+      # Fetch artifacts from scratch-storage
+      - name: Fetch artifacts
+        uses: actions/download-artifact@v4
+        with:
+          pattern: artifacts-*
+          path: target/distrib/
+          merge-multiple: true
+      # This is a harmless no-op for GitHub Releases, hosting for that happens in "announce"
+      - id: host
+        shell: bash
+        run: |
+          cargo dist host ${{ needs.plan.outputs.tag-flag }} --steps=upload --steps=release --output-format=json > dist-manifest.json
+          echo "artifacts uploaded and released successfully"
+          cat dist-manifest.json
+          echo "manifest=$(jq -c "." dist-manifest.json)" >> "$GITHUB_OUTPUT"
+      - name: "Upload dist-manifest.json"
+        uses: actions/upload-artifact@v4
+        with:
+          # Overwrite the previous copy
+          name: artifacts-dist-manifest
+          path: dist-manifest.json
+
+  custom-publish-pypi:
+    needs:
+      - plan
+      - host
+    if: ${{ !fromJson(needs.plan.outputs.val).announcement_is_prerelease || fromJson(needs.plan.outputs.val).publish_prereleases }}
+    uses: ./.github/workflows/publish-pypi.yml
+    with:
+      plan: ${{ needs.plan.outputs.val }}
+    secrets: inherit
+    # publish jobs get escalated permissions
+    permissions:
+      id-token: write
+      packages: write
+
+  # Create a GitHub Release while uploading all files to it
+  announce:
+    needs:
+      - plan
+      - host
+      - custom-publish-pypi
+    # use "always() && ..." to allow us to wait for all publish jobs while
+    # still allowing individual publish jobs to skip themselves (for prereleases).
+    # "host" however must run to completion, no skipping allowed!
+    if: ${{ always() && needs.host.result == 'success' && (needs.custom-publish-pypi.result == 'skipped' || needs.custom-publish-pypi.result == 'success') }}
+    runs-on: "ubuntu-20.04"
+    env:
+      GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+      - name: "Download GitHub Artifacts"
+        uses: actions/download-artifact@v4
+        with:
+          pattern: artifacts-*
+          path: artifacts
+          merge-multiple: true
+      - name: Cleanup
+        run: |
+          # Remove the granular manifests
+          rm -f artifacts/*-dist-manifest.json
+      - name: Create GitHub Release
+        uses: ncipollo/release-action@v1
+        with:
+          tag: ${{ needs.plan.outputs.tag }}
+          name: ${{ fromJson(needs.host.outputs.val).announcement_title }}
+          body: ${{ fromJson(needs.host.outputs.val).announcement_github_body }}
+          prerelease: ${{ fromJson(needs.host.outputs.val).announcement_is_prerelease }}
+          artifacts: "artifacts/*"
+
+  custom-notify-dependents:
+    needs:
+      - plan
+      - announce
+    uses: ./.github/workflows/notify-dependents.yml
+    with:
+      plan: ${{ needs.plan.outputs.val }}
+    secrets: inherit
+
+  custom-publish-docs:
+    needs:
+      - plan
+      - announce
+    uses: ./.github/workflows/publish-docs.yml
+    with:
+      plan: ${{ needs.plan.outputs.val }}
+    secrets: inherit
+
+  custom-publish-playground:
+    needs:
+      - plan
+      - announce
+    uses: ./.github/workflows/publish-playground.yml
+    with:
+      plan: ${{ needs.plan.outputs.val }}
+    secrets: inherit

.github/workflows/sync_typeshed.yaml

@@ -0,0 +1,80 @@
+name: Sync typeshed
+
+on:
+  workflow_dispatch:
+  schedule:
+    # Run on the 1st and the 15th of every month:
+    - cron: "0 0 1,15 * *"
+
+env:
+  FORCE_COLOR: 1
+  GH_TOKEN: ${{ github.token }}
+
+jobs:
+  sync:
+    name: Sync typeshed
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    # Don't run the cron job on forks:
+    if: ${{ github.repository == 'astral-sh/ruff' || github.event_name != 'schedule' }}
+    permissions:
+      contents: write
+      pull-requests: write
+    steps:
+      - uses: actions/checkout@v4
+        name: Checkout Ruff
+        with:
+          path: ruff
+      - uses: actions/checkout@v4
+        name: Checkout typeshed
+        with:
+          repository: python/typeshed
+          path: typeshed
+      - name: Setup git
+        run: |
+          git config --global user.name typeshedbot
+          git config --global user.email '<>'
+      - name: Sync typeshed
+        id: sync
+        run: |
+          rm -rf ruff/crates/red_knot_module_resolver/vendor/typeshed
+          mkdir ruff/crates/red_knot_module_resolver/vendor/typeshed
+          cp typeshed/README.md ruff/crates/red_knot_module_resolver/vendor/typeshed
+          cp typeshed/LICENSE ruff/crates/red_knot_module_resolver/vendor/typeshed
+          cp -r typeshed/stdlib ruff/crates/red_knot_module_resolver/vendor/typeshed/stdlib
+          rm -rf ruff/crates/red_knot_module_resolver/vendor/typeshed/stdlib/@tests
+          git -C typeshed rev-parse HEAD > ruff/crates/red_knot_module_resolver/vendor/typeshed/source_commit.txt
+      - name: Commit the changes
+        id: commit
+        if: ${{ steps.sync.outcome == 'success' }}
+        run: |
+          cd ruff
+          git checkout -b typeshedbot/sync-typeshed
+          git add .
+          git diff --staged --quiet || git commit -m "Sync typeshed. Source commit: https://github.com/python/typeshed/commit/$(git -C ../typeshed rev-parse HEAD)"
+      - name: Create a PR
+        if: ${{ steps.sync.outcome == 'success' && steps.commit.outcome == 'success' }}
+        run: |
+          cd ruff
+          git push --force origin typeshedbot/sync-typeshed
+          gh pr list --repo $GITHUB_REPOSITORY --head typeshedbot/sync-typeshed --json id --jq length | grep 1 && exit 0 # exit if there is existing pr
+          gh pr create --title "Sync vendored typeshed stubs" --body "Close and reopen this PR to trigger CI" --label "internal"
+
+  create-issue-on-failure:
+    name: Create an issue if the typeshed sync failed
+    runs-on: ubuntu-latest
+    needs: [sync]
+    if: ${{ github.repository == 'astral-sh/ruff' && always() && github.event_name == 'schedule' && needs.sync.result == 'failure' }}
+    permissions:
+      issues: write
+    steps:
+      - uses: actions/github-script@v7
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          script: |
+            await github.rest.issues.create({
+              owner: "astral-sh",
+              repo: "ruff",
+              title: `Automated typeshed sync failed on ${new Date().toDateString()}`,
+              body: "Runs are listed here: https://github.com/astral-sh/ruff/actions/workflows/sync_typeshed.yaml",
+            })

.pre-commit-config.yaml

@@ -2,6 +2,7 @@ fail_fast: true
 
 exclude: |
   (?x)^(
+    crates/red_knot_module_resolver/vendor/.*|
     crates/ruff_linter/resources/.*|
     crates/ruff_linter/src/rules/.*/snapshots/.*|
     crates/ruff/resources/.*|
@@ -13,7 +14,7 @@ exclude: |
 
 repos:
   - repo: https://github.com/abravalheri/validate-pyproject
-    rev: v0.16
+    rev: v0.18
     hooks:
       - id: validate-pyproject
 
@@ -31,7 +32,7 @@ repos:
           )$
 
   - repo: https://github.com/igorshubovych/markdownlint-cli
-    rev: v0.39.0
+    rev: v0.41.0
     hooks:
       - id: markdownlint-fix
         exclude: |
@@ -41,7 +42,7 @@ repos:
           )$
 
   - repo: https://github.com/crate-ci/typos
-    rev: v1.20.9
+    rev: v1.22.9
     hooks:
       - id: typos
 
@@ -55,7 +56,7 @@ repos:
         pass_filenames: false # This makes it a lot faster
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.4.1
+    rev: v0.4.10
     hooks:
       - id: ruff-format
       - id: ruff

.prettierignore

@@ -0,0 +1,2 @@
+# Auto-generated by `cargo-dist`.
+.github/workflows/release.yml

.vscode/extensions.json

@@ -0,0 +1,5 @@
+{
+    "recommendations": [
+        "rust-lang.rust-analyzer"
+    ]
+}

.vscode/settings.json

@@ -0,0 +1,6 @@
+{
+    "rust-analyzer.check.extraArgs": [
+        "--all-features"
+    ],
+    "rust-analyzer.check.command": "clippy",
+}

BREAKING_CHANGES.md

@@ -1,5 +1,12 @@
 # Breaking Changes
 
+## 0.5.0
+
+- Follow the XDG specification to discover user-level configurations on macOS (same as on other Unix platforms)
+- Selecting `ALL` now excludes deprecated rules
+- The released archives now include an extra level of nesting, which can be removed with `--strip-components=1` when untarring.
+- The release artifact's file name no longer includes the version tag. This enables users to install via `/latest` URLs on GitHub.
+
 ## 0.3.0
 
 ### Ruff 2024.2 style

CHANGELOG.md

@@ -1,5 +1,432 @@
 # Changelog
 
+## 0.5.0
+
+Check out the [blog post](https://astral.sh/blog/ruff-v0.5.0) for a migration guide and overview of the changes!
+
+### Breaking changes
+
+See also, the "Remapped rules" section which may result in disabled rules.
+
+- Follow the XDG specification to discover user-level configurations on macOS (same as on other Unix platforms)
+- Selecting `ALL` now excludes deprecated rules
+- The released archives now include an extra level of nesting, which can be removed with `--strip-components=1` when untarring.
+- The release artifact's file name no longer includes the version tag. This enables users to install via `/latest` URLs on GitHub.
+- The diagnostic ranges for some `flake8-bandit` rules were modified ([#10667](https://github.com/astral-sh/ruff/pull/10667)).
+
+### Deprecations
+
+The following rules are now deprecated:
+
+- [`syntax-error`](https://docs.astral.sh/ruff/rules/syntax-error/) (`E999`): Syntax errors are now always shown
+
+### Remapped rules
+
+The following rules have been remapped to new rule codes:
+
+- [`blocking-http-call-in-async-function`](https://docs.astral.sh/ruff/rules/blocking-http-call-in-async-function/): `ASYNC100` to `ASYNC210`
+- [`open-sleep-or-subprocess-in-async-function`](https://docs.astral.sh/ruff/rules/open-sleep-or-subprocess-in-async-function/): `ASYNC101` split into `ASYNC220`, `ASYNC221`, `ASYNC230`, and `ASYNC251`
+- [`blocking-os-call-in-async-function`](https://docs.astral.sh/ruff/rules/blocking-os-call-in-async-function/): `ASYNC102` has been merged into `ASYNC220` and `ASYNC221`
+- [`trio-timeout-without-await`](https://docs.astral.sh/ruff/rules/trio-timeout-without-await/): `TRIO100` to `ASYNC100`
+- [`trio-sync-call`](https://docs.astral.sh/ruff/rules/trio-sync-call/): `TRIO105` to `ASYNC105`
+- [`trio-async-function-with-timeout`](https://docs.astral.sh/ruff/rules/trio-async-function-with-timeout/): `TRIO109` to `ASYNC109`
+- [`trio-unneeded-sleep`](https://docs.astral.sh/ruff/rules/trio-unneeded-sleep/): `TRIO110` to `ASYNC110`
+- [`trio-zero-sleep-call`](https://docs.astral.sh/ruff/rules/trio-zero-sleep-call/): `TRIO115` to `ASYNC115`
+- [`repeated-isinstance-calls`](https://docs.astral.sh/ruff/rules/repeated-isinstance-calls/): `PLR1701` to `SIM101`
+
+### Stabilization
+
+The following rules have been stabilized and are no longer in preview:
+
+- [`mutable-fromkeys-value`](https://docs.astral.sh/ruff/rules/mutable-fromkeys-value/) (`RUF024`)
+- [`default-factory-kwarg`](https://docs.astral.sh/ruff/rules/default-factory-kwarg/) (`RUF026`)
+- [`django-extra`](https://docs.astral.sh/ruff/rules/django-extra/) (`S610`)
+- [`manual-dict-comprehension`](https://docs.astral.sh/ruff/rules/manual-dict-comprehension/) (`PERF403`)
+- [`print-empty-string`](https://docs.astral.sh/ruff/rules/print-empty-string/) (`FURB105`)
+- [`readlines-in-for`](https://docs.astral.sh/ruff/rules/readlines-in-for/) (`FURB129`)
+- [`if-expr-min-max`](https://docs.astral.sh/ruff/rules/if-expr-min-max/) (`FURB136`)
+- [`bit-count`](https://docs.astral.sh/ruff/rules/bit-count/) (`FURB161`)
+- [`redundant-log-base`](https://docs.astral.sh/ruff/rules/redundant-log-base/) (`FURB163`)
+- [`regex-flag-alias`](https://docs.astral.sh/ruff/rules/regex-flag-alias/) (`FURB167`)
+- [`isinstance-type-none`](https://docs.astral.sh/ruff/rules/isinstance-type-none/) (`FURB168`)
+- [`type-none-comparison`](https://docs.astral.sh/ruff/rules/type-none-comparison/) (`FURB169`)
+- [`implicit-cwd`](https://docs.astral.sh/ruff/rules/implicit-cwd/) (`FURB177`)
+- [`hashlib-digest-hex`](https://docs.astral.sh/ruff/rules/hashlib-digest-hex/) (`FURB181`)
+- [`list-reverse-copy`](https://docs.astral.sh/ruff/rules/list-reverse-copy/) (`FURB187`)
+- [`bad-open-mode`](https://docs.astral.sh/ruff/rules/bad-open-mode/) (`PLW1501`)
+- [`empty-comment`](https://docs.astral.sh/ruff/rules/empty-comment/) (`PLR2044`)
+- [`global-at-module-level`](https://docs.astral.sh/ruff/rules/global-at-module-level/) (`PLW0604`)
+- [`misplaced-bare-raise`](https://docs.astral.sh/ruff/rules/misplaced-bare-raise%60/) (`PLE0744`)
+- [`non-ascii-import-name`](https://docs.astral.sh/ruff/rules/non-ascii-import-name/) (`PLC2403`)
+- [`non-ascii-name`](https://docs.astral.sh/ruff/rules/non-ascii-name/) (`PLC2401`)
+- [`nonlocal-and-global`](https://docs.astral.sh/ruff/rules/nonlocal-and-global/) (`PLE0115`)
+- [`potential-index-error`](https://docs.astral.sh/ruff/rules/potential-index-error/) (`PLE0643`)
+- [`redeclared-assigned-name`](https://docs.astral.sh/ruff/rules/redeclared-assigned-name/) (`PLW0128`)
+- [`redefined-argument-from-local`](https://docs.astral.sh/ruff/rules/redefined-argument-from-local/) (`PLR1704`)
+- [`repeated-keyword-argument`](https://docs.astral.sh/ruff/rules/repeated-keyword-argument/) (`PLE1132`)
+- [`super-without-brackets`](https://docs.astral.sh/ruff/rules/super-without-brackets/) (`PLW0245`)
+- [`unnecessary-list-index-lookup`](https://docs.astral.sh/ruff/rules/unnecessary-list-index-lookup/) (`PLR1736`)
+- [`useless-exception-statement`](https://docs.astral.sh/ruff/rules/useless-exception-statement/) (`PLW0133`)
+- [`useless-with-lock`](https://docs.astral.sh/ruff/rules/useless-with-lock/) (`PLW2101`)
+
+The following behaviors have been stabilized:
+
+- [`is-literal`](https://docs.astral.sh/ruff/rules/is-literal/) (`F632`) now warns for identity checks against list, set or dictionary literals
+- [`needless-bool`](https://docs.astral.sh/ruff/rules/needless-bool/) (`SIM103`) now detects `if` expressions with implicit `else` branches
+- [`module-import-not-at-top-of-file`](https://docs.astral.sh/ruff/rules/module-import-not-at-top-of-file/) (`E402`) now allows `os.environ` modifications between import statements
+- [`type-comparison`](https://docs.astral.sh/ruff/rules/type-comparison/) (`E721`) now allows idioms such as `type(x) is int`
+- [`yoda-condition`](https://docs.astral.sh/ruff/rules/yoda-conditions/) (`SIM300`) now flags a wider range of expressions
+
+### Removals
+
+The following deprecated settings have been removed:
+
+- `output-format=text`; use `output-format=concise` or `output-format=full`
+- `tab-size`; use `indent-width`
+
+The following deprecated CLI options have been removed:
+
+- `--show-source`; use `--output-format=full`
+- `--no-show-source`; use `--output-format=concise`
+
+The following deprecated CLI commands have been removed:
+
+- `ruff <path>`; use `ruff check <path>`
+- `ruff --clean`; use `ruff clean`
+- `ruff --generate-shell-completion`; use `ruff generate-shell-completion`
+
+### Preview features
+
+- \[`ruff`\] Add `assert-with-print-message` rule ([#11981](https://github.com/astral-sh/ruff/pull/11981))
+
+### CLI
+
+- Use rule name rather than message in `--statistics` ([#11697](https://github.com/astral-sh/ruff/pull/11697))
+- Use the output format `full` by default ([#12010](https://github.com/astral-sh/ruff/pull/12010))
+- Don't log syntax errors to the console ([#11902](https://github.com/astral-sh/ruff/pull/11902))
+
+### Rule changes
+
+- \[`ruff`\] Fix false positives if `gettext` is imported using an alias (`RUF027`) ([#12025](https://github.com/astral-sh/ruff/pull/12025))
+- \[`numpy`\] Update `trapz` and `in1d` deprecation (`NPY201`) ([#11948](https://github.com/astral-sh/ruff/pull/11948))
+- \[`flake8-bandit`\] Modify diagnostic ranges for shell-related rules ([#10667](https://github.com/astral-sh/ruff/pull/10667))
+
+### Server
+
+- Closing an untitled, unsaved notebook document no longer throws an error ([#11942](https://github.com/astral-sh/ruff/pull/11942))
+- Support the usage of tildes and environment variables in `logFile` ([#11945](https://github.com/astral-sh/ruff/pull/11945))
+- Add option to configure whether to show syntax errors ([#12059](https://github.com/astral-sh/ruff/pull/12059))
+
+### Bug fixes
+
+- \[`pycodestyle`\] Avoid `E203` for f-string debug expression ([#12024](https://github.com/astral-sh/ruff/pull/12024))
+- \[`pep8-naming`\] Match import-name ignores against both name and alias (`N812`, `N817`) ([#12033](https://github.com/astral-sh/ruff/pull/12033))
+- \[`pyflakes`\] Detect assignments that shadow definitions (`F811`) ([#11961](https://github.com/astral-sh/ruff/pull/11961))
+
+### Parser
+
+- Emit a syntax error for an empty type parameter list ([#12030](https://github.com/astral-sh/ruff/pull/12030))
+- Avoid consuming the newline for unterminated strings ([#12067](https://github.com/astral-sh/ruff/pull/12067))
+- Do not include the newline in the unterminated string range ([#12017](https://github.com/astral-sh/ruff/pull/12017))
+- Use the correct range to highlight line continuation errors ([#12016](https://github.com/astral-sh/ruff/pull/12016))
+- Consider 2-character EOL before line continuations ([#12035](https://github.com/astral-sh/ruff/pull/12035))
+- Consider line continuation character for re-lexing ([#12008](https://github.com/astral-sh/ruff/pull/12008))
+
+### Other changes
+
+- Upgrade the Unicode table used for measuring the line-length ([#11194](https://github.com/astral-sh/ruff/pull/11194))
+- Remove the deprecation error message for the nursery selector ([#10172](https://github.com/astral-sh/ruff/pull/10172))
+
+## 0.4.10
+
+### Parser
+
+- Implement re-lexing logic for better error recovery ([#11845](https://github.com/astral-sh/ruff/pull/11845))
+
+### Rule changes
+
+- \[`flake8-copyright`\] Update `CPY001` to check the first 4096 bytes instead of 1024 ([#11927](https://github.com/astral-sh/ruff/pull/11927))
+- \[`pycodestyle`\] Update `E999` to show all syntax errors instead of just the first one ([#11900](https://github.com/astral-sh/ruff/pull/11900))
+
+### Server
+
+- Add tracing setup guide to Helix documentation ([#11883](https://github.com/astral-sh/ruff/pull/11883))
+- Add tracing setup guide to Neovim documentation ([#11884](https://github.com/astral-sh/ruff/pull/11884))
+- Defer notebook cell deletion to avoid an error message ([#11864](https://github.com/astral-sh/ruff/pull/11864))
+
+### Security
+
+- Guard against malicious ecosystem comment artifacts ([#11879](https://github.com/astral-sh/ruff/pull/11879))
+
+## 0.4.9
+
+### Preview features
+
+- \[`pylint`\] Implement `consider-dict-items` (`C0206`) ([#11688](https://github.com/astral-sh/ruff/pull/11688))
+- \[`refurb`\] Implement `repeated-global` (`FURB154`) ([#11187](https://github.com/astral-sh/ruff/pull/11187))
+
+### Rule changes
+
+- \[`pycodestyle`\] Adapt fix for `E203` to work identical to `ruff format` ([#10999](https://github.com/astral-sh/ruff/pull/10999))
+
+### Formatter
+
+- Fix formatter instability for lines only consisting of zero-width characters ([#11748](https://github.com/astral-sh/ruff/pull/11748))
+
+### Server
+
+- Add supported commands in server capabilities ([#11850](https://github.com/astral-sh/ruff/pull/11850))
+- Use real file path when available in `ruff server` ([#11800](https://github.com/astral-sh/ruff/pull/11800))
+- Improve error message when a command is run on an unavailable document ([#11823](https://github.com/astral-sh/ruff/pull/11823))
+- Introduce the `ruff.printDebugInformation` command ([#11831](https://github.com/astral-sh/ruff/pull/11831))
+- Tracing system now respects log level and trace level, with options to log to a file ([#11747](https://github.com/astral-sh/ruff/pull/11747))
+
+### CLI
+
+- Handle non-printable characters in diff view ([#11687](https://github.com/astral-sh/ruff/pull/11687))
+
+### Bug fixes
+
+- \[`refurb`\] Avoid suggesting starmap when arguments are used outside call (`FURB140`) ([#11830](https://github.com/astral-sh/ruff/pull/11830))
+- \[`flake8-bugbear`\] Avoid panic in `B909` when checking large loop blocks ([#11772](https://github.com/astral-sh/ruff/pull/11772))
+- \[`refurb`\] Fix misbehavior of `operator.itemgetter` when getter param is a tuple (`FURB118`) ([#11774](https://github.com/astral-sh/ruff/pull/11774))
+
+## 0.4.8
+
+### Performance
+
+- Linter performance has been improved by around 10% on some microbenchmarks by refactoring the lexer and parser to maintain synchronicity between them ([#11457](https://github.com/astral-sh/ruff/pull/11457))
+
+### Preview features
+
+- \[`flake8-bugbear`\] Implement `return-in-generator` (`B901`) ([#11644](https://github.com/astral-sh/ruff/pull/11644))
+- \[`flake8-pyi`\] Implement `PYI063` ([#11699](https://github.com/astral-sh/ruff/pull/11699))
+- \[`pygrep_hooks`\] Check blanket ignores via file-level pragmas (`PGH004`) ([#11540](https://github.com/astral-sh/ruff/pull/11540))
+
+### Rule changes
+
+- \[`pyupgrade`\] Update `UP035` for Python 3.13 and the latest version of `typing_extensions` ([#11693](https://github.com/astral-sh/ruff/pull/11693))
+- \[`numpy`\] Update `NPY001` rule for NumPy 2.0 ([#11735](https://github.com/astral-sh/ruff/pull/11735))
+
+### Server
+
+- Formatting a document with syntax problems no longer spams a visible error popup ([#11745](https://github.com/astral-sh/ruff/pull/11745))
+
+### CLI
+
+- Add RDJson support for `--output-format` flag ([#11682](https://github.com/astral-sh/ruff/pull/11682))
+
+### Bug fixes
+
+- \[`pyupgrade`\] Write empty string in lieu of panic when fixing `UP032` ([#11696](https://github.com/astral-sh/ruff/pull/11696))
+- \[`flake8-simplify`\] Simplify double negatives in `SIM103` ([#11684](https://github.com/astral-sh/ruff/pull/11684))
+- Ensure the expression generator adds a newline before `type` statements ([#11720](https://github.com/astral-sh/ruff/pull/11720))
+- Respect per-file ignores for blanket and redirected noqa rules ([#11728](https://github.com/astral-sh/ruff/pull/11728))
+
+## 0.4.7
+
+### Preview features
+
+- \[`flake8-pyi`\] Implement `PYI064` ([#11325](https://github.com/astral-sh/ruff/pull/11325))
+- \[`flake8-pyi`\] Implement `PYI066` ([#11541](https://github.com/astral-sh/ruff/pull/11541))
+- \[`flake8-pyi`\] Implement `PYI057` ([#11486](https://github.com/astral-sh/ruff/pull/11486))
+- \[`pyflakes`\] Enable `F822` in `__init__.py` files by default ([#11370](https://github.com/astral-sh/ruff/pull/11370))
+
+### Formatter
+
+- Fix incorrect placement of trailing stub function comments ([#11632](https://github.com/astral-sh/ruff/pull/11632))
+
+### Server
+
+- Respect file exclusions in `ruff server` ([#11590](https://github.com/astral-sh/ruff/pull/11590))
+- Add support for documents not exist on disk ([#11588](https://github.com/astral-sh/ruff/pull/11588))
+- Add Vim and Kate setup guide for `ruff server` ([#11615](https://github.com/astral-sh/ruff/pull/11615))
+
+### Bug fixes
+
+- Avoid removing newlines between docstring headers and rST blocks ([#11609](https://github.com/astral-sh/ruff/pull/11609))
+- Infer indentation with imports when logical indent is absent ([#11608](https://github.com/astral-sh/ruff/pull/11608))
+- Use char index rather than position for indent slice ([#11645](https://github.com/astral-sh/ruff/pull/11645))
+- \[`flake8-comprehension`\] Strip parentheses around generators in `C400` ([#11607](https://github.com/astral-sh/ruff/pull/11607))
+- Mark `repeated-isinstance-calls` as unsafe on Python 3.10 and later ([#11622](https://github.com/astral-sh/ruff/pull/11622))
+
+## 0.4.6
+
+### Breaking changes
+
+- Use project-relative paths when calculating GitLab fingerprints ([#11532](https://github.com/astral-sh/ruff/pull/11532))
+- Bump minimum supported Windows version to Windows 10 ([#11613](https://github.com/astral-sh/ruff/pull/11613))
+
+### Preview features
+
+- \[`flake8-async`\] Sleep with >24 hour interval should usually sleep forever (`ASYNC116`) ([#11498](https://github.com/astral-sh/ruff/pull/11498))
+
+### Rule changes
+
+- \[`numpy`\] Add missing functions to NumPy 2.0 migration rule ([#11528](https://github.com/astral-sh/ruff/pull/11528))
+- \[`mccabe`\] Consider irrefutable pattern similar to `if .. else` for `C901` ([#11565](https://github.com/astral-sh/ruff/pull/11565))
+- Consider `match`-`case` statements for `C901`, `PLR0912`, and `PLR0915` ([#11521](https://github.com/astral-sh/ruff/pull/11521))
+- Remove empty strings when converting to f-string (`UP032`) ([#11524](https://github.com/astral-sh/ruff/pull/11524))
+- \[`flake8-bandit`\] `request-without-timeout` should warn for `requests.request` ([#11548](https://github.com/astral-sh/ruff/pull/11548))
+- \[`flake8-self`\] Ignore sunder accesses in `flake8-self` rules ([#11546](https://github.com/astral-sh/ruff/pull/11546))
+- \[`pyupgrade`\] Lint for `TypeAliasType` usages (`UP040`) ([#11530](https://github.com/astral-sh/ruff/pull/11530))
+
+### Server
+
+- Respect excludes in `ruff server` configuration discovery ([#11551](https://github.com/astral-sh/ruff/pull/11551))
+- Use default settings if initialization options is empty or not provided ([#11566](https://github.com/astral-sh/ruff/pull/11566))
+- `ruff server` correctly treats `.pyi` files as stub files ([#11535](https://github.com/astral-sh/ruff/pull/11535))
+- `ruff server` searches for configuration in parent directories ([#11537](https://github.com/astral-sh/ruff/pull/11537))
+- `ruff server`: An empty code action filter no longer returns notebook source actions ([#11526](https://github.com/astral-sh/ruff/pull/11526))
+
+### Bug fixes
+
+- \[`flake8-logging-format`\] Fix autofix title in `logging-warn` (`G010`) ([#11514](https://github.com/astral-sh/ruff/pull/11514))
+- \[`refurb`\] Avoid recommending `operator.itemgetter` with dependence on lambda arguments ([#11574](https://github.com/astral-sh/ruff/pull/11574))
+- \[`flake8-simplify`\] Avoid recommending context manager in `__enter__` implementations ([#11575](https://github.com/astral-sh/ruff/pull/11575))
+- Create intermediary directories for `--output-file` ([#11550](https://github.com/astral-sh/ruff/pull/11550))
+- Propagate reads on global variables ([#11584](https://github.com/astral-sh/ruff/pull/11584))
+- Treat all `singledispatch` arguments as runtime-required ([#11523](https://github.com/astral-sh/ruff/pull/11523))
+
+## 0.4.5
+
+### Ruff's language server is now in Beta
+
+`v0.4.5` marks the official Beta release of `ruff server`, an integrated language server built into Ruff.
+`ruff server` supports the same feature set as `ruff-lsp`, powering linting, formatting, and
+code fixes in Ruff's editor integrations -- but with superior performance and
+no installation required. We'd love your feedback!
+
+You can enable `ruff server` in the [VS Code extension](https://github.com/astral-sh/ruff-vscode?tab=readme-ov-file#enabling-the-rust-based-language-server) today.
+
+To read more about this exciting milestone, check out our [blog post](https://astral.sh/blog/ruff-v0.4.5)!
+
+### Rule changes
+
+- \[`flake8-future-annotations`\] Reword `future-rewritable-type-annotation` (`FA100`) message ([#11381](https://github.com/astral-sh/ruff/pull/11381))
+- \[`isort`\] Expanded the set of standard-library modules to include `_string`, etc. ([#11374](https://github.com/astral-sh/ruff/pull/11374))
+- \[`pycodestyle`\] Consider soft keywords for `E27` rules ([#11446](https://github.com/astral-sh/ruff/pull/11446))
+- \[`pyflakes`\] Recommend adding unused import bindings to `__all__` ([#11314](https://github.com/astral-sh/ruff/pull/11314))
+- \[`pyflakes`\] Update documentation and deprecate `ignore_init_module_imports` ([#11436](https://github.com/astral-sh/ruff/pull/11436))
+- \[`pyupgrade`\] Mark quotes as unnecessary for non-evaluated annotations ([#11485](https://github.com/astral-sh/ruff/pull/11485))
+
+### Formatter
+
+- Avoid multiline quotes warning with `quote-style = preserve` ([#11490](https://github.com/astral-sh/ruff/pull/11490))
+
+### Server
+
+- Support Jupyter Notebook files ([#11206](https://github.com/astral-sh/ruff/pull/11206))
+- Support `noqa` comment code actions ([#11276](https://github.com/astral-sh/ruff/pull/11276))
+- Fix automatic configuration reloading ([#11492](https://github.com/astral-sh/ruff/pull/11492))
+- Fix several issues with configuration in Neovim and Helix ([#11497](https://github.com/astral-sh/ruff/pull/11497))
+
+### CLI
+
+- Add `--output-format` as a CLI option for `ruff config` ([#11438](https://github.com/astral-sh/ruff/pull/11438))
+
+### Bug fixes
+
+- Avoid `PLE0237` for property with setter ([#11377](https://github.com/astral-sh/ruff/pull/11377))
+- Avoid `TCH005` for `if` stmt with `elif`/`else` block ([#11376](https://github.com/astral-sh/ruff/pull/11376))
+- Avoid flagging `__future__` annotations as required for non-evaluated type annotations ([#11414](https://github.com/astral-sh/ruff/pull/11414))
+- Check for ruff executable in 'bin' directory as installed by 'pip install --target'. ([#11450](https://github.com/astral-sh/ruff/pull/11450))
+- Sort edits prior to deduplicating in quotation fix ([#11452](https://github.com/astral-sh/ruff/pull/11452))
+- Treat escaped newline as valid sequence ([#11465](https://github.com/astral-sh/ruff/pull/11465))
+- \[`flake8-pie`\] Preserve parentheses in `unnecessary-dict-kwargs` ([#11372](https://github.com/astral-sh/ruff/pull/11372))
+- \[`pylint`\] Ignore `__slots__` with dynamic values ([#11488](https://github.com/astral-sh/ruff/pull/11488))
+- \[`pylint`\] Remove `try` body from branch counting ([#11487](https://github.com/astral-sh/ruff/pull/11487))
+- \[`refurb`\] Respect operator precedence in `FURB110` ([#11464](https://github.com/astral-sh/ruff/pull/11464))
+
+### Documentation
+
+- Add `--preview` to the README ([#11395](https://github.com/astral-sh/ruff/pull/11395))
+- Add Python 3.13 to list of allowed Python versions ([#11411](https://github.com/astral-sh/ruff/pull/11411))
+- Simplify Neovim setup documentation ([#11489](https://github.com/astral-sh/ruff/pull/11489))
+- Update CONTRIBUTING.md to reflect the new parser ([#11434](https://github.com/astral-sh/ruff/pull/11434))
+- Update server documentation with new migration guide ([#11499](https://github.com/astral-sh/ruff/pull/11499))
+- \[`pycodestyle`\] Clarify motivation for `E713` and `E714` ([#11483](https://github.com/astral-sh/ruff/pull/11483))
+- \[`pyflakes`\] Update docs to describe WAI behavior (F541) ([#11362](https://github.com/astral-sh/ruff/pull/11362))
+- \[`pylint`\] Clearly indicate what is counted as a branch ([#11423](https://github.com/astral-sh/ruff/pull/11423))
+
+## 0.4.4
+
+### Preview features
+
+- \[`pycodestyle`\] Ignore end-of-line comments when determining blank line rules ([#11342](https://github.com/astral-sh/ruff/pull/11342))
+- \[`pylint`\] Detect `pathlib.Path.open` calls in `unspecified-encoding` (`PLW1514`)  ([#11288](https://github.com/astral-sh/ruff/pull/11288))
+- \[`flake8-pyi`\] Implement `PYI059` (`generic-not-last-base-class`) ([#11233](https://github.com/astral-sh/ruff/pull/11233))
+- \[`flake8-pyi`\] Implement `PYI062` (`duplicate-literal-member`) ([#11269](https://github.com/astral-sh/ruff/pull/11269))
+
+### Rule changes
+
+- \[`flake8-boolean-trap`\] Allow passing booleans as positional-only arguments in code such as `set(True)` ([#11287](https://github.com/astral-sh/ruff/pull/11287))
+- \[`flake8-bugbear`\] Ignore enum classes in `cached-instance-method` (`B019`) ([#11312](https://github.com/astral-sh/ruff/pull/11312))
+
+### Server
+
+- Expand tildes when resolving Ruff server configuration file ([#11283](https://github.com/astral-sh/ruff/pull/11283))
+- Fix `ruff server` hanging after Neovim closes ([#11291](https://github.com/astral-sh/ruff/pull/11291))
+- Editor settings are used by default if no file-based configuration exists ([#11266](https://github.com/astral-sh/ruff/pull/11266))
+
+### Bug fixes
+
+- \[`pylint`\] Consider `with` statements for `too-many-branches` (`PLR0912`) ([#11321](https://github.com/astral-sh/ruff/pull/11321))
+- \[`flake8-blind-except`, `tryceratops`\] Respect logged and re-raised expressions in nested statements (`BLE001`, `TRY201`) ([#11301](https://github.com/astral-sh/ruff/pull/11301))
+- Recognise assignments such as `__all__ = builtins.list(["foo", "bar"])` as valid `__all__` definitions ([#11335](https://github.com/astral-sh/ruff/pull/11335))
+
+## 0.4.3
+
+### Enhancements
+
+- Add support for PEP 696 syntax ([#11120](https://github.com/astral-sh/ruff/pull/11120))
+
+### Preview features
+
+- \[`refurb`\] Use function range for `reimplemented-operator` diagnostics ([#11271](https://github.com/astral-sh/ruff/pull/11271))
+- \[`refurb`\] Ignore methods in `reimplemented-operator` (`FURB118`) ([#11270](https://github.com/astral-sh/ruff/pull/11270))
+- \[`refurb`\] Implement `fstring-number-format` (`FURB116`) ([#10921](https://github.com/astral-sh/ruff/pull/10921))
+- \[`ruff`\] Implement `redirected-noqa` (`RUF101`) ([#11052](https://github.com/astral-sh/ruff/pull/11052))
+- \[`pyflakes`\] Distinguish between first-party and third-party imports for fix suggestions ([#11168](https://github.com/astral-sh/ruff/pull/11168))
+
+### Rule changes
+
+- \[`flake8-bugbear`\] Ignore non-abstract class attributes when enforcing `B024` ([#11210](https://github.com/astral-sh/ruff/pull/11210))
+- \[`flake8-logging`\] Include inline instantiations when detecting loggers ([#11154](https://github.com/astral-sh/ruff/pull/11154))
+- \[`pylint`\] Also emit `PLR0206` for properties with variadic parameters ([#11200](https://github.com/astral-sh/ruff/pull/11200))
+- \[`ruff`\] Detect duplicate codes as part of `unused-noqa` (`RUF100`) ([#10850](https://github.com/astral-sh/ruff/pull/10850))
+
+### Formatter
+
+- Avoid multiline expression if format specifier is present ([#11123](https://github.com/astral-sh/ruff/pull/11123))
+
+### LSP
+
+- Write `ruff server` setup guide for Helix ([#11183](https://github.com/astral-sh/ruff/pull/11183))
+- `ruff server` no longer hangs after shutdown ([#11222](https://github.com/astral-sh/ruff/pull/11222))
+- `ruff server` reads from a configuration TOML file in the user configuration directory if no local configuration exists ([#11225](https://github.com/astral-sh/ruff/pull/11225))
+- `ruff server` respects `per-file-ignores` configuration ([#11224](https://github.com/astral-sh/ruff/pull/11224))
+- `ruff server`: Support a custom TOML configuration file ([#11140](https://github.com/astral-sh/ruff/pull/11140))
+- `ruff server`: Support setting to prioritize project configuration over editor configuration ([#11086](https://github.com/astral-sh/ruff/pull/11086))
+
+### Bug fixes
+
+- Avoid debug assertion around NFKC renames ([#11249](https://github.com/astral-sh/ruff/pull/11249))
+- \[`pyflakes`\] Prioritize `redefined-while-unused` over `unused-import` ([#11173](https://github.com/astral-sh/ruff/pull/11173))
+- \[`ruff`\] Respect `async` expressions in comprehension bodies ([#11219](https://github.com/astral-sh/ruff/pull/11219))
+- \[`pygrep_hooks`\] Fix `blanket-noqa` panic when last line has noqa with no newline (`PGH004`) ([#11108](https://github.com/astral-sh/ruff/pull/11108))
+- \[`perflint`\] Ignore list-copy recommendations for async `for` loops ([#11250](https://github.com/astral-sh/ruff/pull/11250))
+- \[`pyflakes`\] Improve `invalid-print-syntax` documentation ([#11171](https://github.com/astral-sh/ruff/pull/11171))
+
+### Performance
+
+- Avoid allocations for isort module names ([#11251](https://github.com/astral-sh/ruff/pull/11251))
+- Build a separate ARM wheel for macOS ([#11149](https://github.com/astral-sh/ruff/pull/11149))
+
+### Windows
+
+- Increase the minimum requirement to Windows 10.
+
 ## 0.4.2
 
 ### Rule changes

CONTRIBUTING.md

@@ -101,6 +101,8 @@ pre-commit run --all-files --show-diff-on-failure  # Rust and Python formatting,
 These checks will run on GitHub Actions when you open your pull request, but running them locally
 will save you time and expedite the merge process.
 
+If you're using VS Code, you can also install the recommended [rust-analyzer](https://marketplace.visualstudio.com/items?itemName=rust-lang.rust-analyzer) extension to get these checks while editing.
+
 Note that many code changes also require updating the snapshot tests, which is done interactively
 after running `cargo test` like so:
 
@@ -344,22 +346,21 @@ even patch releases may contain [non-backwards-compatible changes](https://semve
 1. Run `cargo check`. This should update the lock file with new versions.
 1. Create a pull request with the changelog and version updates
 1. Merge the PR
-1. Run the [release workflow](https://github.com/astral-sh/ruff/actions/workflows/release.yaml) with:
+1. Run the [release workflow](https://github.com/astral-sh/ruff/actions/workflows/release.yml) with:
     - The new version number (without starting `v`)
-    - The commit hash of the merged release pull request on `main`
 1. The release workflow will do the following:
     1. Build all the assets. If this fails (even though we tested in step 4), we haven't tagged or
-        uploaded anything, you can restart after pushing a fix.
+        uploaded anything, you can restart after pushing a fix. If you just need to rerun the build,
+        make sure you're [re-running all the failed
+        jobs](https://docs.github.com/en/actions/managing-workflow-runs/re-running-workflows-and-jobs#re-running-failed-jobs-in-a-workflow) and not just a single failed job.
     1. Upload to PyPI.
     1. Create and push the Git tag (as extracted from `pyproject.toml`). We create the Git tag only
         after building the wheels and uploading to PyPI, since we can't delete or modify the tag ([#4468](https://github.com/astral-sh/ruff/issues/4468)).
     1. Attach artifacts to draft GitHub release
     1. Trigger downstream repositories. This can fail non-catastrophically, as we can run any
         downstream jobs manually if needed.
-1. Publish the GitHub release
-    1. Open the draft release in the GitHub release section
-    1. Copy the changelog for the release into the GitHub release
-        - See previous releases for formatting of section headers
+1. Verify the GitHub release:
+    1. The Changelog should match the content of `CHANGELOG.md`
     1. Append the contributors from the `bump.sh` script
 1. If needed, [update the schemastore](https://github.com/astral-sh/ruff/blob/main/scripts/update_schemastore.py).
     1. One can determine if an update is needed when
@@ -637,11 +638,11 @@ Otherwise, follow the instructions from the linux section.
 `cargo dev` is a shortcut for `cargo run --package ruff_dev --bin ruff_dev`. You can run some useful
 utils with it:
 
-- `cargo dev print-ast <file>`: Print the AST of a python file using the
-    [RustPython parser](https://github.com/astral-sh/ruff/tree/main/crates/ruff_python_parser) that is
-    mainly used in Ruff. For `if True: pass # comment`, you can see the syntax tree, the byte offsets
-    for start and stop of each node and also how the `:` token, the comment and whitespace are not
-    represented anymore:
+- `cargo dev print-ast <file>`: Print the AST of a python file using Ruff's
+    [Python parser](https://github.com/astral-sh/ruff/tree/main/crates/ruff_python_parser).
+    For `if True: pass # comment`, you can see the syntax tree, the byte offsets for start and
+    stop of each node and also how the `:` token, the comment and whitespace are not represented
+    anymore:
 
 ```text
 [

Cargo.toml

@@ -4,7 +4,7 @@ resolver = "2"
 
 [workspace.package]
 edition = "2021"
-rust-version = "1.71"
+rust-version = "1.75"
 homepage = "https://docs.astral.sh/ruff"
 documentation = "https://docs.astral.sh/ruff"
 repository = "https://github.com/astral-sh/ruff"
@@ -12,6 +12,31 @@ authors = ["Charlie Marsh <charlie.r.marsh@gmail.com>"]
 license = "MIT"
 
 [workspace.dependencies]
+ruff = { path = "crates/ruff" }
+ruff_cache = { path = "crates/ruff_cache" }
+ruff_db = { path = "crates/ruff_db" }
+ruff_diagnostics = { path = "crates/ruff_diagnostics" }
+ruff_formatter = { path = "crates/ruff_formatter" }
+ruff_index = { path = "crates/ruff_index" }
+ruff_linter = { path = "crates/ruff_linter" }
+ruff_macros = { path = "crates/ruff_macros" }
+ruff_notebook = { path = "crates/ruff_notebook" }
+ruff_python_ast = { path = "crates/ruff_python_ast" }
+ruff_python_codegen = { path = "crates/ruff_python_codegen" }
+ruff_python_formatter = { path = "crates/ruff_python_formatter" }
+ruff_python_index = { path = "crates/ruff_python_index" }
+ruff_python_literal = { path = "crates/ruff_python_literal" }
+ruff_python_parser = { path = "crates/ruff_python_parser" }
+ruff_python_semantic = { path = "crates/ruff_python_semantic" }
+ruff_python_stdlib = { path = "crates/ruff_python_stdlib" }
+ruff_python_trivia = { path = "crates/ruff_python_trivia" }
+ruff_server = { path = "crates/ruff_server" }
+ruff_source_file = { path = "crates/ruff_source_file" }
+ruff_text_size = { path = "crates/ruff_text_size" }
+ruff_workspace = { path = "crates/ruff_workspace" }
+
+red_knot_module_resolver = { path = "crates/red_knot_module_resolver" }
+
 aho-corasick = { version = "1.1.3" }
 annotate-snippets = { version = "0.9.2", features = ["color"] }
 anyhow = { version = "1.0.80" }
@@ -20,6 +45,7 @@ bincode = { version = "1.3.3" }
 bitflags = { version = "2.5.0" }
 bstr = { version = "1.9.1" }
 cachedir = { version = "0.3.1" }
+camino = { version = "1.1.7" }
 chrono = { version = "0.4.35", default-features = false, features = ["clock"] }
 clap = { version = "4.5.3", features = ["derive"] }
 clap_complete_command = { version = "0.5.1" }
@@ -30,43 +56,41 @@ console_error_panic_hook = { version = "0.1.7" }
 console_log = { version = "1.0.0" }
 countme = { version = "3.0.1" }
 criterion = { version = "0.5.1", default-features = false }
-crossbeam-channel = { version = "0.5.12" }
+crossbeam = { version = "0.8.4" }
 dashmap = { version = "5.5.3" }
-dirs = { version = "5.0.0" }
 drop_bomb = { version = "0.1.5" }
 env_logger = { version = "0.11.0" }
+etcetera = { version = "0.8.0" }
 fern = { version = "0.6.1" }
 filetime = { version = "0.2.23" }
-fs-err = { version = "2.11.0" }
 glob = { version = "0.3.1" }
 globset = { version = "0.4.14" }
 hashbrown = "0.14.3"
-hexf-parse = { version = "0.2.1" }
 ignore = { version = "0.4.22" }
 imara-diff = { version = "0.1.5" }
 imperative = { version = "1.0.4" }
 indexmap = { version = "2.2.6" }
 indicatif = { version = "0.17.8" }
 indoc = { version = "2.0.4" }
-insta = { version = "1.35.1", feature = ["filters", "glob"] }
+insta = { version = "1.35.1" }
 insta-cmd = { version = "0.6.0" }
 is-macro = { version = "0.3.5" }
 is-wsl = { version = "0.4.0" }
-itertools = { version = "0.12.1" }
+itertools = { version = "0.13.0" }
 js-sys = { version = "0.3.69" }
 jod-thread = { version = "0.1.2" }
-lexical-parse-float = { version = "0.8.0", features = ["format"] }
 libc = { version = "0.2.153" }
 libcst = { version = "1.1.0", default-features = false }
 log = { version = "0.4.17" }
 lsp-server = { version = "0.7.6" }
-lsp-types = { version = "0.95.0", features = ["proposed"] }
+lsp-types = { git = "https://github.com/astral-sh/lsp-types.git", rev = "3512a9f", features = [
+  "proposed",
+] }
 matchit = { version = "0.8.1" }
 memchr = { version = "2.7.1" }
 mimalloc = { version = "0.1.39" }
 natord = { version = "1.0.9" }
 notify = { version = "6.1.1" }
-num_cpus = { version = "1.16.0" }
 once_cell = { version = "1.19.0" }
 path-absolutize = { version = "3.1.1" }
 path-slash = { version = "0.2.1" }
@@ -81,19 +105,21 @@ quote = { version = "1.0.23" }
 rand = { version = "0.8.5" }
 rayon = { version = "1.10.0" }
 regex = { version = "1.10.2" }
-result-like = { version = "0.5.0" }
-rustc-hash = { version = "1.1.0" }
+rustc-hash = { version = "2.0.0" }
+salsa = { git = "https://github.com/salsa-rs/salsa.git", rev = "f706aa2d32d473ee633a77c1af01d180c85da308" }
 schemars = { version = "0.8.16" }
 seahash = { version = "4.1.0" }
 serde = { version = "1.0.197", features = ["derive"] }
 serde-wasm-bindgen = { version = "0.6.4" }
 serde_json = { version = "1.0.113" }
 serde_test = { version = "1.0.152" }
-serde_with = { version = "3.6.0", default-features = false, features = ["macros"] }
+serde_with = { version = "3.6.0", default-features = false, features = [
+  "macros",
+] }
 shellexpand = { version = "3.0.0" }
-shlex = { version = "1.3.0" }
 similar = { version = "2.4.0", features = ["inline"] }
 smallvec = { version = "1.13.2" }
+smol_str = { version = "0.2.2" }
 static_assertions = "1.1.0"
 strum = { version = "0.26.0", features = ["strum_macros"] }
 strum_macros = { version = "0.26.0" }
@@ -115,11 +141,17 @@ unicode_names2 = { version = "1.2.2" }
 unicode-normalization = { version = "0.1.23" }
 ureq = { version = "2.9.6" }
 url = { version = "2.5.0" }
-uuid = { version = "1.6.1", features = ["v4", "fast-rng", "macro-diagnostics", "js"] }
+uuid = { version = "1.6.1", features = [
+  "v4",
+  "fast-rng",
+  "macro-diagnostics",
+  "js",
+] }
 walkdir = { version = "2.3.2" }
 wasm-bindgen = { version = "0.2.92" }
 wasm-bindgen-test = { version = "0.3.42" }
 wild = { version = "2" }
+zip = { version = "0.6.6", default-features = false, features = ["zstd"] }
 
 [workspace.lints.rust]
 unsafe_code = "warn"
@@ -132,6 +164,7 @@ char_lit_as_u8 = "allow"
 collapsible_else_if = "allow"
 collapsible_if = "allow"
 implicit_hasher = "allow"
+map_unwrap_or = "allow"
 match_same_arms = "allow"
 missing_errors_doc = "allow"
 missing_panics_doc = "allow"
@@ -186,3 +219,60 @@ opt-level = 1
 [profile.profiling]
 inherits = "release"
 debug = 1
+
+# The profile that 'cargo dist' will build with.
+[profile.dist]
+inherits = "release"
+
+# Config for 'cargo dist'
+[workspace.metadata.dist]
+# The preferred cargo-dist version to use in CI (Cargo.toml SemVer syntax)
+cargo-dist-version = "0.14.0"
+# CI backends to support
+ci = ["github"]
+# The installers to generate for each app
+installers = ["shell", "powershell"]
+# The archive format to use for windows builds (defaults .zip)
+windows-archive = ".zip"
+# The archive format to use for non-windows builds (defaults .tar.xz)
+unix-archive = ".tar.gz"
+# Target platforms to build apps for (Rust target-triple syntax)
+targets = [
+  "aarch64-apple-darwin",
+  "aarch64-pc-windows-msvc",
+  "aarch64-unknown-linux-gnu",
+  "aarch64-unknown-linux-musl",
+  "arm-unknown-linux-musleabihf",
+  "armv7-unknown-linux-gnueabihf",
+  "armv7-unknown-linux-musleabihf",
+  "i686-pc-windows-msvc",
+  "i686-unknown-linux-gnu",
+  "i686-unknown-linux-musl",
+  "powerpc64-unknown-linux-gnu",
+  "powerpc64le-unknown-linux-gnu",
+  "s390x-unknown-linux-gnu",
+  "x86_64-apple-darwin",
+  "x86_64-pc-windows-msvc",
+  "x86_64-unknown-linux-gnu",
+  "x86_64-unknown-linux-musl",
+]
+# Whether to auto-include files like READMEs, LICENSEs, and CHANGELOGs (default true)
+auto-includes = false
+# Whether cargo-dist should create a Github Release or use an existing draft
+create-release = true
+# Publish jobs to run in CI
+pr-run-mode = "skip"
+# Whether CI should trigger releases with dispatches instead of tag pushes
+dispatch-releases = true
+# Whether CI should include auto-generated code to build local artifacts
+build-local-artifacts = false
+# Local artifacts jobs to run in CI
+local-artifacts-jobs = ["./build-binaries", "./build-docker"]
+# Publish jobs to run in CI
+publish-jobs = ["./publish-pypi"]
+# Announcement jobs to run in CI
+post-announce-jobs = ["./notify-dependents", "./publish-docs", "./publish-playground"]
+# Skip checking whether the specified configuration files are up to date
+allow-dirty = ["ci"]
+# Whether to install an updater program
+install-updater = false

README.md

@@ -28,7 +28,7 @@ An extremely fast Python linter and code formatter, written in Rust.
 - ⚡️ 10-100x faster than existing linters (like Flake8) and formatters (like Black)
 - 🐍 Installable via `pip`
 - 🛠️ `pyproject.toml` support
-- 🤝 Python 3.12 compatibility
+- 🤝 Python 3.13 compatibility
 - ⚖️ Drop-in parity with [Flake8](https://docs.astral.sh/ruff/faq/#how-does-ruff-compare-to-flake8), isort, and Black
 - 📦 Built-in caching, to avoid re-analyzing unchanged files
 - 🔧 Fix support, for automatic error correction (e.g., automatically remove unused imports)
@@ -152,7 +152,7 @@ Ruff can also be used as a [pre-commit](https://pre-commit.com/) hook via [`ruff
 ```yaml
 - repo: https://github.com/astral-sh/ruff-pre-commit
   # Ruff version.
-  rev: v0.4.2
+  rev: v0.5.0
   hooks:
     # Run the linter.
     - id: ruff
@@ -266,6 +266,11 @@ The remaining configuration options can be provided through a catch-all `--confi
 ruff check --config "lint.per-file-ignores = {'some_file.py' = ['F841']}"
 ```
 
+To opt in to the latest lint rules, formatter style changes, interface updates, and more, enable
+[preview mode](https://docs.astral.sh/ruff/rules/) by setting `preview = true` in your configuration
+file or passing `--preview` on the command line. Preview mode enables a collection of unstable
+features that may change prior to stabilization.
+
 See `ruff help` for more on Ruff's top-level commands, or `ruff help check` and `ruff help format`
 for more on the linting and formatting commands, respectively.
 
@@ -329,7 +334,6 @@ quality tools, including:
 - [flake8-super](https://pypi.org/project/flake8-super/)
 - [flake8-tidy-imports](https://pypi.org/project/flake8-tidy-imports/)
 - [flake8-todos](https://pypi.org/project/flake8-todos/)
-- [flake8-trio](https://pypi.org/project/flake8-trio/)
 - [flake8-type-checking](https://pypi.org/project/flake8-type-checking/)
 - [flake8-use-pathlib](https://pypi.org/project/flake8-use-pathlib/)
 - [flynt](https://pypi.org/project/flynt/) ([#2102](https://github.com/astral-sh/ruff/issues/2102))
@@ -403,6 +407,7 @@ Ruff is used by a number of major open-source projects and companies, including:
 - [Dagster](https://github.com/dagster-io/dagster)
 - Databricks ([MLflow](https://github.com/mlflow/mlflow))
 - [FastAPI](https://github.com/tiangolo/fastapi)
+- [Godot](https://github.com/godotengine/godot)
 - [Gradio](https://github.com/gradio-app/gradio)
 - [Great Expectations](https://github.com/great-expectations/great_expectations)
 - [HTTPX](https://github.com/encode/httpx)
@@ -425,9 +430,10 @@ Ruff is used by a number of major open-source projects and companies, including:
 - Microsoft ([Semantic Kernel](https://github.com/microsoft/semantic-kernel),
     [ONNX Runtime](https://github.com/microsoft/onnxruntime),
     [LightGBM](https://github.com/microsoft/LightGBM))
-- Modern Treasury ([Python SDK](https://github.com/Modern-Treasury/modern-treasury-python-sdk))
+- Modern Treasury ([Python SDK](https://github.com/Modern-Treasury/modern-treasury-python))
 - Mozilla ([Firefox](https://github.com/mozilla/gecko-dev))
 - [Mypy](https://github.com/python/mypy)
+- [Nautobot](https://github.com/nautobot/nautobot)
 - Netflix ([Dispatch](https://github.com/Netflix/dispatch))
 - [Neon](https://github.com/neondatabase/neon)
 - [Nokia](https://nokia.com/)
@@ -435,6 +441,7 @@ Ruff is used by a number of major open-source projects and companies, including:
 - [NumPyro](https://github.com/pyro-ppl/numpyro)
 - [ONNX](https://github.com/onnx/onnx)
 - [OpenBB](https://github.com/OpenBB-finance/OpenBBTerminal)
+- [Open Wine Components](https://github.com/Open-Wine-Components/umu-launcher)
 - [PDM](https://github.com/pdm-project/pdm)
 - [PaddlePaddle](https://github.com/PaddlePaddle/Paddle)
 - [Pandas](https://github.com/pandas-dev/pandas)
@@ -462,6 +469,7 @@ Ruff is used by a number of major open-source projects and companies, including:
 - [Sphinx](https://github.com/sphinx-doc/sphinx)
 - [Stable Baselines3](https://github.com/DLR-RM/stable-baselines3)
 - [Starlette](https://github.com/encode/starlette)
+- [Streamlit](https://github.com/streamlit/streamlit)
 - [The Algorithms](https://github.com/TheAlgorithms/Python)
 - [Vega-Altair](https://github.com/altair-viz/altair)
 - WordPress ([Openverse](https://github.com/WordPress/openverse))

_typos.toml

@@ -1,6 +1,6 @@
 [files]
 # https://github.com/crate-ci/typos/issues/868
-extend-exclude = ["**/resources/**/*", "**/snapshots/**/*"]
+extend-exclude = ["crates/red_knot_module_resolver/vendor/**/*", "**/resources/**/*", "**/snapshots/**/*"]
 
 [default.extend-words]
 "arange" = "arange"  # e.g. `numpy.arange`
@@ -11,9 +11,11 @@ ned = "ned"
 pn = "pn"  # `import panel as pd` is a thing
 poit = "poit"
 BA = "BA" # acronym for "Bad Allowed", used in testing.
+jod = "jod" # e.g., `jod-thread`
 
 [default]
 extend-ignore-re = [
   # Line ignore with trailing "spellchecker:disable-line"
-  "(?Rm)^.*#\\s*spellchecker:disable-line$"
+  "(?Rm)^.*#\\s*spellchecker:disable-line$",
+  "LICENSEs",
 ]

clippy.toml

@@ -1,7 +1,13 @@
 doc-valid-idents = [
-  "StackOverflow",
-  "CodeQL",
-  "IPython",
-  "NumPy",
   "..",
+  "CodeQL",
+  "FastAPI",
+  "IPython",
+  "LangChain",
+  "LibCST",
+  "McCabe",
+  "NumPy",
+  "SCREAMING_SNAKE_CASE",
+  "SQLAlchemy",
+  "StackOverflow",
 ]

crates/red_knot/Cargo.toml

@@ -12,33 +12,32 @@ license.workspace = true
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-ruff_python_parser = { path = "../ruff_python_parser" }
-ruff_python_ast = { path = "../ruff_python_ast" }
-ruff_python_trivia = { path = "../ruff_python_trivia" }
-ruff_text_size = { path = "../ruff_text_size" }
-ruff_index = { path = "../ruff_index" }
-ruff_notebook = { path = "../ruff_notebook" }
+red_knot_module_resolver = { workspace = true }
+
+ruff_python_parser = { workspace = true }
+ruff_python_ast = { workspace = true }
+ruff_text_size = { workspace = true }
+ruff_index = { workspace = true }
+ruff_notebook = { workspace = true }
 
 anyhow = { workspace = true }
 bitflags = { workspace = true }
-ctrlc = "3.4.4"
-crossbeam-channel = { workspace = true }
+crossbeam = { workspace = true }
+ctrlc = { version = "3.4.4" }
 dashmap = { workspace = true }
 hashbrown = { workspace = true }
 indexmap = { workspace = true }
-log = { workspace = true }
+is-macro = { workspace = true }
 notify = { workspace = true }
 parking_lot = { workspace = true }
 rayon = { workspace = true }
 rustc-hash = { workspace = true }
-smallvec = { workspace = true }
-smol_str = "0.2.1"
+smol_str = { version = "0.2.1" }
 tracing = { workspace = true }
 tracing-subscriber = { workspace = true }
 tracing-tree = { workspace = true }
 
 [dev-dependencies]
-textwrap = "0.16.1"
 tempfile = { workspace = true }
 
 [lints]

crates/red_knot/src/ast_ids.rs

@@ -6,8 +6,8 @@ use std::marker::PhantomData;
 use rustc_hash::FxHashMap;
 
 use ruff_index::{Idx, IndexVec};
-use ruff_python_ast::visitor::preorder;
-use ruff_python_ast::visitor::preorder::{PreorderVisitor, TraversalSignal};
+use ruff_python_ast::visitor::source_order;
+use ruff_python_ast::visitor::source_order::{SourceOrderVisitor, TraversalSignal};
 use ruff_python_ast::{
     AnyNodeRef, AstNode, ExceptHandler, ExceptHandlerExceptHandler, Expr, MatchCase, ModModule,
     NodeKind, Parameter, Stmt, StmtAnnAssign, StmtAssign, StmtAugAssign, StmtClassDef,
@@ -91,9 +91,9 @@ impl AstIds {
         while let Some(deferred) = visitor.deferred.pop() {
             match deferred {
                 DeferredNode::FunctionDefinition(def) => {
-                    def.visit_preorder(&mut visitor);
+                    def.visit_source_order(&mut visitor);
                 }
-                DeferredNode::ClassDefinition(def) => def.visit_preorder(&mut visitor),
+                DeferredNode::ClassDefinition(def) => def.visit_source_order(&mut visitor),
             }
         }
 
@@ -182,7 +182,7 @@ impl<'a> AstIdsVisitor<'a> {
     }
 }
 
-impl<'a> PreorderVisitor<'a> for AstIdsVisitor<'a> {
+impl<'a> SourceOrderVisitor<'a> for AstIdsVisitor<'a> {
     fn visit_stmt(&mut self, stmt: &'a Stmt) {
         match stmt {
             Stmt::FunctionDef(def) => {
@@ -226,14 +226,14 @@ impl<'a> PreorderVisitor<'a> for AstIdsVisitor<'a> {
             Stmt::IpyEscapeCommand(_) => {}
         }
 
-        preorder::walk_stmt(self, stmt);
+        source_order::walk_stmt(self, stmt);
     }
 
     fn visit_expr(&mut self, _expr: &'a Expr) {}
 
     fn visit_parameter(&mut self, parameter: &'a Parameter) {
         self.create_id(parameter);
-        preorder::walk_parameter(self, parameter);
+        source_order::walk_parameter(self, parameter);
     }
 
     fn visit_except_handler(&mut self, except_handler: &'a ExceptHandler) {
@@ -243,17 +243,17 @@ impl<'a> PreorderVisitor<'a> for AstIdsVisitor<'a> {
             }
         }
 
-        preorder::walk_except_handler(self, except_handler);
+        source_order::walk_except_handler(self, except_handler);
     }
 
     fn visit_with_item(&mut self, with_item: &'a WithItem) {
         self.create_id(with_item);
-        preorder::walk_with_item(self, with_item);
+        source_order::walk_with_item(self, with_item);
     }
 
     fn visit_match_case(&mut self, match_case: &'a MatchCase) {
         self.create_id(match_case);
-        preorder::walk_match_case(self, match_case);
+        source_order::walk_match_case(self, match_case);
     }
 
     fn visit_type_param(&mut self, type_param: &'a TypeParam) {
@@ -275,10 +275,7 @@ pub struct TypedNodeKey<N: AstNode> {
 
 impl<N: AstNode> TypedNodeKey<N> {
     pub fn from_node(node: &N) -> Self {
-        let inner = NodeKey {
-            kind: node.as_any_node_ref().kind(),
-            range: node.range(),
-        };
+        let inner = NodeKey::from_node(node.as_any_node_ref());
         Self {
             inner,
             _marker: PhantomData,
@@ -312,7 +309,7 @@ struct FindNodeKeyVisitor<'a> {
     result: Option<AnyNodeRef<'a>>,
 }
 
-impl<'a> PreorderVisitor<'a> for FindNodeKeyVisitor<'a> {
+impl<'a> SourceOrderVisitor<'a> for FindNodeKeyVisitor<'a> {
     fn enter_node(&mut self, node: AnyNodeRef<'a>) -> TraversalSignal {
         if self.result.is_some() {
             return TraversalSignal::Skip;
@@ -352,6 +349,12 @@ pub struct NodeKey {
 }
 
 impl NodeKey {
+    pub fn from_node(node: AnyNodeRef) -> Self {
+        NodeKey {
+            kind: node.kind(),
+            range: node.range(),
+        }
+    }
     pub fn resolve<'a>(&self, root: AnyNodeRef<'a>) -> Option<AnyNodeRef<'a>> {
         // We need to do a binary search here. Only traverse into a node if the range is withint the node
         let mut visitor = FindNodeKeyVisitor {

crates/red_knot/src/cache.rs

@@ -2,6 +2,7 @@ use std::fmt::Formatter;
 use std::hash::Hash;
 use std::sync::atomic::{AtomicUsize, Ordering};
 
+use crate::db::QueryResult;
 use dashmap::mapref::entry::Entry;
 
 use crate::FxDashMap;
@@ -27,11 +28,11 @@ where
         }
     }
 
-    pub fn get<F>(&self, key: &K, compute: F) -> V
+    pub fn get<F>(&self, key: &K, compute: F) -> QueryResult<V>
     where
-        F: FnOnce(&K) -> V,
+        F: FnOnce(&K) -> QueryResult<V>,
     {
-        match self.map.entry(key.clone()) {
+        Ok(match self.map.entry(key.clone()) {
             Entry::Occupied(cached) => {
                 self.statistics.hit();
 
@@ -40,11 +41,11 @@ where
             Entry::Vacant(vacant) => {
                 self.statistics.miss();
 
-                let value = compute(key);
+                let value = compute(key)?;
                 vacant.insert(value.clone());
                 value
             }
-        }
+        })
     }
 
     pub fn set(&mut self, key: K, value: V) {
@@ -117,23 +118,29 @@ pub type CacheStatistics = DebugStatistics;
 #[cfg(not(debug_assertions))]
 pub type CacheStatistics = ReleaseStatistics;
 
+pub trait StatisticsRecorder {
+    fn hit(&self);
+    fn miss(&self);
+    fn to_statistics(&self) -> Option<Statistics>;
+}
+
 #[derive(Debug, Default)]
 pub struct DebugStatistics {
     hits: AtomicUsize,
     misses: AtomicUsize,
 }
 
-impl DebugStatistics {
+impl StatisticsRecorder for DebugStatistics {
     // TODO figure out appropriate Ordering
-    pub fn hit(&self) {
+    fn hit(&self) {
         self.hits.fetch_add(1, Ordering::SeqCst);
     }
 
-    pub fn miss(&self) {
+    fn miss(&self) {
         self.misses.fetch_add(1, Ordering::SeqCst);
     }
 
-    pub fn to_statistics(&self) -> Option<Statistics> {
+    fn to_statistics(&self) -> Option<Statistics> {
         let hits = self.hits.load(Ordering::SeqCst);
         let misses = self.misses.load(Ordering::SeqCst);
 
@@ -144,15 +151,15 @@ impl DebugStatistics {
 #[derive(Debug, Default)]
 pub struct ReleaseStatistics;
 
-impl ReleaseStatistics {
+impl StatisticsRecorder for ReleaseStatistics {
     #[inline]
-    pub const fn hit(&self) {}
+    fn hit(&self) {}
 
     #[inline]
-    pub const fn miss(&self) {}
+    fn miss(&self) {}
 
     #[inline]
-    pub const fn to_statistics(&self) -> Option<Statistics> {
+    fn to_statistics(&self) -> Option<Statistics> {
         None
     }
 }

crates/red_knot/src/cancellation.rs

@@ -1,35 +1,25 @@
-use std::sync::{Arc, Condvar, Mutex};
+use std::sync::atomic::AtomicBool;
+use std::sync::Arc;
 
-#[derive(Debug, Default)]
+#[derive(Debug, Clone, Default)]
 pub struct CancellationTokenSource {
-    signal: Arc<(Mutex<bool>, Condvar)>,
+    signal: Arc<AtomicBool>,
 }
 
 impl CancellationTokenSource {
     pub fn new() -> Self {
         Self {
-            signal: Arc::new((Mutex::new(false), Condvar::default())),
+            signal: Arc::new(AtomicBool::new(false)),
         }
     }
 
-    #[tracing::instrument(level = "trace")]
+    #[tracing::instrument(level = "trace", skip_all)]
     pub fn cancel(&self) {
-        let (cancelled, condvar) = &*self.signal;
-
-        let mut cancelled = cancelled.lock().unwrap();
-
-        if *cancelled {
-            return;
-        }
-
-        *cancelled = true;
-        condvar.notify_all();
+        self.signal.store(true, std::sync::atomic::Ordering::SeqCst);
     }
 
     pub fn is_cancelled(&self) -> bool {
-        let (cancelled, _) = &*self.signal;
-
-        *cancelled.lock().unwrap()
+        self.signal.load(std::sync::atomic::Ordering::SeqCst)
     }
 
     pub fn token(&self) -> CancellationToken {
@@ -41,26 +31,12 @@ impl CancellationTokenSource {
 
 #[derive(Clone, Debug)]
 pub struct CancellationToken {
-    signal: Arc<(Mutex<bool>, Condvar)>,
+    signal: Arc<AtomicBool>,
 }
 
 impl CancellationToken {
     /// Returns `true` if cancellation has been requested.
     pub fn is_cancelled(&self) -> bool {
-        let (cancelled, _) = &*self.signal;
-
-        *cancelled.lock().unwrap()
-    }
-
-    pub fn wait(&self) {
-        let (bool, condvar) = &*self.signal;
-
-        let lock = condvar
-            .wait_while(bool.lock().unwrap(), |bool| !*bool)
-            .unwrap();
-
-        debug_assert!(*lock);
-
-        drop(lock);
+        self.signal.load(std::sync::atomic::Ordering::SeqCst)
     }
 }

crates/red_knot/src/db.rs

@@ -1,98 +1,151 @@
-use std::path::Path;
 use std::sync::Arc;
 
-use crate::files::FileId;
-use crate::lint::{Diagnostics, LintSyntaxStorage};
-use crate::module::{Module, ModuleData, ModuleName, ModuleResolver, ModuleSearchPath};
-use crate::parse::{Parsed, ParsedStorage};
-use crate::source::{Source, SourceStorage};
-use crate::symbols::{SymbolId, SymbolTable, SymbolTablesStorage};
-use crate::types::{Type, TypeStore};
+pub use jars::{HasJar, HasJars};
+pub use query::{QueryError, QueryResult};
+pub use runtime::DbRuntime;
+pub use storage::JarsStorage;
 
-pub trait SourceDb {
+use crate::files::FileId;
+use crate::lint::{LintSemanticStorage, LintSyntaxStorage};
+use crate::module::ModuleResolver;
+use crate::parse::ParsedStorage;
+use crate::semantic::SemanticIndexStorage;
+use crate::semantic::TypeStore;
+use crate::source::SourceStorage;
+
+mod jars;
+mod query;
+mod runtime;
+mod storage;
+
+pub trait Database {
+    /// Returns a reference to the runtime of the current worker.
+    fn runtime(&self) -> &DbRuntime;
+
+    /// Returns a mutable reference to the runtime. Only one worker can hold a mutable reference to the runtime.
+    fn runtime_mut(&mut self) -> &mut DbRuntime;
+
+    /// Returns `Ok` if the queries have not been cancelled and `Err(QueryError::Cancelled)` otherwise.
+    fn cancelled(&self) -> QueryResult<()> {
+        self.runtime().cancelled()
+    }
+
+    /// Returns `true` if the queries have been cancelled.
+    fn is_cancelled(&self) -> bool {
+        self.runtime().is_cancelled()
+    }
+}
+
+/// Database that supports running queries from multiple threads.
+pub trait ParallelDatabase: Database + Send {
+    /// Creates a snapshot of the database state that can be used to query the database in another thread.
+    ///
+    /// The snapshot is a read-only view of the database but query results are shared between threads.
+    /// All queries will be automatically cancelled when applying any mutations (calling [`HasJars::jars_mut`])
+    /// to the database (not the snapshot, because they're readonly).
+    ///
+    /// ## Creating a snapshot
+    ///
+    /// Creating a snapshot of the database's jars is cheap but creating a snapshot of
+    /// other state stored on the database might require deep-cloning data. That's why you should
+    /// avoid creating snapshots in a hot function (e.g. don't create a snapshot for each file, instead
+    /// create a snapshot when scheduling the check of an entire program).
+    ///
+    /// ## Salsa compatibility
+    /// Salsa prohibits creating a snapshot while running a local query (it's fine if other workers run a query) [[source](https://github.com/salsa-rs/salsa/issues/80)].
+    /// We should avoid creating snapshots while running a query because we might want to adopt Salsa in the future (if we can figure out persistent caching).
+    /// Unfortunately, the infrastructure doesn't provide an automated way of knowing when a query is run, that's
+    /// why we have to "enforce" this constraint manually.
+    #[must_use]
+    fn snapshot(&self) -> Snapshot<Self>;
+}
+
+pub trait DbWithJar<Jar>: Database + HasJar<Jar> {}
+
+/// Readonly snapshot of a database.
+///
+/// ## Dead locks
+/// A snapshot should always be dropped as soon as it is no longer necessary to run queries.
+/// Storing the snapshot without running a query or periodically checking if cancellation was requested
+/// can lead to deadlocks because mutating the [`Database`] requires cancels all pending queries
+/// and waiting for all [`Snapshot`]s to be dropped.
+#[derive(Debug)]
+pub struct Snapshot<DB: ?Sized>
+where
+    DB: ParallelDatabase,
+{
+    db: DB,
+}
+
+impl<DB> Snapshot<DB>
+where
+    DB: ParallelDatabase,
+{
+    pub fn new(db: DB) -> Self {
+        Snapshot { db }
+    }
+}
+
+impl<DB> std::ops::Deref for Snapshot<DB>
+where
+    DB: ParallelDatabase,
+{
+    type Target = DB;
+
+    fn deref(&self) -> &DB {
+        &self.db
+    }
+}
+
+pub trait Upcast<T: ?Sized> {
+    fn upcast(&self) -> &T;
+}
+
+// Red knot specific databases code.
+
+pub trait SourceDb: DbWithJar<SourceJar> {
     // queries
     fn file_id(&self, path: &std::path::Path) -> FileId;
 
     fn file_path(&self, file_id: FileId) -> Arc<std::path::Path>;
-
-    fn source(&self, file_id: FileId) -> Source;
-
-    fn parse(&self, file_id: FileId) -> Parsed;
-
-    fn lint_syntax(&self, file_id: FileId) -> Diagnostics;
 }
 
-pub trait SemanticDb: SourceDb {
-    // queries
-    fn resolve_module(&self, name: ModuleName) -> Option<Module>;
+pub trait SemanticDb: SourceDb + DbWithJar<SemanticJar> + Upcast<dyn SourceDb> {}
 
-    fn file_to_module(&self, file_id: FileId) -> Option<Module>;
+pub trait LintDb: SemanticDb + DbWithJar<LintJar> + Upcast<dyn SemanticDb> {}
 
-    fn path_to_module(&self, path: &Path) -> Option<Module>;
-
-    fn symbol_table(&self, file_id: FileId) -> Arc<SymbolTable>;
-
-    // mutations
-
-    fn add_module(&mut self, path: &Path) -> Option<(Module, Vec<Arc<ModuleData>>)>;
-
-    fn set_module_search_paths(&mut self, paths: Vec<ModuleSearchPath>);
-
-    fn infer_symbol_type(&mut self, file_id: FileId, symbol_id: SymbolId) -> Type;
-}
-
-pub trait Db: SemanticDb {}
+pub trait Db: LintDb + Upcast<dyn LintDb> {}
 
 #[derive(Debug, Default)]
 pub struct SourceJar {
     pub sources: SourceStorage,
     pub parsed: ParsedStorage,
-    pub lint_syntax: LintSyntaxStorage,
 }
 
 #[derive(Debug, Default)]
 pub struct SemanticJar {
     pub module_resolver: ModuleResolver,
-    pub symbol_tables: SymbolTablesStorage,
+    pub semantic_indices: SemanticIndexStorage,
     pub type_store: TypeStore,
 }
 
-/// Gives access to a specific jar in the database.
-///
-/// Nope, the terminology isn't borrowed from Java but from Salsa <https://salsa-rs.github.io/salsa/>,
-/// which is an analogy to storing the salsa in different jars.
-///
-/// The basic idea is that each crate can define its own jar and the jars can be combined to a single
-/// database in the top level crate. Each crate also defines its own `Database` trait. The combination of
-/// `Database` trait and the jar allows to write queries in isolation without having to know how they get composed at the upper levels.
-///
-/// Salsa further defines a `HasIngredient` trait which slices the jar to a specific storage (e.g. a specific cache).
-/// We don't need this just jet because we write our queries by hand. We may want a similar trait if we decide
-/// to use a macro to generate the queries.
-pub trait HasJar<T> {
-    /// Gives a read-only reference to the jar.
-    fn jar(&self) -> &T;
-
-    /// Gives a mutable reference to the jar.
-    fn jar_mut(&mut self) -> &mut T;
+#[derive(Debug, Default)]
+pub struct LintJar {
+    pub lint_syntax: LintSyntaxStorage,
+    pub lint_semantic: LintSemanticStorage,
 }
 
 #[cfg(test)]
 pub(crate) mod tests {
-    use crate::db::{HasJar, SourceDb, SourceJar};
-    use crate::files::{FileId, Files};
-    use crate::lint::{lint_syntax, Diagnostics};
-    use crate::module::{
-        add_module, file_to_module, path_to_module, resolve_module, set_module_search_paths,
-        Module, ModuleData, ModuleName, ModuleSearchPath,
-    };
-    use crate::parse::{parse, Parsed};
-    use crate::source::{source_text, Source};
-    use crate::symbols::{symbol_table, SymbolId, SymbolTable};
-    use crate::types::{infer_symbol_type, Type};
     use std::path::Path;
     use std::sync::Arc;
 
+    use crate::db::{
+        Database, DbRuntime, DbWithJar, HasJar, HasJars, JarsStorage, LintDb, LintJar, QueryResult,
+        SourceDb, SourceJar, Upcast,
+    };
+    use crate::files::{FileId, Files};
+
     use super::{SemanticDb, SemanticJar};
 
     // This can be a partial database used in a single crate for testing.
@@ -100,27 +153,36 @@ pub(crate) mod tests {
     #[derive(Debug, Default)]
     pub(crate) struct TestDb {
         files: Files,
-        source: SourceJar,
-        semantic: SemanticJar,
+        jars: JarsStorage<Self>,
     }
 
     impl HasJar<SourceJar> for TestDb {
-        fn jar(&self) -> &SourceJar {
-            &self.source
+        fn jar(&self) -> QueryResult<&SourceJar> {
+            Ok(&self.jars()?.0)
         }
 
         fn jar_mut(&mut self) -> &mut SourceJar {
-            &mut self.source
+            &mut self.jars_mut().0
         }
     }
 
     impl HasJar<SemanticJar> for TestDb {
-        fn jar(&self) -> &SemanticJar {
-            &self.semantic
+        fn jar(&self) -> QueryResult<&SemanticJar> {
+            Ok(&self.jars()?.1)
         }
 
         fn jar_mut(&mut self) -> &mut SemanticJar {
-            &mut self.semantic
+            &mut self.jars_mut().1
+        }
+    }
+
+    impl HasJar<LintJar> for TestDb {
+        fn jar(&self) -> QueryResult<&LintJar> {
+            Ok(&self.jars()?.2)
+        }
+
+        fn jar_mut(&mut self) -> &mut LintJar {
+            &mut self.jars_mut().2
         }
     }
 
@@ -132,47 +194,55 @@ pub(crate) mod tests {
         fn file_path(&self, file_id: FileId) -> Arc<Path> {
             self.files.path(file_id)
         }
+    }
 
-        fn source(&self, file_id: FileId) -> Source {
-            source_text(self, file_id)
-        }
+    impl DbWithJar<SourceJar> for TestDb {}
 
-        fn parse(&self, file_id: FileId) -> Parsed {
-            parse(self, file_id)
-        }
-
-        fn lint_syntax(&self, file_id: FileId) -> Diagnostics {
-            lint_syntax(self, file_id)
+    impl Upcast<dyn SourceDb> for TestDb {
+        fn upcast(&self) -> &(dyn SourceDb + 'static) {
+            self
         }
     }
 
-    impl SemanticDb for TestDb {
-        fn resolve_module(&self, name: ModuleName) -> Option<Module> {
-            resolve_module(self, name)
+    impl SemanticDb for TestDb {}
+
+    impl DbWithJar<SemanticJar> for TestDb {}
+
+    impl Upcast<dyn SemanticDb> for TestDb {
+        fn upcast(&self) -> &(dyn SemanticDb + 'static) {
+            self
+        }
+    }
+
+    impl LintDb for TestDb {}
+
+    impl Upcast<dyn LintDb> for TestDb {
+        fn upcast(&self) -> &(dyn LintDb + 'static) {
+            self
+        }
+    }
+
+    impl DbWithJar<LintJar> for TestDb {}
+
+    impl HasJars for TestDb {
+        type Jars = (SourceJar, SemanticJar, LintJar);
+
+        fn jars(&self) -> QueryResult<&Self::Jars> {
+            self.jars.jars()
         }
 
-        fn file_to_module(&self, file_id: FileId) -> Option<Module> {
-            file_to_module(self, file_id)
+        fn jars_mut(&mut self) -> &mut Self::Jars {
+            self.jars.jars_mut()
+        }
+    }
+
+    impl Database for TestDb {
+        fn runtime(&self) -> &DbRuntime {
+            self.jars.runtime()
         }
 
-        fn path_to_module(&self, path: &Path) -> Option<Module> {
-            path_to_module(self, path)
-        }
-
-        fn symbol_table(&self, file_id: FileId) -> Arc<SymbolTable> {
-            symbol_table(self, file_id)
-        }
-
-        fn add_module(&mut self, path: &Path) -> Option<(Module, Vec<Arc<ModuleData>>)> {
-            add_module(self, path)
-        }
-
-        fn set_module_search_paths(&mut self, paths: Vec<ModuleSearchPath>) {
-            set_module_search_paths(self, paths);
-        }
-
-        fn infer_symbol_type(&mut self, file_id: FileId, symbol_id: SymbolId) -> Type {
-            infer_symbol_type(self, file_id, symbol_id)
+        fn runtime_mut(&mut self) -> &mut DbRuntime {
+            self.jars.runtime_mut()
         }
     }
 }

crates/red_knot/src/db/jars.rs

@@ -0,0 +1,37 @@
+use crate::db::query::QueryResult;
+
+/// Gives access to a specific jar in the database.
+///
+/// Nope, the terminology isn't borrowed from Java but from Salsa <https://salsa-rs.github.io/salsa/>,
+/// which is an analogy to storing the salsa in different jars.
+///
+/// The basic idea is that each crate can define its own jar and the jars can be combined to a single
+/// database in the top level crate. Each crate also defines its own `Database` trait. The combination of
+/// `Database` trait and the jar allows to write queries in isolation without having to know how they get composed at the upper levels.
+///
+/// Salsa further defines a `HasIngredient` trait which slices the jar to a specific storage (e.g. a specific cache).
+/// We don't need this just yet because we write our queries by hand. We may want a similar trait if we decide
+/// to use a macro to generate the queries.
+pub trait HasJar<T> {
+    /// Gives a read-only reference to the jar.
+    fn jar(&self) -> QueryResult<&T>;
+
+    /// Gives a mutable reference to the jar.
+    fn jar_mut(&mut self) -> &mut T;
+}
+
+/// Gives access to the jars in a database.
+pub trait HasJars {
+    /// A type storing the jars.
+    ///
+    /// Most commonly, this is a tuple where each jar is a tuple element.
+    type Jars: Default;
+
+    /// Gives access to the underlying jars but tests if the queries have been cancelled.
+    ///
+    /// Returns `Err(QueryError::Cancelled)` if the queries have been cancelled.
+    fn jars(&self) -> QueryResult<&Self::Jars>;
+
+    /// Gives mutable access to the underlying jars.
+    fn jars_mut(&mut self) -> &mut Self::Jars;
+}

crates/red_knot/src/db/query.rs

@@ -0,0 +1,20 @@
+use std::fmt::{Display, Formatter};
+
+/// Reason why a db query operation failed.
+#[derive(Debug, Clone, Copy)]
+pub enum QueryError {
+    /// The query was cancelled because the DB was mutated or the query was cancelled by the host (e.g. on a file change or when pressing CTRL+C).
+    Cancelled,
+}
+
+impl Display for QueryError {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        match self {
+            QueryError::Cancelled => f.write_str("query was cancelled"),
+        }
+    }
+}
+
+impl std::error::Error for QueryError {}
+
+pub type QueryResult<T> = Result<T, QueryError>;

crates/red_knot/src/db/runtime.rs

@@ -0,0 +1,41 @@
+use crate::cancellation::CancellationTokenSource;
+use crate::db::{QueryError, QueryResult};
+
+/// Holds the jar agnostic state of the database.
+#[derive(Debug, Default)]
+pub struct DbRuntime {
+    /// The cancellation token source used to signal other works that the queries should be aborted and
+    /// exit at the next possible point.
+    cancellation_token: CancellationTokenSource,
+}
+
+impl DbRuntime {
+    pub(super) fn snapshot(&self) -> Self {
+        Self {
+            cancellation_token: self.cancellation_token.clone(),
+        }
+    }
+
+    /// Cancels the pending queries of other workers. The current worker cannot have any pending
+    /// queries because we're holding a mutable reference to the runtime.
+    pub(super) fn cancel_other_workers(&mut self) {
+        self.cancellation_token.cancel();
+        // Set a new cancellation token so that we're in a non-cancelled state again when running the next
+        // query.
+        self.cancellation_token = CancellationTokenSource::default();
+    }
+
+    /// Returns `Ok` if the queries have not been cancelled and `Err(QueryError::Cancelled)` otherwise.
+    pub(super) fn cancelled(&self) -> QueryResult<()> {
+        if self.cancellation_token.is_cancelled() {
+            Err(QueryError::Cancelled)
+        } else {
+            Ok(())
+        }
+    }
+
+    /// Returns `true` if the queries have been cancelled.
+    pub(super) fn is_cancelled(&self) -> bool {
+        self.cancellation_token.is_cancelled()
+    }
+}

crates/red_knot/src/db/storage.rs

@@ -0,0 +1,117 @@
+use std::fmt::Formatter;
+use std::sync::Arc;
+
+use crossbeam::sync::WaitGroup;
+
+use crate::db::query::QueryResult;
+use crate::db::runtime::DbRuntime;
+use crate::db::{HasJars, ParallelDatabase};
+
+/// Stores the jars of a database and the state for each worker.
+///
+/// Today, all state is shared across all workers, but it may be desired to store data per worker in the future.
+pub struct JarsStorage<T>
+where
+    T: HasJars + Sized,
+{
+    // It's important that `jars_wait_group` is declared after `jars` to ensure that `jars` is dropped first.
+    // See https://doc.rust-lang.org/reference/destructors.html
+    /// Stores the jars of the database.
+    jars: Arc<T::Jars>,
+
+    /// Used to count the references to `jars`. Allows implementing `jars_mut` without requiring to clone `jars`.
+    jars_wait_group: WaitGroup,
+
+    /// The data agnostic state.
+    runtime: DbRuntime,
+}
+
+impl<Db> JarsStorage<Db>
+where
+    Db: HasJars,
+{
+    pub(super) fn new() -> Self {
+        Self {
+            jars: Arc::new(Db::Jars::default()),
+            jars_wait_group: WaitGroup::default(),
+            runtime: DbRuntime::default(),
+        }
+    }
+
+    /// Creates a snapshot of the jars.
+    ///
+    /// Creating the snapshot is cheap because it doesn't clone the jars, it only increments a ref counter.
+    #[must_use]
+    pub fn snapshot(&self) -> JarsStorage<Db>
+    where
+        Db: ParallelDatabase,
+    {
+        Self {
+            jars: self.jars.clone(),
+            jars_wait_group: self.jars_wait_group.clone(),
+            runtime: self.runtime.snapshot(),
+        }
+    }
+
+    pub(crate) fn jars(&self) -> QueryResult<&Db::Jars> {
+        self.runtime.cancelled()?;
+        Ok(&self.jars)
+    }
+
+    /// Returns a mutable reference to the jars without cloning their content.
+    ///
+    /// The method cancels any pending queries of other works and waits for them to complete so that
+    /// this instance is the only instance holding a reference to the jars.
+    pub(crate) fn jars_mut(&mut self) -> &mut Db::Jars {
+        // We have a mutable ref here, so no more workers can be spawned between calling this function and taking the mut ref below.
+        self.cancel_other_workers();
+
+        // Now all other references to `self.jars` should have been released. We can now safely return a mutable reference
+        // to the Arc's content.
+        let jars =
+            Arc::get_mut(&mut self.jars).expect("All references to jars should have been released");
+
+        jars
+    }
+
+    pub(crate) fn runtime(&self) -> &DbRuntime {
+        &self.runtime
+    }
+
+    pub(crate) fn runtime_mut(&mut self) -> &mut DbRuntime {
+        // Note: This method may need to use a similar trick to `jars_mut` if `DbRuntime` is ever to store data that is shared between workers.
+        &mut self.runtime
+    }
+
+    #[tracing::instrument(level = "trace", skip(self))]
+    fn cancel_other_workers(&mut self) {
+        self.runtime.cancel_other_workers();
+
+        // Wait for all other works to complete.
+        let existing_wait = std::mem::take(&mut self.jars_wait_group);
+        existing_wait.wait();
+    }
+}
+
+impl<Db> Default for JarsStorage<Db>
+where
+    Db: HasJars,
+{
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl<T> std::fmt::Debug for JarsStorage<T>
+where
+    T: HasJars,
+    <T as HasJars>::Jars: std::fmt::Debug,
+{
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("SharedStorage")
+            .field("jars", &self.jars)
+            .field("jars_wait_group", &self.jars_wait_group)
+            .field("runtime", &self.runtime)
+            .finish()
+    }
+}

crates/red_knot/src/files.rs

@@ -15,9 +15,9 @@ type Map<K, V> = hashbrown::HashMap<K, V, ()>;
 pub struct FileId;
 
 // TODO we'll need a higher level virtual file system abstraction that allows testing if a file exists
-// or retrieving its content (ideally lazily and in a way that the memory can be retained later)
-// I suspect that we'll end up with a FileSystem trait and our own Path abstraction.
-#[derive(Clone, Default)]
+//  or retrieving its content (ideally lazily and in a way that the memory can be retained later)
+//  I suspect that we'll end up with a FileSystem trait and our own Path abstraction.
+#[derive(Default)]
 pub struct Files {
     inner: Arc<RwLock<FilesInner>>,
 }
@@ -36,6 +36,16 @@ impl Files {
     pub fn path(&self, id: FileId) -> Arc<Path> {
         self.inner.read().path(id)
     }
+
+    /// Snapshots files for a new database snapshot.
+    ///
+    /// This method should not be used outside a database snapshot.
+    #[must_use]
+    pub fn snapshot(&self) -> Files {
+        Files {
+            inner: self.inner.clone(),
+        }
+    }
 }
 
 impl Debug for Files {
@@ -63,7 +73,7 @@ struct FilesInner {
     by_path: Map<FileId, ()>,
     // TODO should we use a map here to reclaim the space for removed files?
     // TODO I think we should use our own path abstraction here to avoid having to normalize paths
-    // and dealing with non-utf paths everywhere.
+    //  and dealing with non-utf paths everywhere.
     by_id: IndexVec<FileId, Arc<Path>>,
 }
 
@@ -71,9 +81,7 @@ impl FilesInner {
     /// Inserts the path and returns a new id for it or returns the id if it is an existing path.
     // TODO should this accept Path or PathBuf?
     pub(crate) fn intern(&mut self, path: &Path) -> FileId {
-        let mut hasher = FxHasher::default();
-        path.hash(&mut hasher);
-        let hash = hasher.finish();
+        let hash = FilesInner::hash_path(path);
 
         let entry = self
             .by_path
@@ -84,12 +92,20 @@ impl FilesInner {
             RawEntryMut::Occupied(entry) => *entry.key(),
             RawEntryMut::Vacant(entry) => {
                 let id = self.by_id.push(Arc::from(path));
-                entry.insert_with_hasher(hash, id, (), |_| hash);
+                entry.insert_with_hasher(hash, id, (), |file| {
+                    FilesInner::hash_path(&self.by_id[*file])
+                });
                 id
             }
         }
     }
 
+    fn hash_path(path: &Path) -> u64 {
+        let mut hasher = FxHasher::default();
+        path.hash(&mut hasher);
+        hasher.finish()
+    }
+
     pub(crate) fn try_get(&self, path: &Path) -> Option<FileId> {
         let mut hasher = FxHasher::default();
         path.hash(&mut hasher);
@@ -145,4 +161,20 @@ mod tests {
         let id2 = files.intern(&path2);
         assert_ne!(id1, id2);
     }
+
+    #[test]
+    fn four_files() {
+        let files = Files::default();
+        let foo_path = PathBuf::from("foo");
+        let foo_id = files.intern(&foo_path);
+        let bar_path = PathBuf::from("bar");
+        files.intern(&bar_path);
+        let baz_path = PathBuf::from("baz");
+        files.intern(&baz_path);
+        let qux_path = PathBuf::from("qux");
+        files.intern(&qux_path);
+
+        let foo_id_2 = files.try_get(&foo_path).expect("foo_path to be found");
+        assert_eq!(foo_id_2, foo_id);
+    }
 }

crates/red_knot/src/hir.rs

@@ -1,6 +1,6 @@
 //! Key observations
 //!
-//! The HIR avoids allocations to large extends by:
+//! The HIR (High-Level Intermediate Representation) avoids allocations to large extends by:
 //! * Using an arena per node type
 //! * using ids and id ranges to reference items.
 //!

crates/red_knot/src/lib.rs

@@ -1,3 +1,4 @@
+use std::fmt::Formatter;
 use std::hash::BuildHasherDefault;
 use std::ops::Deref;
 use std::path::{Path, PathBuf};
@@ -16,9 +17,8 @@ pub mod lint;
 pub mod module;
 mod parse;
 pub mod program;
+mod semantic;
 pub mod source;
-mod symbols;
-mod types;
 pub mod watch;
 
 pub(crate) type FxDashMap<K, V> = dashmap::DashMap<K, V, BuildHasherDefault<FxHasher>>;
@@ -26,7 +26,7 @@ pub(crate) type FxDashMap<K, V> = dashmap::DashMap<K, V, BuildHasherDefault<FxHa
 pub(crate) type FxDashSet<V> = dashmap::DashSet<V, BuildHasherDefault<FxHasher>>;
 pub(crate) type FxIndexSet<V> = indexmap::set::IndexSet<V, BuildHasherDefault<FxHasher>>;
 
-#[derive(Debug)]
+#[derive(Debug, Clone)]
 pub struct Workspace {
     /// TODO this should be a resolved path. We should probably use a newtype wrapper that guarantees that
     /// PATH is a UTF-8 path and is normalized.
@@ -100,3 +100,9 @@ where
         Self(value.into())
     }
 }
+
+impl std::fmt::Display for Name {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        f.write_str(self.as_str())
+    }
+}

crates/red_knot/src/lint.rs

@@ -1,30 +1,47 @@
+use std::cell::RefCell;
 use std::ops::{Deref, DerefMut};
 use std::sync::Arc;
+use std::time::Duration;
 
 use ruff_python_ast::visitor::Visitor;
-use ruff_python_ast::StringLiteral;
+use ruff_python_ast::{ModModule, StringLiteral};
+use ruff_python_parser::Parsed;
 
 use crate::cache::KeyValueCache;
-use crate::db::{HasJar, SourceDb, SourceJar};
+use crate::db::{LintDb, LintJar, QueryResult};
 use crate::files::FileId;
+use crate::module::{resolve_module, ModuleName};
+use crate::parse::parse;
+use crate::semantic::{infer_definition_type, infer_symbol_public_type, Type};
+use crate::semantic::{
+    resolve_global_symbol, semantic_index, Definition, GlobalSymbolId, SemanticIndex, SymbolId,
+};
+use crate::source::{source_text, Source};
 
 #[tracing::instrument(level = "debug", skip(db))]
-pub(crate) fn lint_syntax<Db>(db: &Db, file_id: FileId) -> Diagnostics
-where
-    Db: SourceDb + HasJar<SourceJar>,
-{
-    let storage = &db.jar().lint_syntax;
+pub(crate) fn lint_syntax(db: &dyn LintDb, file_id: FileId) -> QueryResult<Diagnostics> {
+    let lint_jar: &LintJar = db.jar()?;
+    let storage = &lint_jar.lint_syntax;
+
+    #[allow(clippy::print_stdout)]
+    if std::env::var("RED_KNOT_SLOW_LINT").is_ok() {
+        for i in 0..10 {
+            db.cancelled()?;
+            println!("RED_KNOT_SLOW_LINT is set, sleeping for {i}/10 seconds");
+            std::thread::sleep(Duration::from_secs(1));
+        }
+    }
 
     storage.get(&file_id, |file_id| {
         let mut diagnostics = Vec::new();
 
-        let source = db.source(*file_id);
+        let source = source_text(db.upcast(), *file_id)?;
         lint_lines(source.text(), &mut diagnostics);
 
-        let parsed = db.parse(*file_id);
+        let parsed = parse(db.upcast(), *file_id)?;
 
         if parsed.errors().is_empty() {
-            let ast = parsed.ast();
+            let ast = parsed.syntax();
 
             let mut visitor = SyntaxLintVisitor {
                 diagnostics,
@@ -36,11 +53,11 @@ where
             diagnostics.extend(parsed.errors().iter().map(std::string::ToString::to_string));
         }
 
-        Diagnostics::from(diagnostics)
+        Ok(Diagnostics::from(diagnostics))
     })
 }
 
-pub(crate) fn lint_lines(source: &str, diagnostics: &mut Vec<String>) {
+fn lint_lines(source: &str, diagnostics: &mut Vec<String>) {
     for (line_number, line) in source.lines().enumerate() {
         if line.len() < 88 {
             continue;
@@ -57,6 +74,179 @@ pub(crate) fn lint_lines(source: &str, diagnostics: &mut Vec<String>) {
     }
 }
 
+#[tracing::instrument(level = "debug", skip(db))]
+pub(crate) fn lint_semantic(db: &dyn LintDb, file_id: FileId) -> QueryResult<Diagnostics> {
+    let lint_jar: &LintJar = db.jar()?;
+    let storage = &lint_jar.lint_semantic;
+
+    storage.get(&file_id, |file_id| {
+        let source = source_text(db.upcast(), *file_id)?;
+        let parsed = parse(db.upcast(), *file_id)?;
+        let semantic_index = semantic_index(db.upcast(), *file_id)?;
+
+        let context = SemanticLintContext {
+            file_id: *file_id,
+            source,
+            parsed: &parsed,
+            semantic_index,
+            db,
+            diagnostics: RefCell::new(Vec::new()),
+        };
+
+        lint_unresolved_imports(&context)?;
+        lint_bad_overrides(&context)?;
+
+        Ok(Diagnostics::from(context.diagnostics.take()))
+    })
+}
+
+fn lint_unresolved_imports(context: &SemanticLintContext) -> QueryResult<()> {
+    // TODO: Consider iterating over the dependencies (imports) only instead of all definitions.
+    for (symbol, definition) in context.semantic_index().symbol_table().all_definitions() {
+        match definition {
+            Definition::Import(import) => {
+                let ty = context.infer_symbol_public_type(symbol)?;
+
+                if ty.is_unknown() {
+                    context.push_diagnostic(format!("Unresolved module {}", import.module));
+                }
+            }
+            Definition::ImportFrom(import) => {
+                let ty = context.infer_symbol_public_type(symbol)?;
+
+                if ty.is_unknown() {
+                    let module_name = import.module().map(Deref::deref).unwrap_or_default();
+                    let message = if import.level() > 0 {
+                        format!(
+                            "Unresolved relative import '{}' from {}{}",
+                            import.name(),
+                            ".".repeat(import.level() as usize),
+                            module_name
+                        )
+                    } else {
+                        format!(
+                            "Unresolved import '{}' from '{}'",
+                            import.name(),
+                            module_name
+                        )
+                    };
+
+                    context.push_diagnostic(message);
+                }
+            }
+            _ => {}
+        }
+    }
+
+    Ok(())
+}
+
+fn lint_bad_overrides(context: &SemanticLintContext) -> QueryResult<()> {
+    // TODO we should have a special marker on the real typing module (from typeshed) so if you
+    // have your own "typing" module in your project, we don't consider it THE typing module (and
+    // same for other stdlib modules that our lint rules care about)
+    let Some(typing_override) = context.resolve_global_symbol("typing", "override")? else {
+        // TODO once we bundle typeshed, this should be unreachable!()
+        return Ok(());
+    };
+
+    // TODO we should maybe index definitions by type instead of iterating all, or else iterate all
+    // just once, match, and branch to all lint rules that care about a type of definition
+    for (symbol, definition) in context.semantic_index().symbol_table().all_definitions() {
+        if !matches!(definition, Definition::FunctionDef(_)) {
+            continue;
+        }
+        let ty = infer_definition_type(
+            context.db.upcast(),
+            GlobalSymbolId {
+                file_id: context.file_id,
+                symbol_id: symbol,
+            },
+            definition.clone(),
+        )?;
+        let Type::Function(func) = ty else {
+            unreachable!("type of a FunctionDef should always be a Function");
+        };
+        let Some(class) = func.get_containing_class(context.db.upcast())? else {
+            // not a method of a class
+            continue;
+        };
+        if func.has_decorator(context.db.upcast(), typing_override)? {
+            let method_name = func.name(context.db.upcast())?;
+            if class
+                .get_super_class_member(context.db.upcast(), &method_name)?
+                .is_none()
+            {
+                // TODO should have a qualname() method to support nested classes
+                context.push_diagnostic(
+                    format!(
+                        "Method {}.{} is decorated with `typing.override` but does not override any base class method",
+                        class.name(context.db.upcast())?,
+                        method_name,
+                    ));
+            }
+        }
+    }
+    Ok(())
+}
+
+pub struct SemanticLintContext<'a> {
+    file_id: FileId,
+    source: Source,
+    parsed: &'a Parsed<ModModule>,
+    semantic_index: Arc<SemanticIndex>,
+    db: &'a dyn LintDb,
+    diagnostics: RefCell<Vec<String>>,
+}
+
+impl<'a> SemanticLintContext<'a> {
+    pub fn source_text(&self) -> &str {
+        self.source.text()
+    }
+
+    pub fn file_id(&self) -> FileId {
+        self.file_id
+    }
+
+    pub fn ast(&self) -> &'a ModModule {
+        self.parsed.syntax()
+    }
+
+    pub fn semantic_index(&self) -> &SemanticIndex {
+        &self.semantic_index
+    }
+
+    pub fn infer_symbol_public_type(&self, symbol_id: SymbolId) -> QueryResult<Type> {
+        infer_symbol_public_type(
+            self.db.upcast(),
+            GlobalSymbolId {
+                file_id: self.file_id,
+                symbol_id,
+            },
+        )
+    }
+
+    pub fn push_diagnostic(&self, diagnostic: String) {
+        self.diagnostics.borrow_mut().push(diagnostic);
+    }
+
+    pub fn extend_diagnostics(&mut self, diagnostics: impl IntoIterator<Item = String>) {
+        self.diagnostics.get_mut().extend(diagnostics);
+    }
+
+    pub fn resolve_global_symbol(
+        &self,
+        module: &str,
+        symbol_name: &str,
+    ) -> QueryResult<Option<GlobalSymbolId>> {
+        let Some(module) = resolve_module(self.db.upcast(), ModuleName::new(module))? else {
+            return Ok(None);
+        };
+
+        resolve_global_symbol(self.db.upcast(), module, symbol_name)
+    }
+}
+
 #[derive(Debug)]
 struct SyntaxLintVisitor<'a> {
     diagnostics: Vec<String>,
@@ -123,3 +313,20 @@ impl DerefMut for LintSyntaxStorage {
         &mut self.0
     }
 }
+
+#[derive(Default, Debug)]
+pub struct LintSemanticStorage(KeyValueCache<FileId, Diagnostics>);
+
+impl Deref for LintSemanticStorage {
+    type Target = KeyValueCache<FileId, Diagnostics>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+impl DerefMut for LintSemanticStorage {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.0
+    }
+}

crates/red_knot/src/main.rs

@@ -1,10 +1,9 @@
 #![allow(clippy::dbg_macro)]
 
-use std::collections::hash_map::Entry;
 use std::path::Path;
 use std::sync::Mutex;
 
-use rustc_hash::FxHashMap;
+use crossbeam::channel as crossbeam_channel;
 use tracing::subscriber::Interest;
 use tracing::{Level, Metadata};
 use tracing_subscriber::filter::LevelFilter;
@@ -12,12 +11,10 @@ use tracing_subscriber::layer::{Context, Filter, SubscriberExt};
 use tracing_subscriber::{Layer, Registry};
 use tracing_tree::time::Uptime;
 
-use red_knot::cancellation::CancellationTokenSource;
-use red_knot::db::{HasJar, SourceDb, SourceJar};
-use red_knot::files::FileId;
-use red_knot::module::{ModuleSearchPath, ModuleSearchPathKind};
-use red_knot::program::check::{CheckError, RayonCheckScheduler};
-use red_knot::program::{FileChange, FileChangeKind, Program};
+use red_knot::db::{HasJar, ParallelDatabase, QueryError, SourceDb, SourceJar};
+use red_knot::module::{set_module_search_paths, ModuleResolutionInputs};
+use red_knot::program::check::ExecutionMode;
+use red_knot::program::{FileWatcherChange, Program};
 use red_knot::watch::FileWatcher;
 use red_knot::Workspace;
 
@@ -47,11 +44,17 @@ fn main() -> anyhow::Result<()> {
     let workspace_folder = entry_point.parent().unwrap();
     let workspace = Workspace::new(workspace_folder.to_path_buf());
 
-    let workspace_search_path = ModuleSearchPath::new(
-        workspace.root().to_path_buf(),
-        ModuleSearchPathKind::FirstParty,
-    );
-    let mut program = Program::new(workspace, vec![workspace_search_path]);
+    let workspace_search_path = workspace.root().to_path_buf();
+
+    let search_paths = ModuleResolutionInputs {
+        extra_paths: vec![],
+        workspace_root: workspace_search_path,
+        site_packages: None,
+        custom_typeshed: None,
+    };
+
+    let mut program = Program::new(workspace);
+    set_module_search_paths(&mut program, search_paths);
 
     let entry_id = program.file_id(entry_point);
     program.workspace_mut().open_file(entry_id);
@@ -71,18 +74,15 @@ fn main() -> anyhow::Result<()> {
     let file_changes_notifier = main_loop.file_changes_notifier();
 
     // Watch for file changes and re-trigger the analysis.
-    let mut file_watcher = FileWatcher::new(
-        move |changes| {
-            file_changes_notifier.notify(changes);
-        },
-        program.files().clone(),
-    )?;
+    let mut file_watcher = FileWatcher::new(move |changes| {
+        file_changes_notifier.notify(changes);
+    })?;
 
     file_watcher.watch_folder(workspace_folder)?;
 
     main_loop.run(&mut program);
 
-    let source_jar: &SourceJar = program.jar();
+    let source_jar: &SourceJar = program.jar().unwrap();
 
     dbg!(source_jar.parsed.statistics());
     dbg!(source_jar.sources.statistics());
@@ -101,10 +101,9 @@ impl MainLoop {
         let (main_loop_sender, main_loop_receiver) = crossbeam_channel::bounded(1);
 
         let mut orchestrator = Orchestrator {
-            pending_analysis: None,
             receiver: orchestrator_receiver,
             sender: main_loop_sender.clone(),
-            aggregated_changes: AggregatedChanges::default(),
+            revision: 0,
         };
 
         std::thread::spawn(move || {
@@ -137,35 +136,27 @@ impl MainLoop {
             tracing::trace!("Main Loop: Tick");
 
             match message {
-                MainLoopMessage::CheckProgram => {
-                    // Remove mutability from program.
-                    let program = &*program;
-                    let run_cancellation_token_source = CancellationTokenSource::new();
-                    let run_cancellation_token = run_cancellation_token_source.token();
-                    let sender = &self.orchestrator_sender;
+                MainLoopMessage::CheckProgram { revision } => {
+                    let program = program.snapshot();
+                    let sender = self.orchestrator_sender.clone();
 
-                    sender
-                        .send(OrchestratorMessage::CheckProgramStarted {
-                            cancellation_token: run_cancellation_token_source,
-                        })
-                        .unwrap();
-
-                    rayon::in_place_scope(|scope| {
-                        let scheduler = RayonCheckScheduler::new(program, scope);
-
-                        let result = program.check(&scheduler, run_cancellation_token);
-                        match result {
-                            Ok(result) => sender
-                                .send(OrchestratorMessage::CheckProgramCompleted(result))
-                                .unwrap(),
-                            Err(CheckError::Cancelled) => sender
-                                .send(OrchestratorMessage::CheckProgramCancelled)
-                                .unwrap(),
+                    // Spawn a new task that checks the program. This needs to be done in a separate thread
+                    // to prevent blocking the main loop here.
+                    rayon::spawn(move || match program.check(ExecutionMode::ThreadPool) {
+                        Ok(result) => {
+                            sender
+                                .send(OrchestratorMessage::CheckProgramCompleted {
+                                    diagnostics: result,
+                                    revision,
+                                })
+                                .unwrap();
                         }
+                        Err(QueryError::Cancelled) => {}
                     });
                 }
                 MainLoopMessage::ApplyChanges(changes) => {
-                    program.apply_changes(changes.iter());
+                    // Automatically cancels any pending queries and waits for them to complete.
+                    program.apply_changes(changes);
                 }
                 MainLoopMessage::CheckCompleted(diagnostics) => {
                     dbg!(diagnostics);
@@ -192,7 +183,7 @@ struct FileChangesNotifier {
 }
 
 impl FileChangesNotifier {
-    fn notify(&self, changes: Vec<FileChange>) {
+    fn notify(&self, changes: Vec<FileWatcherChange>) {
         self.sender
             .send(OrchestratorMessage::FileChanges(changes))
             .unwrap();
@@ -211,13 +202,11 @@ impl MainLoopCancellationToken {
 }
 
 struct Orchestrator {
-    aggregated_changes: AggregatedChanges,
-    pending_analysis: Option<PendingAnalysisState>,
-
     /// Sends messages to the main loop.
     sender: crossbeam_channel::Sender<MainLoopMessage>,
     /// Receives messages from the main loop.
     receiver: crossbeam_channel::Receiver<OrchestratorMessage>,
+    revision: usize,
 }
 
 impl Orchestrator {
@@ -225,51 +214,33 @@ impl Orchestrator {
         while let Ok(message) = self.receiver.recv() {
             match message {
                 OrchestratorMessage::Run => {
-                    self.pending_analysis = None;
-                    self.sender.send(MainLoopMessage::CheckProgram).unwrap();
-                }
-
-                OrchestratorMessage::CheckProgramStarted { cancellation_token } => {
-                    debug_assert!(self.pending_analysis.is_none());
-
-                    self.pending_analysis = Some(PendingAnalysisState { cancellation_token });
-                }
-
-                OrchestratorMessage::CheckProgramCompleted(diagnostics) => {
-                    self.pending_analysis
-                        .take()
-                        .expect("Expected a pending analysis.");
-
                     self.sender
-                        .send(MainLoopMessage::CheckCompleted(diagnostics))
+                        .send(MainLoopMessage::CheckProgram {
+                            revision: self.revision,
+                        })
                         .unwrap();
                 }
 
-                OrchestratorMessage::CheckProgramCancelled => {
-                    self.pending_analysis
-                        .take()
-                        .expect("Expected a pending analysis.");
-
-                    self.debounce_changes();
+                OrchestratorMessage::CheckProgramCompleted {
+                    diagnostics,
+                    revision,
+                } => {
+                    // Only take the diagnostics if they are for the latest revision.
+                    if self.revision == revision {
+                        self.sender
+                            .send(MainLoopMessage::CheckCompleted(diagnostics))
+                            .unwrap();
+                    } else {
+                        tracing::debug!("Discarding diagnostics for outdated revision {revision} (current: {}).", self.revision);
+                    }
                 }
 
                 OrchestratorMessage::FileChanges(changes) => {
                     // Request cancellation, but wait until all analysis tasks have completed to
                     // avoid stale messages in the next main loop.
-                    let pending = if let Some(pending_state) = self.pending_analysis.as_ref() {
-                        pending_state.cancellation_token.cancel();
-                        true
-                    } else {
-                        false
-                    };
 
-                    self.aggregated_changes.extend(changes);
-
-                    // If there are no pending analysis tasks, apply the file changes. Otherwise
-                    // keep running until all file checks have completed.
-                    if !pending {
-                        self.debounce_changes();
-                    }
+                    self.revision += 1;
+                    self.debounce_changes(changes);
                 }
                 OrchestratorMessage::Shutdown => {
                     return self.shutdown();
@@ -278,9 +249,7 @@ impl Orchestrator {
         }
     }
 
-    fn debounce_changes(&mut self) {
-        debug_assert!(self.pending_analysis.is_none());
-
+    fn debounce_changes(&self, mut changes: Vec<FileWatcherChange>) {
         loop {
             // Consume possibly incoming file change messages before running a new analysis, but don't wait for more than 100ms.
             crossbeam_channel::select! {
@@ -290,10 +259,12 @@ impl Orchestrator {
                             return self.shutdown();
                         }
                         Ok(OrchestratorMessage::FileChanges(file_changes)) => {
-                            self.aggregated_changes.extend(file_changes);
+                            changes.extend(file_changes);
                         }
 
-                        Ok(OrchestratorMessage::CheckProgramStarted {..}| OrchestratorMessage::CheckProgramCompleted(_) | OrchestratorMessage::CheckProgramCancelled) => unreachable!("No program check should be running while debouncing changes."),
+                        Ok(OrchestratorMessage::CheckProgramCompleted { .. })=> {
+                            // disregard any outdated completion message.
+                        }
                         Ok(OrchestratorMessage::Run) => unreachable!("The orchestrator is already running."),
 
                         Err(_) => {
@@ -302,10 +273,10 @@ impl Orchestrator {
                         }
                     }
                 },
-                default(std::time::Duration::from_millis(100)) => {
-                    // No more file changes after 100 ms, send the changes and schedule a new analysis
-                    self.sender.send(MainLoopMessage::ApplyChanges(std::mem::take(&mut self.aggregated_changes))).unwrap();
-                    self.sender.send(MainLoopMessage::CheckProgram).unwrap();
+                default(std::time::Duration::from_millis(10)) => {
+                    // No more file changes after 10 ms, send the changes and schedule a new analysis
+                    self.sender.send(MainLoopMessage::ApplyChanges(changes)).unwrap();
+                    self.sender.send(MainLoopMessage::CheckProgram { revision: self.revision}).unwrap();
                     return;
                 }
             }
@@ -318,17 +289,12 @@ impl Orchestrator {
     }
 }
 
-#[derive(Debug)]
-struct PendingAnalysisState {
-    cancellation_token: CancellationTokenSource,
-}
-
 /// Message sent from the orchestrator to the main loop.
 #[derive(Debug)]
 enum MainLoopMessage {
-    CheckProgram,
+    CheckProgram { revision: usize },
     CheckCompleted(Vec<String>),
-    ApplyChanges(AggregatedChanges),
+    ApplyChanges(Vec<FileWatcherChange>),
     Exit,
 }
 
@@ -337,83 +303,12 @@ enum OrchestratorMessage {
     Run,
     Shutdown,
 
-    CheckProgramStarted {
-        cancellation_token: CancellationTokenSource,
+    CheckProgramCompleted {
+        diagnostics: Vec<String>,
+        revision: usize,
     },
-    CheckProgramCompleted(Vec<String>),
-    CheckProgramCancelled,
 
-    FileChanges(Vec<FileChange>),
-}
-
-#[derive(Default, Debug)]
-struct AggregatedChanges {
-    changes: FxHashMap<FileId, FileChangeKind>,
-}
-
-impl AggregatedChanges {
-    fn add(&mut self, change: FileChange) {
-        match self.changes.entry(change.file_id()) {
-            Entry::Occupied(mut entry) => {
-                let merged = entry.get_mut();
-
-                match (merged, change.kind()) {
-                    (FileChangeKind::Created, FileChangeKind::Deleted) => {
-                        // Deletion after creations means that ruff never saw the file.
-                        entry.remove();
-                    }
-                    (FileChangeKind::Created, FileChangeKind::Modified) => {
-                        // No-op, for ruff, modifying a file that it doesn't yet know that it exists is still considered a creation.
-                    }
-
-                    (FileChangeKind::Modified, FileChangeKind::Created) => {
-                        // Uhh, that should probably not happen. Continue considering it a modification.
-                    }
-
-                    (FileChangeKind::Modified, FileChangeKind::Deleted) => {
-                        *entry.get_mut() = FileChangeKind::Deleted;
-                    }
-
-                    (FileChangeKind::Deleted, FileChangeKind::Created) => {
-                        *entry.get_mut() = FileChangeKind::Modified;
-                    }
-
-                    (FileChangeKind::Deleted, FileChangeKind::Modified) => {
-                        // That's weird, but let's consider it a modification.
-                        *entry.get_mut() = FileChangeKind::Modified;
-                    }
-
-                    (FileChangeKind::Created, FileChangeKind::Created)
-                    | (FileChangeKind::Modified, FileChangeKind::Modified)
-                    | (FileChangeKind::Deleted, FileChangeKind::Deleted) => {
-                        // No-op transitions. Some of them should be impossible but we handle them anyway.
-                    }
-                }
-            }
-            Entry::Vacant(entry) => {
-                entry.insert(change.kind());
-            }
-        }
-    }
-
-    fn extend<I>(&mut self, changes: I)
-    where
-        I: IntoIterator<Item = FileChange>,
-        I::IntoIter: ExactSizeIterator,
-    {
-        let iter = changes.into_iter();
-        self.changes.reserve(iter.len());
-
-        for change in iter {
-            self.add(change);
-        }
-    }
-
-    fn iter(&self) -> impl Iterator<Item = FileChange> + '_ {
-        self.changes
-            .iter()
-            .map(|(id, kind)| FileChange::new(*id, *kind))
-    }
+    FileChanges(Vec<FileWatcherChange>),
 }
 
 fn setup_tracing() {

crates/red_knot/src/parse.rs

@@ -1,87 +1,33 @@
 use std::ops::{Deref, DerefMut};
 use std::sync::Arc;
 
-use ruff_python_ast as ast;
-use ruff_python_parser::{Mode, ParseError};
-use ruff_text_size::{Ranged, TextRange};
+use ruff_python_ast::ModModule;
+use ruff_python_parser::Parsed;
 
 use crate::cache::KeyValueCache;
-use crate::db::{HasJar, SourceDb, SourceJar};
+use crate::db::{QueryResult, SourceDb};
 use crate::files::FileId;
-
-#[derive(Debug, Clone, PartialEq)]
-pub struct Parsed {
-    inner: Arc<ParsedInner>,
-}
-
-#[derive(Debug, PartialEq)]
-struct ParsedInner {
-    ast: ast::ModModule,
-    errors: Vec<ParseError>,
-}
-
-impl Parsed {
-    fn new(ast: ast::ModModule, errors: Vec<ParseError>) -> Self {
-        Self {
-            inner: Arc::new(ParsedInner { ast, errors }),
-        }
-    }
-
-    pub(crate) fn from_text(text: &str) -> Self {
-        let result = ruff_python_parser::parse(text, Mode::Module);
-
-        let (module, errors) = match result {
-            Ok(ast::Mod::Module(module)) => (module, vec![]),
-            Ok(ast::Mod::Expression(expression)) => (
-                ast::ModModule {
-                    range: expression.range(),
-                    body: vec![ast::Stmt::Expr(ast::StmtExpr {
-                        range: expression.range(),
-                        value: expression.body,
-                    })],
-                },
-                vec![],
-            ),
-            Err(errors) => (
-                ast::ModModule {
-                    range: TextRange::default(),
-                    body: Vec::new(),
-                },
-                vec![errors],
-            ),
-        };
-
-        Parsed::new(module, errors)
-    }
-
-    pub fn ast(&self) -> &ast::ModModule {
-        &self.inner.ast
-    }
-
-    pub fn errors(&self) -> &[ParseError] {
-        &self.inner.errors
-    }
-}
+use crate::source::source_text;
 
 #[tracing::instrument(level = "debug", skip(db))]
-pub(crate) fn parse<Db>(db: &Db, file_id: FileId) -> Parsed
-where
-    Db: SourceDb + HasJar<SourceJar>,
-{
-    let parsed = db.jar();
+pub(crate) fn parse(db: &dyn SourceDb, file_id: FileId) -> QueryResult<Arc<Parsed<ModModule>>> {
+    let jar = db.jar()?;
 
-    parsed.parsed.get(&file_id, |file_id| {
-        let source = db.source(*file_id);
+    jar.parsed.get(&file_id, |file_id| {
+        let source = source_text(db, *file_id)?;
 
-        Parsed::from_text(source.text())
+        Ok(Arc::new(ruff_python_parser::parse_unchecked_source(
+            source.text(),
+            source.kind().into(),
+        )))
     })
 }
 
 #[derive(Debug, Default)]
-pub struct ParsedStorage(KeyValueCache<FileId, Parsed>);
+pub struct ParsedStorage(KeyValueCache<FileId, Arc<Parsed<ModModule>>>);
 
 impl Deref for ParsedStorage {
-    type Target = KeyValueCache<FileId, Parsed>;
+    type Target = KeyValueCache<FileId, Arc<Parsed<ModModule>>>;
 
     fn deref(&self) -> &Self::Target {
         &self.0

crates/red_knot/src/program/check.rs

@@ -1,66 +1,59 @@
-use crate::cancellation::CancellationToken;
-use crate::db::{SemanticDb, SourceDb};
-use crate::files::FileId;
-use crate::lint::Diagnostics;
-use crate::program::Program;
-use rayon::max_num_threads;
+use rayon::{current_num_threads, yield_local};
 use rustc_hash::FxHashSet;
-use std::num::NonZeroUsize;
+
+use crate::db::{Database, QueryError, QueryResult};
+use crate::files::FileId;
+use crate::lint::{lint_semantic, lint_syntax, Diagnostics};
+use crate::module::{file_to_module, resolve_module};
+use crate::program::Program;
+use crate::semantic::{semantic_index, Dependency};
 
 impl Program {
     /// Checks all open files in the workspace and its dependencies.
     #[tracing::instrument(level = "debug", skip_all)]
-    pub fn check(
-        &self,
-        scheduler: &dyn CheckScheduler,
-        cancellation_token: CancellationToken,
-    ) -> Result<Vec<String>, CheckError> {
-        let check_loop = CheckFilesLoop::new(scheduler, cancellation_token);
+    pub fn check(&self, mode: ExecutionMode) -> QueryResult<Vec<String>> {
+        self.cancelled()?;
 
-        check_loop.run(self.workspace().open_files.iter().copied())
-    }
+        let mut context = CheckContext::new(self);
 
-    /// Checks a single file and its dependencies.
-    #[tracing::instrument(level = "debug", skip(self, scheduler, cancellation_token))]
-    pub fn check_file(
-        &self,
-        file: FileId,
-        scheduler: &dyn CheckScheduler,
-        cancellation_token: CancellationToken,
-    ) -> Result<Vec<String>, CheckError> {
-        let check_loop = CheckFilesLoop::new(scheduler, cancellation_token);
+        match mode {
+            ExecutionMode::SingleThreaded => SingleThreadedExecutor.run(&mut context)?,
+            ExecutionMode::ThreadPool => ThreadPoolExecutor.run(&mut context)?,
+        };
 
-        check_loop.run([file].into_iter())
+        Ok(context.finish())
     }
 
     #[tracing::instrument(level = "debug", skip(self, context))]
-    fn do_check_file(
-        &self,
-        file: FileId,
-        context: &CheckContext,
-    ) -> Result<Diagnostics, CheckError> {
-        context.cancelled_ok()?;
+    fn check_file(&self, file: FileId, context: &CheckFileContext) -> QueryResult<Diagnostics> {
+        self.cancelled()?;
 
-        let symbol_table = self.symbol_table(file);
-        let dependencies = symbol_table.dependencies();
+        let index = semantic_index(self, file)?;
+        let dependencies = index.symbol_table().dependencies();
 
         if !dependencies.is_empty() {
-            let module = self.file_to_module(file);
+            let module = file_to_module(self, file)?;
 
             // TODO scheduling all dependencies here is wasteful if we don't infer any types on them
             //  but I think that's unlikely, so it is okay?
             //  Anyway, we need to figure out a way to retrieve the dependencies of a module
             //  from the persistent cache. So maybe it should be a separate query after all.
             for dependency in dependencies {
-                let dependency_name = dependency.module_name(self, module);
+                let dependency_name = match dependency {
+                    Dependency::Module(name) => Some(name.clone()),
+                    Dependency::Relative { .. } => match &module {
+                        Some(module) => module.resolve_dependency(self, dependency)?,
+                        None => None,
+                    },
+                };
 
                 if let Some(dependency_name) = dependency_name {
                     // TODO We may want to have a different check functions for non-first-party
                     //   files because we only need to index them and not check them.
                     //   Supporting non-first-party code also requires supporting typing stubs.
-                    if let Some(dependency) = self.resolve_module(dependency_name) {
-                        if dependency.path(self).root().kind().is_first_party() {
-                            context.schedule_check_file(dependency.path(self).file());
+                    if let Some(dependency) = resolve_module(self, dependency_name)? {
+                        if dependency.path(self)?.root().kind().is_first_party() {
+                            context.schedule_dependency(dependency.path(self)?.file());
                         }
                     }
                 }
@@ -70,237 +63,351 @@ impl Program {
         let mut diagnostics = Vec::new();
 
         if self.workspace().is_file_open(file) {
-            diagnostics.extend_from_slice(&self.lint_syntax(file));
+            diagnostics.extend_from_slice(&lint_syntax(self, file)?);
+            diagnostics.extend_from_slice(&lint_semantic(self, file)?);
         }
 
         Ok(Diagnostics::from(diagnostics))
     }
 }
 
-/// Schedules checks for files.
-pub trait CheckScheduler {
-    /// Schedules a check for a file.
-    ///
-    /// The check can either be run immediately on the current thread or the check can be queued
-    /// in a thread pool and ran asynchronously.
-    ///
-    /// The order in which scheduled checks are executed is not guaranteed.
-    ///
-    /// The implementation should call [`CheckFileTask::run`] to execute the check.
-    fn check_file(&self, file_task: CheckFileTask);
-
-    /// The maximum number of checks that can be run concurrently.
-    ///
-    /// Returns `None` if the checks run on the current thread (no concurrency).
-    fn max_concurrency(&self) -> Option<NonZeroUsize>;
+#[derive(Copy, Clone, Debug, PartialEq, Eq)]
+pub enum ExecutionMode {
+    SingleThreaded,
+    ThreadPool,
 }
 
-/// Scheduler that runs checks on a rayon thread pool.
-pub struct RayonCheckScheduler<'program, 'scope_ref, 'scope> {
-    program: &'program Program,
-    scope: &'scope_ref rayon::Scope<'scope>,
+/// Context that stores state information about the entire check operation.
+struct CheckContext<'a> {
+    /// IDs of the files that have been queued for checking.
+    ///
+    /// Used to avoid queuing the same file twice.
+    scheduled_files: FxHashSet<FileId>,
+
+    /// Reference to the program that is checked.
+    program: &'a Program,
+
+    /// The aggregated diagnostics
+    diagnostics: Vec<String>,
 }
 
-impl<'program, 'scope_ref, 'scope> RayonCheckScheduler<'program, 'scope_ref, 'scope> {
-    pub fn new(program: &'program Program, scope: &'scope_ref rayon::Scope<'scope>) -> Self {
-        Self { program, scope }
+impl<'a> CheckContext<'a> {
+    fn new(program: &'a Program) -> Self {
+        Self {
+            scheduled_files: FxHashSet::default(),
+            program,
+            diagnostics: Vec::new(),
+        }
+    }
+
+    /// Returns the tasks to check all open files in the workspace.
+    fn check_open_files(&mut self) -> Vec<CheckOpenFileTask> {
+        self.scheduled_files
+            .extend(self.program.workspace().open_files());
+
+        self.program
+            .workspace()
+            .open_files()
+            .map(|file_id| CheckOpenFileTask { file_id })
+            .collect()
+    }
+
+    /// Returns the task to check a dependency.
+    fn check_dependency(&mut self, file_id: FileId) -> Option<CheckDependencyTask> {
+        if self.scheduled_files.insert(file_id) {
+            Some(CheckDependencyTask { file_id })
+        } else {
+            None
+        }
+    }
+
+    /// Pushes the result for a single file check operation
+    fn push_diagnostics(&mut self, diagnostics: &Diagnostics) {
+        self.diagnostics.extend_from_slice(diagnostics);
+    }
+
+    /// Returns a reference to the program that is being checked.
+    fn program(&self) -> &'a Program {
+        self.program
+    }
+
+    /// Creates a task context that is used to check a single file.
+    fn task_context<'b, S>(&self, dependency_scheduler: &'b S) -> CheckTaskContext<'a, 'b, S>
+    where
+        S: ScheduleDependency,
+    {
+        CheckTaskContext {
+            program: self.program,
+            dependency_scheduler,
+        }
+    }
+
+    fn finish(self) -> Vec<String> {
+        self.diagnostics
     }
 }
 
-impl<'program, 'scope_ref, 'scope> CheckScheduler
-    for RayonCheckScheduler<'program, 'scope_ref, 'scope>
+/// Trait that abstracts away how a dependency of a file gets scheduled for checking.
+trait ScheduleDependency {
+    /// Schedules the file with the given ID for checking.
+    fn schedule(&self, file_id: FileId);
+}
+
+impl<T> ScheduleDependency for T
 where
-    'program: 'scope,
+    T: Fn(FileId),
 {
-    fn check_file(&self, check_file_task: CheckFileTask) {
-        let child_span =
-            tracing::trace_span!("check_file", file_id = check_file_task.file_id.as_u32());
-        let program = self.program;
-
-        self.scope
-            .spawn(move |_| child_span.in_scope(|| check_file_task.run(program)));
-    }
-
-    fn max_concurrency(&self) -> Option<NonZeroUsize> {
-        Some(NonZeroUsize::new(max_num_threads()).unwrap_or(NonZeroUsize::MIN))
+    fn schedule(&self, file_id: FileId) {
+        let f = self;
+        f(file_id);
     }
 }
 
-/// Scheduler that runs all checks on the current thread.
-pub struct SameThreadCheckScheduler<'a> {
+/// Context that is used to run a single file check task.
+///
+/// The task is generic over `S` because it is passed across thread boundaries and
+/// we don't want to add the requirement that [`ScheduleDependency`] must be [`Send`].
+struct CheckTaskContext<'a, 'scheduler, S>
+where
+    S: ScheduleDependency,
+{
+    dependency_scheduler: &'scheduler S,
     program: &'a Program,
 }
 
-impl<'a> SameThreadCheckScheduler<'a> {
-    pub fn new(program: &'a Program) -> Self {
-        Self { program }
+impl<'a, 'scheduler, S> CheckTaskContext<'a, 'scheduler, S>
+where
+    S: ScheduleDependency,
+{
+    fn as_file_context(&self) -> CheckFileContext<'scheduler> {
+        CheckFileContext {
+            dependency_scheduler: self.dependency_scheduler,
+        }
     }
 }
 
-impl CheckScheduler for SameThreadCheckScheduler<'_> {
-    fn check_file(&self, task: CheckFileTask) {
-        task.run(self.program);
-    }
-
-    fn max_concurrency(&self) -> Option<NonZeroUsize> {
-        None
-    }
+/// Context passed when checking a single file.
+///
+/// This is a trimmed down version of [`CheckTaskContext`] with the type parameter `S` erased
+/// to avoid monomorphization of [`Program:check_file`].
+struct CheckFileContext<'a> {
+    dependency_scheduler: &'a dyn ScheduleDependency,
 }
 
-#[derive(Debug, Clone)]
-pub enum CheckError {
-    Cancelled,
+impl<'a> CheckFileContext<'a> {
+    fn schedule_dependency(&self, file_id: FileId) {
+        self.dependency_scheduler.schedule(file_id);
+    }
 }
 
 #[derive(Debug)]
-pub struct CheckFileTask {
-    file_id: FileId,
-    context: CheckContext,
+enum CheckFileTask {
+    OpenFile(CheckOpenFileTask),
+    Dependency(CheckDependencyTask),
 }
 
 impl CheckFileTask {
-    /// Runs the check and communicates the result to the orchestrator.
-    pub fn run(self, program: &Program) {
-        match program.do_check_file(self.file_id, &self.context) {
-            Ok(diagnostics) => self
-                .context
-                .sender
-                .send(CheckFileMessage::Completed(diagnostics))
-                .unwrap(),
-            Err(CheckError::Cancelled) => self
-                .context
-                .sender
-                .send(CheckFileMessage::Cancelled)
-                .unwrap(),
+    /// Runs the task and returns the results for checking this file.
+    fn run<S>(&self, context: &CheckTaskContext<S>) -> QueryResult<Diagnostics>
+    where
+        S: ScheduleDependency,
+    {
+        match self {
+            Self::OpenFile(task) => task.run(context),
+            Self::Dependency(task) => task.run(context),
+        }
+    }
+
+    fn file_id(&self) -> FileId {
+        match self {
+            CheckFileTask::OpenFile(task) => task.file_id,
+            CheckFileTask::Dependency(task) => task.file_id,
         }
     }
 }
 
-#[derive(Clone, Debug)]
-struct CheckContext {
-    cancellation_token: CancellationToken,
-    sender: crossbeam_channel::Sender<CheckFileMessage>,
+/// Task to check an open file.
+
+#[derive(Debug)]
+struct CheckOpenFileTask {
+    file_id: FileId,
 }
 
-impl CheckContext {
-    fn new(
-        cancellation_token: CancellationToken,
-        sender: crossbeam_channel::Sender<CheckFileMessage>,
-    ) -> Self {
-        Self {
-            cancellation_token,
-            sender,
-        }
-    }
-
-    /// Queues a new file for checking using the [`CheckScheduler`].
-    #[allow(unused)]
-    fn schedule_check_file(&self, file_id: FileId) {
-        self.sender.send(CheckFileMessage::Queue(file_id)).unwrap();
-    }
-
-    /// Returns `true` if the check has been cancelled.
-    fn is_cancelled(&self) -> bool {
-        self.cancellation_token.is_cancelled()
-    }
-
-    fn cancelled_ok(&self) -> Result<(), CheckError> {
-        if self.is_cancelled() {
-            Err(CheckError::Cancelled)
-        } else {
-            Ok(())
-        }
+impl CheckOpenFileTask {
+    fn run<S>(&self, context: &CheckTaskContext<S>) -> QueryResult<Diagnostics>
+    where
+        S: ScheduleDependency,
+    {
+        context
+            .program
+            .check_file(self.file_id, &context.as_file_context())
     }
 }
 
-struct CheckFilesLoop<'a> {
-    scheduler: &'a dyn CheckScheduler,
-    cancellation_token: CancellationToken,
-    pending: usize,
-    queued_files: FxHashSet<FileId>,
+/// Task to check a dependency file.
+#[derive(Debug)]
+struct CheckDependencyTask {
+    file_id: FileId,
 }
 
-impl<'a> CheckFilesLoop<'a> {
-    fn new(scheduler: &'a dyn CheckScheduler, cancellation_token: CancellationToken) -> Self {
-        Self {
-            scheduler,
-            cancellation_token,
-
-            queued_files: FxHashSet::default(),
-            pending: 0,
-        }
+impl CheckDependencyTask {
+    fn run<S>(&self, context: &CheckTaskContext<S>) -> QueryResult<Diagnostics>
+    where
+        S: ScheduleDependency,
+    {
+        context
+            .program
+            .check_file(self.file_id, &context.as_file_context())
     }
+}
 
-    fn run(mut self, files: impl Iterator<Item = FileId>) -> Result<Vec<String>, CheckError> {
-        let (sender, receiver) = if let Some(max_concurrency) = self.scheduler.max_concurrency() {
-            crossbeam_channel::bounded(max_concurrency.get())
-        } else {
-            // The checks run on the current thread. That means it is necessary to store all messages
-            // or we risk deadlocking when the main loop never gets a chance to read the messages.
-            crossbeam_channel::unbounded()
-        };
+/// Executor that schedules the checking of individual program files.
+trait CheckExecutor {
+    fn run(self, context: &mut CheckContext) -> QueryResult<()>;
+}
 
-        let context = CheckContext::new(self.cancellation_token.clone(), sender.clone());
+/// Executor that runs all check operations on the current thread.
+///
+/// The executor does not schedule dependencies for checking.
+/// The main motivation for scheduling dependencies
+/// in a multithreaded environment is to parse and index the dependencies concurrently.
+/// However, that doesn't make sense in a single threaded environment, because the dependencies then compute
+/// with checking the open files. Checking dependencies in a single threaded environment is more likely
+/// to hurt performance because we end up analyzing files in their entirety, even if we only need to type check parts of them.
+#[derive(Debug, Default)]
+struct SingleThreadedExecutor;
 
-        for file in files {
-            self.queue_file(file, context.clone())?;
-        }
+impl CheckExecutor for SingleThreadedExecutor {
+    fn run(self, context: &mut CheckContext) -> QueryResult<()> {
+        let mut queue = context.check_open_files();
 
-        self.run_impl(receiver, &context)
-    }
+        let noop_schedule_dependency = |_| {};
 
-    fn run_impl(
-        mut self,
-        receiver: crossbeam_channel::Receiver<CheckFileMessage>,
-        context: &CheckContext,
-    ) -> Result<Vec<String>, CheckError> {
-        if self.cancellation_token.is_cancelled() {
-            return Err(CheckError::Cancelled);
-        }
+        while let Some(file) = queue.pop() {
+            context.program().cancelled()?;
 
-        let mut result = Vec::default();
-
-        for message in receiver {
-            match message {
-                CheckFileMessage::Completed(diagnostics) => {
-                    result.extend_from_slice(&diagnostics);
-
-                    self.pending -= 1;
-
-                    if self.pending == 0 {
-                        break;
-                    }
-                }
-                CheckFileMessage::Queue(id) => {
-                    self.queue_file(id, context.clone())?;
-                }
-                CheckFileMessage::Cancelled => {
-                    return Err(CheckError::Cancelled);
-                }
-            }
-        }
-
-        Ok(result)
-    }
-
-    fn queue_file(&mut self, file_id: FileId, context: CheckContext) -> Result<(), CheckError> {
-        if context.is_cancelled() {
-            return Err(CheckError::Cancelled);
-        }
-
-        if self.queued_files.insert(file_id) {
-            self.pending += 1;
-
-            self.scheduler
-                .check_file(CheckFileTask { file_id, context });
+            let task_context = context.task_context(&noop_schedule_dependency);
+            context.push_diagnostics(&file.run(&task_context)?);
         }
 
         Ok(())
     }
 }
 
-enum CheckFileMessage {
-    Completed(Diagnostics),
-    Queue(FileId),
-    Cancelled,
+/// Executor that runs the check operations on a thread pool.
+///
+/// The executor runs each check operation as its own task using a thread pool.
+///
+/// Other than [`SingleThreadedExecutor`], this executor schedules dependencies for checking. It
+/// even schedules dependencies for checking when the thread pool size is 1 for a better debugging experience.
+#[derive(Debug, Default)]
+struct ThreadPoolExecutor;
+
+impl CheckExecutor for ThreadPoolExecutor {
+    fn run(self, context: &mut CheckContext) -> QueryResult<()> {
+        let num_threads = current_num_threads();
+        let single_threaded = num_threads == 1;
+        let span = tracing::trace_span!("ThreadPoolExecutor::run", num_threads);
+        let _ = span.enter();
+
+        let mut queue: Vec<_> = context
+            .check_open_files()
+            .into_iter()
+            .map(CheckFileTask::OpenFile)
+            .collect();
+
+        let (sender, receiver) = if single_threaded {
+            // Use an unbounded queue for single threaded execution to prevent deadlocks
+            // when a single file schedules multiple dependencies.
+            crossbeam::channel::unbounded()
+        } else {
+            // Use a bounded queue to apply backpressure when the orchestration thread isn't able to keep
+            // up processing messages from the worker threads.
+            crossbeam::channel::bounded(num_threads)
+        };
+
+        let schedule_sender = sender.clone();
+        let schedule_dependency = move |file_id| {
+            schedule_sender
+                .send(ThreadPoolMessage::ScheduleDependency(file_id))
+                .unwrap();
+        };
+
+        let result = rayon::in_place_scope(|scope| {
+            let mut pending = 0usize;
+
+            loop {
+                context.program().cancelled()?;
+
+                // 1. Try to get a queued message to ensure that we have always remaining space in the channel to prevent blocking the worker threads.
+                // 2. Try to process a queued file
+                // 3. If there's no queued file wait for the next incoming message.
+                // 4. Exit if there are no more messages and no senders.
+                let message = if let Ok(message) = receiver.try_recv() {
+                    message
+                } else if let Some(task) = queue.pop() {
+                    pending += 1;
+
+                    let task_context = context.task_context(&schedule_dependency);
+                    let sender = sender.clone();
+                    let task_span = tracing::trace_span!(
+                        parent: &span,
+                        "CheckFileTask::run",
+                        file_id = task.file_id().as_u32(),
+                    );
+
+                    scope.spawn(move |_| {
+                        task_span.in_scope(|| match task.run(&task_context) {
+                            Ok(result) => {
+                                sender.send(ThreadPoolMessage::Completed(result)).unwrap();
+                            }
+                            Err(err) => sender.send(ThreadPoolMessage::Errored(err)).unwrap(),
+                        });
+                    });
+
+                    // If this is a single threaded rayon thread pool, yield the current thread
+                    // or we never start processing the work items.
+                    if single_threaded {
+                        yield_local();
+                    }
+
+                    continue;
+                } else if let Ok(message) = receiver.recv() {
+                    message
+                } else {
+                    break;
+                };
+
+                match message {
+                    ThreadPoolMessage::ScheduleDependency(dependency) => {
+                        if let Some(task) = context.check_dependency(dependency) {
+                            queue.push(CheckFileTask::Dependency(task));
+                        }
+                    }
+                    ThreadPoolMessage::Completed(diagnostics) => {
+                        context.push_diagnostics(&diagnostics);
+                        pending -= 1;
+
+                        if pending == 0 && queue.is_empty() {
+                            break;
+                        }
+                    }
+                    ThreadPoolMessage::Errored(err) => {
+                        return Err(err);
+                    }
+                }
+            }
+
+            Ok(())
+        });
+
+        result
+    }
+}
+
+#[derive(Debug)]
+enum ThreadPoolMessage {
+    ScheduleDependency(FileId),
+    Completed(Diagnostics),
+    Errored(QueryError),
 }

crates/red_knot/src/program/mod.rs

@@ -1,42 +1,29 @@
-pub mod check;
-
-use std::path::Path;
+use std::collections::hash_map::Entry;
+use std::path::{Path, PathBuf};
 use std::sync::Arc;
 
-use crate::db::{Db, HasJar, SemanticDb, SemanticJar, SourceDb, SourceJar};
-use crate::files::{FileId, Files};
-use crate::lint::{lint_syntax, Diagnostics, LintSyntaxStorage};
-use crate::module::{
-    add_module, file_to_module, path_to_module, resolve_module, set_module_search_paths, Module,
-    ModuleData, ModuleName, ModuleResolver, ModuleSearchPath,
+use rustc_hash::FxHashMap;
+
+use crate::db::{
+    Database, Db, DbRuntime, DbWithJar, HasJar, HasJars, JarsStorage, LintDb, LintJar,
+    ParallelDatabase, QueryResult, SemanticDb, SemanticJar, Snapshot, SourceDb, SourceJar, Upcast,
 };
-use crate::parse::{parse, Parsed, ParsedStorage};
-use crate::source::{source_text, Source, SourceStorage};
-use crate::symbols::{symbol_table, SymbolId, SymbolTable, SymbolTablesStorage};
-use crate::types::{infer_symbol_type, Type, TypeStore};
+use crate::files::{FileId, Files};
 use crate::Workspace;
 
+pub mod check;
+
 #[derive(Debug)]
 pub struct Program {
+    jars: JarsStorage<Program>,
     files: Files,
-    source: SourceJar,
-    semantic: SemanticJar,
     workspace: Workspace,
 }
 
 impl Program {
-    pub fn new(workspace: Workspace, module_search_paths: Vec<ModuleSearchPath>) -> Self {
+    pub fn new(workspace: Workspace) -> Self {
         Self {
-            source: SourceJar {
-                sources: SourceStorage::default(),
-                parsed: ParsedStorage::default(),
-                lint_syntax: LintSyntaxStorage::default(),
-            },
-            semantic: SemanticJar {
-                module_resolver: ModuleResolver::new(module_search_paths),
-                symbol_tables: SymbolTablesStorage::default(),
-                type_store: TypeStore::default(),
-            },
+            jars: JarsStorage::default(),
             files: Files::default(),
             workspace,
         }
@@ -44,18 +31,25 @@ impl Program {
 
     pub fn apply_changes<I>(&mut self, changes: I)
     where
-        I: IntoIterator<Item = FileChange>,
+        I: IntoIterator<Item = FileWatcherChange>,
     {
-        for change in changes {
-            self.semantic
-                .module_resolver
-                .remove_module(&self.file_path(change.id));
-            self.semantic.symbol_tables.remove(&change.id);
-            self.source.sources.remove(&change.id);
-            self.source.parsed.remove(&change.id);
-            self.source.lint_syntax.remove(&change.id);
+        let mut aggregated_changes = AggregatedChanges::default();
+
+        aggregated_changes.extend(changes.into_iter().map(|change| FileChange {
+            id: self.files.intern(&change.path),
+            kind: change.kind,
+        }));
+
+        let (source, semantic, lint) = self.jars_mut();
+        for change in aggregated_changes.iter() {
+            semantic.module_resolver.remove_module_by_file(change.id);
+            semantic.semantic_indices.remove(&change.id);
+            source.sources.remove(&change.id);
+            source.parsed.remove(&change.id);
             // TODO: remove all dependent modules as well
-            self.semantic.type_store.remove_module(change.id);
+            semantic.type_store.remove_module(change.id);
+            lint.lint_syntax.remove(&change.id);
+            lint.lint_semantic.remove(&change.id);
         }
     }
 
@@ -80,90 +74,124 @@ impl SourceDb for Program {
     fn file_path(&self, file_id: FileId) -> Arc<Path> {
         self.files.path(file_id)
     }
+}
 
-    fn source(&self, file_id: FileId) -> Source {
-        source_text(self, file_id)
-    }
+impl DbWithJar<SourceJar> for Program {}
 
-    fn parse(&self, file_id: FileId) -> Parsed {
-        parse(self, file_id)
-    }
+impl SemanticDb for Program {}
 
-    fn lint_syntax(&self, file_id: FileId) -> Diagnostics {
-        lint_syntax(self, file_id)
+impl DbWithJar<SemanticJar> for Program {}
+
+impl LintDb for Program {}
+
+impl DbWithJar<LintJar> for Program {}
+
+impl Upcast<dyn SemanticDb> for Program {
+    fn upcast(&self) -> &(dyn SemanticDb + 'static) {
+        self
     }
 }
 
-impl SemanticDb for Program {
-    fn resolve_module(&self, name: ModuleName) -> Option<Module> {
-        resolve_module(self, name)
+impl Upcast<dyn SourceDb> for Program {
+    fn upcast(&self) -> &(dyn SourceDb + 'static) {
+        self
     }
+}
 
-    fn file_to_module(&self, file_id: FileId) -> Option<Module> {
-        file_to_module(self, file_id)
-    }
-
-    fn path_to_module(&self, path: &Path) -> Option<Module> {
-        path_to_module(self, path)
-    }
-
-    fn symbol_table(&self, file_id: FileId) -> Arc<SymbolTable> {
-        symbol_table(self, file_id)
-    }
-
-    // Mutations
-
-    fn add_module(&mut self, path: &Path) -> Option<(Module, Vec<Arc<ModuleData>>)> {
-        add_module(self, path)
-    }
-
-    fn set_module_search_paths(&mut self, paths: Vec<ModuleSearchPath>) {
-        set_module_search_paths(self, paths);
-    }
-
-    fn infer_symbol_type(&mut self, file_id: FileId, symbol_id: SymbolId) -> Type {
-        infer_symbol_type(self, file_id, symbol_id)
+impl Upcast<dyn LintDb> for Program {
+    fn upcast(&self) -> &(dyn LintDb + 'static) {
+        self
     }
 }
 
 impl Db for Program {}
 
+impl Database for Program {
+    fn runtime(&self) -> &DbRuntime {
+        self.jars.runtime()
+    }
+
+    fn runtime_mut(&mut self) -> &mut DbRuntime {
+        self.jars.runtime_mut()
+    }
+}
+
+impl ParallelDatabase for Program {
+    fn snapshot(&self) -> Snapshot<Self> {
+        Snapshot::new(Self {
+            jars: self.jars.snapshot(),
+            files: self.files.snapshot(),
+            workspace: self.workspace.clone(),
+        })
+    }
+}
+
+impl HasJars for Program {
+    type Jars = (SourceJar, SemanticJar, LintJar);
+
+    fn jars(&self) -> QueryResult<&Self::Jars> {
+        self.jars.jars()
+    }
+
+    fn jars_mut(&mut self) -> &mut Self::Jars {
+        self.jars.jars_mut()
+    }
+}
+
 impl HasJar<SourceJar> for Program {
-    fn jar(&self) -> &SourceJar {
-        &self.source
+    fn jar(&self) -> QueryResult<&SourceJar> {
+        Ok(&self.jars()?.0)
     }
 
     fn jar_mut(&mut self) -> &mut SourceJar {
-        &mut self.source
+        &mut self.jars_mut().0
     }
 }
 
 impl HasJar<SemanticJar> for Program {
-    fn jar(&self) -> &SemanticJar {
-        &self.semantic
+    fn jar(&self) -> QueryResult<&SemanticJar> {
+        Ok(&self.jars()?.1)
     }
 
     fn jar_mut(&mut self) -> &mut SemanticJar {
-        &mut self.semantic
+        &mut self.jars_mut().1
+    }
+}
+
+impl HasJar<LintJar> for Program {
+    fn jar(&self) -> QueryResult<&LintJar> {
+        Ok(&self.jars()?.2)
+    }
+
+    fn jar_mut(&mut self) -> &mut LintJar {
+        &mut self.jars_mut().2
+    }
+}
+
+#[derive(Clone, Debug)]
+pub struct FileWatcherChange {
+    path: PathBuf,
+    kind: FileChangeKind,
+}
+
+impl FileWatcherChange {
+    pub fn new(path: PathBuf, kind: FileChangeKind) -> Self {
+        Self { path, kind }
     }
 }
 
 #[derive(Copy, Clone, Debug)]
-pub struct FileChange {
+struct FileChange {
     id: FileId,
     kind: FileChangeKind,
 }
 
 impl FileChange {
-    pub fn new(file_id: FileId, kind: FileChangeKind) -> Self {
-        Self { id: file_id, kind }
-    }
-
-    pub fn file_id(&self) -> FileId {
+    fn file_id(self) -> FileId {
         self.id
     }
 
-    pub fn kind(&self) -> FileChangeKind {
+    fn kind(self) -> FileChangeKind {
         self.kind
     }
 }
@@ -174,3 +202,74 @@ pub enum FileChangeKind {
     Modified,
     Deleted,
 }
+
+#[derive(Default, Debug)]
+struct AggregatedChanges {
+    changes: FxHashMap<FileId, FileChangeKind>,
+}
+
+impl AggregatedChanges {
+    fn add(&mut self, change: FileChange) {
+        match self.changes.entry(change.file_id()) {
+            Entry::Occupied(mut entry) => {
+                let merged = entry.get_mut();
+
+                match (merged, change.kind()) {
+                    (FileChangeKind::Created, FileChangeKind::Deleted) => {
+                        // Deletion after creations means that ruff never saw the file.
+                        entry.remove();
+                    }
+                    (FileChangeKind::Created, FileChangeKind::Modified) => {
+                        // No-op, for ruff, modifying a file that it doesn't yet know that it exists is still considered a creation.
+                    }
+
+                    (FileChangeKind::Modified, FileChangeKind::Created) => {
+                        // Uhh, that should probably not happen. Continue considering it a modification.
+                    }
+
+                    (FileChangeKind::Modified, FileChangeKind::Deleted) => {
+                        *entry.get_mut() = FileChangeKind::Deleted;
+                    }
+
+                    (FileChangeKind::Deleted, FileChangeKind::Created) => {
+                        *entry.get_mut() = FileChangeKind::Modified;
+                    }
+
+                    (FileChangeKind::Deleted, FileChangeKind::Modified) => {
+                        // That's weird, but let's consider it a modification.
+                        *entry.get_mut() = FileChangeKind::Modified;
+                    }
+
+                    (FileChangeKind::Created, FileChangeKind::Created)
+                    | (FileChangeKind::Modified, FileChangeKind::Modified)
+                    | (FileChangeKind::Deleted, FileChangeKind::Deleted) => {
+                        // No-op transitions. Some of them should be impossible but we handle them anyway.
+                    }
+                }
+            }
+            Entry::Vacant(entry) => {
+                entry.insert(change.kind());
+            }
+        }
+    }
+
+    fn extend<I>(&mut self, changes: I)
+    where
+        I: IntoIterator<Item = FileChange>,
+    {
+        let iter = changes.into_iter();
+        let (lower, _) = iter.size_hint();
+        self.changes.reserve(lower);
+
+        for change in iter {
+            self.add(change);
+        }
+    }
+
+    fn iter(&self) -> impl Iterator<Item = FileChange> + '_ {
+        self.changes.iter().map(|(id, kind)| FileChange {
+            id: *id,
+            kind: *kind,
+        })
+    }
+}

crates/red_knot/src/semantic.rs

@@ -0,0 +1,882 @@
+use std::num::NonZeroU32;
+
+use ruff_python_ast as ast;
+use ruff_python_ast::visitor::source_order::SourceOrderVisitor;
+use ruff_python_ast::AstNode;
+
+use crate::ast_ids::{NodeKey, TypedNodeKey};
+use crate::cache::KeyValueCache;
+use crate::db::{QueryResult, SemanticDb, SemanticJar};
+use crate::files::FileId;
+use crate::module::Module;
+use crate::module::ModuleName;
+use crate::parse::parse;
+use crate::Name;
+pub(crate) use definitions::Definition;
+use definitions::{ImportDefinition, ImportFromDefinition};
+pub(crate) use flow_graph::ConstrainedDefinition;
+use flow_graph::{FlowGraph, FlowGraphBuilder, FlowNodeId, ReachableDefinitionsIterator};
+use ruff_index::{newtype_index, IndexVec};
+use rustc_hash::FxHashMap;
+use std::ops::{Deref, DerefMut};
+use std::sync::Arc;
+pub(crate) use symbol_table::{Dependency, SymbolId};
+use symbol_table::{ScopeId, ScopeKind, SymbolFlags, SymbolTable, SymbolTableBuilder};
+pub(crate) use types::{infer_definition_type, infer_symbol_public_type, Type, TypeStore};
+
+mod definitions;
+mod flow_graph;
+mod symbol_table;
+mod types;
+
+#[tracing::instrument(level = "debug", skip(db))]
+pub fn semantic_index(db: &dyn SemanticDb, file_id: FileId) -> QueryResult<Arc<SemanticIndex>> {
+    let jar: &SemanticJar = db.jar()?;
+
+    jar.semantic_indices.get(&file_id, |_| {
+        let parsed = parse(db.upcast(), file_id)?;
+        Ok(Arc::from(SemanticIndex::from_ast(parsed.syntax())))
+    })
+}
+
+#[tracing::instrument(level = "debug", skip(db))]
+pub fn resolve_global_symbol(
+    db: &dyn SemanticDb,
+    module: Module,
+    name: &str,
+) -> QueryResult<Option<GlobalSymbolId>> {
+    let file_id = module.path(db)?.file();
+    let symbol_table = &semantic_index(db, file_id)?.symbol_table;
+    let Some(symbol_id) = symbol_table.root_symbol_id_by_name(name) else {
+        return Ok(None);
+    };
+    Ok(Some(GlobalSymbolId { file_id, symbol_id }))
+}
+
+#[newtype_index]
+pub struct ExpressionId;
+
+#[derive(Copy, Clone, Debug, PartialEq, Eq)]
+pub struct GlobalSymbolId {
+    pub(crate) file_id: FileId,
+    pub(crate) symbol_id: SymbolId,
+}
+
+#[derive(Debug)]
+pub struct SemanticIndex {
+    symbol_table: SymbolTable,
+    flow_graph: FlowGraph,
+    expressions: FxHashMap<NodeKey, ExpressionId>,
+    expressions_by_id: IndexVec<ExpressionId, NodeKey>,
+}
+
+impl SemanticIndex {
+    pub fn from_ast(module: &ast::ModModule) -> Self {
+        let root_scope_id = SymbolTable::root_scope_id();
+        let mut indexer = SemanticIndexer {
+            symbol_table_builder: SymbolTableBuilder::new(),
+            flow_graph_builder: FlowGraphBuilder::new(),
+            scopes: vec![ScopeState {
+                scope_id: root_scope_id,
+                current_flow_node_id: FlowGraph::start(),
+            }],
+            expressions: FxHashMap::default(),
+            expressions_by_id: IndexVec::default(),
+            current_definition: None,
+        };
+        indexer.visit_body(&module.body);
+        indexer.finish()
+    }
+
+    fn resolve_expression_id<'a>(
+        &self,
+        ast: &'a ast::ModModule,
+        expression_id: ExpressionId,
+    ) -> ast::AnyNodeRef<'a> {
+        let node_key = self.expressions_by_id[expression_id];
+        node_key
+            .resolve(ast.as_any_node_ref())
+            .expect("node to resolve")
+    }
+
+    /// Return an iterator over all definitions of `symbol_id` reachable from `use_expr`. The value
+    /// of `symbol_id` in `use_expr` must originate from one of the iterated definitions (or from
+    /// an external reassignment of the name outside of this scope).
+    pub fn reachable_definitions(
+        &self,
+        symbol_id: SymbolId,
+        use_expr: &ast::Expr,
+    ) -> ReachableDefinitionsIterator {
+        let expression_id = self.expression_id(use_expr);
+        ReachableDefinitionsIterator::new(
+            &self.flow_graph,
+            symbol_id,
+            self.flow_graph.for_expr(expression_id),
+        )
+    }
+
+    pub fn expression_id(&self, expression: &ast::Expr) -> ExpressionId {
+        self.expressions[&NodeKey::from_node(expression.into())]
+    }
+
+    pub fn symbol_table(&self) -> &SymbolTable {
+        &self.symbol_table
+    }
+}
+
+#[derive(Debug)]
+struct ScopeState {
+    scope_id: ScopeId,
+    current_flow_node_id: FlowNodeId,
+}
+
+#[derive(Debug)]
+struct SemanticIndexer {
+    symbol_table_builder: SymbolTableBuilder,
+    flow_graph_builder: FlowGraphBuilder,
+    scopes: Vec<ScopeState>,
+    /// the definition whose target(s) we are currently walking
+    current_definition: Option<Definition>,
+    expressions: FxHashMap<NodeKey, ExpressionId>,
+    expressions_by_id: IndexVec<ExpressionId, NodeKey>,
+}
+
+impl SemanticIndexer {
+    pub(crate) fn finish(mut self) -> SemanticIndex {
+        let SemanticIndexer {
+            flow_graph_builder,
+            symbol_table_builder,
+            ..
+        } = self;
+        self.expressions.shrink_to_fit();
+        self.expressions_by_id.shrink_to_fit();
+        SemanticIndex {
+            flow_graph: flow_graph_builder.finish(),
+            symbol_table: symbol_table_builder.finish(),
+            expressions: self.expressions,
+            expressions_by_id: self.expressions_by_id,
+        }
+    }
+
+    fn set_current_flow_node(&mut self, new_flow_node_id: FlowNodeId) {
+        let scope_state = self.scopes.last_mut().expect("scope stack is never empty");
+        scope_state.current_flow_node_id = new_flow_node_id;
+    }
+
+    fn current_flow_node(&self) -> FlowNodeId {
+        self.scopes
+            .last()
+            .expect("scope stack is never empty")
+            .current_flow_node_id
+    }
+
+    fn add_or_update_symbol(&mut self, identifier: &str, flags: SymbolFlags) -> SymbolId {
+        self.symbol_table_builder
+            .add_or_update_symbol(self.cur_scope(), identifier, flags)
+    }
+
+    fn add_or_update_symbol_with_def(
+        &mut self,
+        identifier: &str,
+        definition: Definition,
+    ) -> SymbolId {
+        let symbol_id = self.add_or_update_symbol(identifier, SymbolFlags::IS_DEFINED);
+        self.symbol_table_builder
+            .add_definition(symbol_id, definition.clone());
+        let new_flow_node_id =
+            self.flow_graph_builder
+                .add_definition(symbol_id, definition, self.current_flow_node());
+        self.set_current_flow_node(new_flow_node_id);
+        symbol_id
+    }
+
+    fn push_scope(
+        &mut self,
+        name: &str,
+        kind: ScopeKind,
+        definition: Option<Definition>,
+        defining_symbol: Option<SymbolId>,
+    ) -> ScopeId {
+        let scope_id = self.symbol_table_builder.add_child_scope(
+            self.cur_scope(),
+            name,
+            kind,
+            definition,
+            defining_symbol,
+        );
+        self.scopes.push(ScopeState {
+            scope_id,
+            current_flow_node_id: FlowGraph::start(),
+        });
+        scope_id
+    }
+
+    fn pop_scope(&mut self) -> ScopeId {
+        self.scopes
+            .pop()
+            .expect("Scope stack should never be empty")
+            .scope_id
+    }
+
+    fn cur_scope(&self) -> ScopeId {
+        self.scopes
+            .last()
+            .expect("Scope stack should never be empty")
+            .scope_id
+    }
+
+    fn record_scope_for_node(&mut self, node_key: NodeKey, scope_id: ScopeId) {
+        self.symbol_table_builder
+            .record_scope_for_node(node_key, scope_id);
+    }
+
+    fn insert_constraint(&mut self, expr: &ast::Expr) {
+        let node_key = NodeKey::from_node(expr.into());
+        let expression_id = self.expressions[&node_key];
+        let constraint = self
+            .flow_graph_builder
+            .add_constraint(self.current_flow_node(), expression_id);
+        self.set_current_flow_node(constraint);
+    }
+
+    fn with_type_params(
+        &mut self,
+        name: &str,
+        params: &Option<Box<ast::TypeParams>>,
+        definition: Option<Definition>,
+        defining_symbol: Option<SymbolId>,
+        nested: impl FnOnce(&mut Self) -> ScopeId,
+    ) -> ScopeId {
+        if let Some(type_params) = params {
+            self.push_scope(name, ScopeKind::Annotation, definition, defining_symbol);
+            for type_param in &type_params.type_params {
+                let name = match type_param {
+                    ast::TypeParam::TypeVar(ast::TypeParamTypeVar { name, .. }) => name,
+                    ast::TypeParam::ParamSpec(ast::TypeParamParamSpec { name, .. }) => name,
+                    ast::TypeParam::TypeVarTuple(ast::TypeParamTypeVarTuple { name, .. }) => name,
+                };
+                self.add_or_update_symbol(name, SymbolFlags::IS_DEFINED);
+            }
+        }
+        let scope_id = nested(self);
+        if params.is_some() {
+            self.pop_scope();
+        }
+        scope_id
+    }
+}
+
+impl SourceOrderVisitor<'_> for SemanticIndexer {
+    fn visit_expr(&mut self, expr: &ast::Expr) {
+        let node_key = NodeKey::from_node(expr.into());
+        let expression_id = self.expressions_by_id.push(node_key);
+
+        let flow_expression_id = self
+            .flow_graph_builder
+            .record_expr(self.current_flow_node());
+        debug_assert_eq!(expression_id, flow_expression_id);
+
+        let symbol_expression_id = self
+            .symbol_table_builder
+            .record_expression(self.cur_scope());
+
+        debug_assert_eq!(expression_id, symbol_expression_id);
+
+        self.expressions.insert(node_key, expression_id);
+
+        match expr {
+            ast::Expr::Name(ast::ExprName { id, ctx, .. }) => {
+                let flags = match ctx {
+                    ast::ExprContext::Load => SymbolFlags::IS_USED,
+                    ast::ExprContext::Store => SymbolFlags::IS_DEFINED,
+                    ast::ExprContext::Del => SymbolFlags::IS_DEFINED,
+                    ast::ExprContext::Invalid => SymbolFlags::empty(),
+                };
+                self.add_or_update_symbol(id, flags);
+                if flags.contains(SymbolFlags::IS_DEFINED) {
+                    if let Some(curdef) = self.current_definition.clone() {
+                        self.add_or_update_symbol_with_def(id, curdef);
+                    }
+                }
+                ast::visitor::source_order::walk_expr(self, expr);
+            }
+            ast::Expr::Named(node) => {
+                debug_assert!(self.current_definition.is_none());
+                self.current_definition =
+                    Some(Definition::NamedExpr(TypedNodeKey::from_node(node)));
+                // TODO walrus in comprehensions is implicitly nonlocal
+                self.visit_expr(&node.target);
+                self.current_definition = None;
+                self.visit_expr(&node.value);
+            }
+            ast::Expr::If(ast::ExprIf {
+                body, test, orelse, ..
+            }) => {
+                // TODO detect statically known truthy or falsy test (via type inference, not naive
+                // AST inspection, so we can't simplify here, need to record test expression in CFG
+                // for later checking)
+
+                self.visit_expr(test);
+
+                let if_branch = self.flow_graph_builder.add_branch(self.current_flow_node());
+
+                self.set_current_flow_node(if_branch);
+                self.insert_constraint(test);
+                self.visit_expr(body);
+
+                let post_body = self.current_flow_node();
+
+                self.set_current_flow_node(if_branch);
+                self.visit_expr(orelse);
+
+                let post_else = self
+                    .flow_graph_builder
+                    .add_phi(self.current_flow_node(), post_body);
+
+                self.set_current_flow_node(post_else);
+            }
+            _ => {
+                ast::visitor::source_order::walk_expr(self, expr);
+            }
+        }
+    }
+
+    fn visit_stmt(&mut self, stmt: &ast::Stmt) {
+        // TODO need to capture more definition statements here
+        match stmt {
+            ast::Stmt::ClassDef(node) => {
+                let node_key = TypedNodeKey::from_node(node);
+                let def = Definition::ClassDef(node_key.clone());
+                let symbol_id = self.add_or_update_symbol_with_def(&node.name, def.clone());
+                for decorator in &node.decorator_list {
+                    self.visit_decorator(decorator);
+                }
+                let scope_id = self.with_type_params(
+                    &node.name,
+                    &node.type_params,
+                    Some(def.clone()),
+                    Some(symbol_id),
+                    |indexer| {
+                        if let Some(arguments) = &node.arguments {
+                            indexer.visit_arguments(arguments);
+                        }
+                        let scope_id = indexer.push_scope(
+                            &node.name,
+                            ScopeKind::Class,
+                            Some(def.clone()),
+                            Some(symbol_id),
+                        );
+                        indexer.visit_body(&node.body);
+                        indexer.pop_scope();
+                        scope_id
+                    },
+                );
+                self.record_scope_for_node(*node_key.erased(), scope_id);
+            }
+            ast::Stmt::FunctionDef(node) => {
+                let node_key = TypedNodeKey::from_node(node);
+                let def = Definition::FunctionDef(node_key.clone());
+                let symbol_id = self.add_or_update_symbol_with_def(&node.name, def.clone());
+                for decorator in &node.decorator_list {
+                    self.visit_decorator(decorator);
+                }
+                let scope_id = self.with_type_params(
+                    &node.name,
+                    &node.type_params,
+                    Some(def.clone()),
+                    Some(symbol_id),
+                    |indexer| {
+                        indexer.visit_parameters(&node.parameters);
+                        for expr in &node.returns {
+                            indexer.visit_annotation(expr);
+                        }
+                        let scope_id = indexer.push_scope(
+                            &node.name,
+                            ScopeKind::Function,
+                            Some(def.clone()),
+                            Some(symbol_id),
+                        );
+                        indexer.visit_body(&node.body);
+                        indexer.pop_scope();
+                        scope_id
+                    },
+                );
+                self.record_scope_for_node(*node_key.erased(), scope_id);
+            }
+            ast::Stmt::Import(ast::StmtImport { names, .. }) => {
+                for alias in names {
+                    let symbol_name = if let Some(asname) = &alias.asname {
+                        asname.id.as_str()
+                    } else {
+                        alias.name.id.split('.').next().unwrap()
+                    };
+
+                    let module = ModuleName::new(&alias.name.id);
+
+                    let def = Definition::Import(ImportDefinition {
+                        module: module.clone(),
+                    });
+                    self.add_or_update_symbol_with_def(symbol_name, def);
+                    self.symbol_table_builder
+                        .add_dependency(Dependency::Module(module));
+                }
+            }
+            ast::Stmt::ImportFrom(ast::StmtImportFrom {
+                module,
+                names,
+                level,
+                ..
+            }) => {
+                let module = module.as_ref().map(|m| ModuleName::new(&m.id));
+
+                for alias in names {
+                    let symbol_name = if let Some(asname) = &alias.asname {
+                        asname.id.as_str()
+                    } else {
+                        alias.name.id.as_str()
+                    };
+                    let def = Definition::ImportFrom(ImportFromDefinition {
+                        module: module.clone(),
+                        name: Name::new(&alias.name.id),
+                        level: *level,
+                    });
+                    self.add_or_update_symbol_with_def(symbol_name, def);
+                }
+
+                let dependency = if let Some(module) = module {
+                    match NonZeroU32::new(*level) {
+                        Some(level) => Dependency::Relative {
+                            level,
+                            module: Some(module),
+                        },
+                        None => Dependency::Module(module),
+                    }
+                } else {
+                    Dependency::Relative {
+                        level: NonZeroU32::new(*level)
+                            .expect("Import without a module to have a level > 0"),
+                        module,
+                    }
+                };
+
+                self.symbol_table_builder.add_dependency(dependency);
+            }
+            ast::Stmt::Assign(node) => {
+                debug_assert!(self.current_definition.is_none());
+                self.visit_expr(&node.value);
+                self.current_definition =
+                    Some(Definition::Assignment(TypedNodeKey::from_node(node)));
+                for expr in &node.targets {
+                    self.visit_expr(expr);
+                }
+
+                self.current_definition = None;
+            }
+            ast::Stmt::If(node) => {
+                // TODO detect statically known truthy or falsy test (via type inference, not naive
+                // AST inspection, so we can't simplify here, need to record test expression in CFG
+                // for later checking)
+
+                // we visit the if "test" condition first regardless
+                self.visit_expr(&node.test);
+
+                // create branch node: does the if test pass or not?
+                let if_branch = self.flow_graph_builder.add_branch(self.current_flow_node());
+
+                // visit the body of the `if` clause
+                self.set_current_flow_node(if_branch);
+                self.insert_constraint(&node.test);
+                self.visit_body(&node.body);
+
+                // Flow node for the last if/elif condition branch; represents the "no branch
+                // taken yet" possibility (where "taking a branch" means that the condition in an
+                // if or elif evaluated to true and control flow went into that clause).
+                let mut prior_branch = if_branch;
+
+                // Flow node for the state after the prior if/elif/else clause; represents "we have
+                // taken one of the branches up to this point." Initially set to the post-if-clause
+                // state, later will be set to the phi node joining that possible path with the
+                // possibility that we took a later if/elif/else clause instead.
+                let mut post_prior_clause = self.current_flow_node();
+
+                // Flag to mark if the final clause is an "else" -- if so, that means the "match no
+                // clauses" path is not possible, we have to go through one of the clauses.
+                let mut last_branch_is_else = false;
+
+                for clause in &node.elif_else_clauses {
+                    if let Some(test) = &clause.test {
+                        self.visit_expr(test);
+                        // This is an elif clause. Create a new branch node. Its predecessor is the
+                        // previous branch node, because we can only take one branch in an entire
+                        // if/elif/else chain, so if we take this branch, it can only be because we
+                        // didn't take the previous one.
+                        prior_branch = self.flow_graph_builder.add_branch(prior_branch);
+                        self.set_current_flow_node(prior_branch);
+                        self.insert_constraint(test);
+                    } else {
+                        // This is an else clause. No need to create a branch node; there's no
+                        // branch here, if we haven't taken any previous branch, we definitely go
+                        // into the "else" clause.
+                        self.set_current_flow_node(prior_branch);
+                        last_branch_is_else = true;
+                    }
+                    self.visit_elif_else_clause(clause);
+                    // Update `post_prior_clause` to a new phi node joining the possibility that we
+                    // took any of the previous branches with the possibility that we took the one
+                    // just visited.
+                    post_prior_clause = self
+                        .flow_graph_builder
+                        .add_phi(self.current_flow_node(), post_prior_clause);
+                }
+
+                if !last_branch_is_else {
+                    // Final branch was not an "else", which means it's possible we took zero
+                    // branches in the entire if/elif chain, so we need one more phi node to join
+                    // the "no branches taken" possibility.
+                    post_prior_clause = self
+                        .flow_graph_builder
+                        .add_phi(post_prior_clause, prior_branch);
+                }
+
+                // Onward, with current flow node set to our final Phi node.
+                self.set_current_flow_node(post_prior_clause);
+            }
+            _ => {
+                ast::visitor::source_order::walk_stmt(self, stmt);
+            }
+        }
+    }
+}
+
+#[derive(Debug, Default)]
+pub struct SemanticIndexStorage(KeyValueCache<FileId, Arc<SemanticIndex>>);
+
+impl Deref for SemanticIndexStorage {
+    type Target = KeyValueCache<FileId, Arc<SemanticIndex>>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+impl DerefMut for SemanticIndexStorage {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.0
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use crate::semantic::symbol_table::{Symbol, SymbolIterator};
+    use ruff_python_ast as ast;
+    use ruff_python_ast::ModModule;
+    use ruff_python_parser::{Mode, Parsed};
+
+    use super::{Definition, ScopeKind, SemanticIndex, SymbolId};
+
+    fn parse(code: &str) -> Parsed<ModModule> {
+        ruff_python_parser::parse_unchecked(code, Mode::Module)
+            .try_into_module()
+            .unwrap()
+    }
+
+    fn names<I>(it: SymbolIterator<I>) -> Vec<&str>
+    where
+        I: Iterator<Item = SymbolId>,
+    {
+        let mut symbols: Vec<_> = it.map(Symbol::name).collect();
+        symbols.sort_unstable();
+        symbols
+    }
+
+    #[test]
+    fn empty() {
+        let parsed = parse("");
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()).len(), 0);
+    }
+
+    #[test]
+    fn simple() {
+        let parsed = parse("x");
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["x"]);
+        assert_eq!(
+            table
+                .definitions(table.root_symbol_id_by_name("x").unwrap())
+                .len(),
+            0
+        );
+    }
+
+    #[test]
+    fn annotation_only() {
+        let parsed = parse("x: int");
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["int", "x"]);
+        // TODO record definition
+    }
+
+    #[test]
+    fn import() {
+        let parsed = parse("import foo");
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["foo"]);
+        assert_eq!(
+            table
+                .definitions(table.root_symbol_id_by_name("foo").unwrap())
+                .len(),
+            1
+        );
+    }
+
+    #[test]
+    fn import_sub() {
+        let parsed = parse("import foo.bar");
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["foo"]);
+    }
+
+    #[test]
+    fn import_as() {
+        let parsed = parse("import foo.bar as baz");
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["baz"]);
+    }
+
+    #[test]
+    fn import_from() {
+        let parsed = parse("from bar import foo");
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["foo"]);
+        assert_eq!(
+            table
+                .definitions(table.root_symbol_id_by_name("foo").unwrap())
+                .len(),
+            1
+        );
+        assert!(
+            table.root_symbol_id_by_name("foo").is_some_and(|sid| {
+                let s = sid.symbol(&table);
+                s.is_defined() || !s.is_used()
+            }),
+            "symbols that are defined get the defined flag"
+        );
+    }
+
+    #[test]
+    fn assign() {
+        let parsed = parse("x = foo");
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["foo", "x"]);
+        assert_eq!(
+            table
+                .definitions(table.root_symbol_id_by_name("x").unwrap())
+                .len(),
+            1
+        );
+        assert!(
+            table.root_symbol_id_by_name("foo").is_some_and(|sid| {
+                let s = sid.symbol(&table);
+                !s.is_defined() && s.is_used()
+            }),
+            "a symbol used but not defined in a scope should have only the used flag"
+        );
+    }
+
+    #[test]
+    fn class_scope() {
+        let parsed = parse(
+            "
+                class C:
+                    x = 1
+                y = 2
+                ",
+        );
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["C", "y"]);
+        let scopes = table.root_child_scope_ids();
+        assert_eq!(scopes.len(), 1);
+        let c_scope = scopes[0].scope(&table);
+        assert_eq!(c_scope.kind(), ScopeKind::Class);
+        assert_eq!(c_scope.name(), "C");
+        assert_eq!(names(table.symbols_for_scope(scopes[0])), vec!["x"]);
+        assert_eq!(
+            table
+                .definitions(table.root_symbol_id_by_name("C").unwrap())
+                .len(),
+            1
+        );
+    }
+
+    #[test]
+    fn func_scope() {
+        let parsed = parse(
+            "
+                def func():
+                    x = 1
+                y = 2
+                ",
+        );
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["func", "y"]);
+        let scopes = table.root_child_scope_ids();
+        assert_eq!(scopes.len(), 1);
+        let func_scope = scopes[0].scope(&table);
+        assert_eq!(func_scope.kind(), ScopeKind::Function);
+        assert_eq!(func_scope.name(), "func");
+        assert_eq!(names(table.symbols_for_scope(scopes[0])), vec!["x"]);
+        assert_eq!(
+            table
+                .definitions(table.root_symbol_id_by_name("func").unwrap())
+                .len(),
+            1
+        );
+    }
+
+    #[test]
+    fn dupes() {
+        let parsed = parse(
+            "
+                def func():
+                    x = 1
+                def func():
+                    y = 2
+                ",
+        );
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["func"]);
+        let scopes = table.root_child_scope_ids();
+        assert_eq!(scopes.len(), 2);
+        let func_scope_1 = scopes[0].scope(&table);
+        let func_scope_2 = scopes[1].scope(&table);
+        assert_eq!(func_scope_1.kind(), ScopeKind::Function);
+        assert_eq!(func_scope_1.name(), "func");
+        assert_eq!(func_scope_2.kind(), ScopeKind::Function);
+        assert_eq!(func_scope_2.name(), "func");
+        assert_eq!(names(table.symbols_for_scope(scopes[0])), vec!["x"]);
+        assert_eq!(names(table.symbols_for_scope(scopes[1])), vec!["y"]);
+        assert_eq!(
+            table
+                .definitions(table.root_symbol_id_by_name("func").unwrap())
+                .len(),
+            2
+        );
+    }
+
+    #[test]
+    fn generic_func() {
+        let parsed = parse(
+            "
+                def func[T]():
+                    x = 1
+                ",
+        );
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["func"]);
+        let scopes = table.root_child_scope_ids();
+        assert_eq!(scopes.len(), 1);
+        let ann_scope_id = scopes[0];
+        let ann_scope = ann_scope_id.scope(&table);
+        assert_eq!(ann_scope.kind(), ScopeKind::Annotation);
+        assert_eq!(ann_scope.name(), "func");
+        assert_eq!(names(table.symbols_for_scope(ann_scope_id)), vec!["T"]);
+        let scopes = table.child_scope_ids_of(ann_scope_id);
+        assert_eq!(scopes.len(), 1);
+        let func_scope_id = scopes[0];
+        let func_scope = func_scope_id.scope(&table);
+        assert_eq!(func_scope.kind(), ScopeKind::Function);
+        assert_eq!(func_scope.name(), "func");
+        assert_eq!(names(table.symbols_for_scope(func_scope_id)), vec!["x"]);
+    }
+
+    #[test]
+    fn generic_class() {
+        let parsed = parse(
+            "
+                class C[T]:
+                    x = 1
+                ",
+        );
+        let table = SemanticIndex::from_ast(parsed.syntax()).symbol_table;
+        assert_eq!(names(table.root_symbols()), vec!["C"]);
+        let scopes = table.root_child_scope_ids();
+        assert_eq!(scopes.len(), 1);
+        let ann_scope_id = scopes[0];
+        let ann_scope = ann_scope_id.scope(&table);
+        assert_eq!(ann_scope.kind(), ScopeKind::Annotation);
+        assert_eq!(ann_scope.name(), "C");
+        assert_eq!(names(table.symbols_for_scope(ann_scope_id)), vec!["T"]);
+        assert!(
+            table
+                .symbol_by_name(ann_scope_id, "T")
+                .is_some_and(|s| s.is_defined() && !s.is_used()),
+            "type parameters are defined by the scope that introduces them"
+        );
+        let scopes = table.child_scope_ids_of(ann_scope_id);
+        assert_eq!(scopes.len(), 1);
+        let func_scope_id = scopes[0];
+        let func_scope = func_scope_id.scope(&table);
+        assert_eq!(func_scope.kind(), ScopeKind::Class);
+        assert_eq!(func_scope.name(), "C");
+        assert_eq!(names(table.symbols_for_scope(func_scope_id)), vec!["x"]);
+    }
+
+    #[test]
+    fn reachability_trivial() {
+        let parsed = parse("x = 1; x");
+        let ast = parsed.syntax();
+        let index = SemanticIndex::from_ast(ast);
+        let table = &index.symbol_table;
+        let x_sym = table
+            .root_symbol_id_by_name("x")
+            .expect("x symbol should exist");
+        let ast::Stmt::Expr(ast::StmtExpr { value: x_use, .. }) = &ast.body[1] else {
+            panic!("should be an expr")
+        };
+        let x_defs: Vec<_> = index
+            .reachable_definitions(x_sym, x_use)
+            .map(|constrained_definition| constrained_definition.definition)
+            .collect();
+        assert_eq!(x_defs.len(), 1);
+        let Definition::Assignment(node_key) = &x_defs[0] else {
+            panic!("def should be an assignment")
+        };
+        let Some(def_node) = node_key.resolve(ast.into()) else {
+            panic!("node key should resolve")
+        };
+        let ast::Expr::NumberLiteral(ast::ExprNumberLiteral {
+            value: ast::Number::Int(num),
+            ..
+        }) = &*def_node.value
+        else {
+            panic!("should be a number literal")
+        };
+        assert_eq!(*num, 1);
+    }
+
+    #[test]
+    fn expression_scope() {
+        let parsed = parse("x = 1;\ndef test():\n  y = 4");
+        let ast = parsed.syntax();
+        let index = SemanticIndex::from_ast(ast);
+        let table = &index.symbol_table;
+
+        let x_sym = table
+            .root_symbol_by_name("x")
+            .expect("x symbol should exist");
+
+        let x_stmt = ast.body[0].as_assign_stmt().unwrap();
+
+        let x_id = index.expression_id(&x_stmt.targets[0]);
+
+        assert_eq!(table.scope_of_expression(x_id).kind(), ScopeKind::Module);
+        assert_eq!(table.scope_id_of_expression(x_id), x_sym.scope_id());
+
+        let def = ast.body[1].as_function_def_stmt().unwrap();
+        let y_stmt = def.body[0].as_assign_stmt().unwrap();
+        let y_id = index.expression_id(&y_stmt.targets[0]);
+
+        assert_eq!(table.scope_of_expression(y_id).kind(), ScopeKind::Function);
+    }
+}

crates/red_knot/src/semantic/definitions.rs

@@ -0,0 +1,52 @@
+use crate::ast_ids::TypedNodeKey;
+use crate::semantic::ModuleName;
+use crate::Name;
+use ruff_python_ast as ast;
+
+// TODO storing TypedNodeKey for definitions means we have to search to find them again in the AST;
+// this is at best O(log n). If looking up definitions is a bottleneck we should look for
+// alternatives here.
+// TODO intern Definitions in SymbolTable and reference using IDs?
+#[derive(Clone, Debug)]
+pub enum Definition {
+    // For the import cases, we don't need reference to any arbitrary AST subtrees (annotations,
+    // RHS), and referencing just the import statement node is imprecise (a single import statement
+    // can assign many symbols, we'd have to re-search for the one we care about), so we just copy
+    // the small amount of information we need from the AST.
+    Import(ImportDefinition),
+    ImportFrom(ImportFromDefinition),
+    ClassDef(TypedNodeKey<ast::StmtClassDef>),
+    FunctionDef(TypedNodeKey<ast::StmtFunctionDef>),
+    Assignment(TypedNodeKey<ast::StmtAssign>),
+    AnnotatedAssignment(TypedNodeKey<ast::StmtAnnAssign>),
+    NamedExpr(TypedNodeKey<ast::ExprNamed>),
+    /// represents the implicit initial definition of every name as "unbound"
+    Unbound,
+    // TODO with statements, except handlers, function args...
+}
+
+#[derive(Clone, Debug)]
+pub struct ImportDefinition {
+    pub module: ModuleName,
+}
+
+#[derive(Clone, Debug)]
+pub struct ImportFromDefinition {
+    pub module: Option<ModuleName>,
+    pub name: Name,
+    pub level: u32,
+}
+
+impl ImportFromDefinition {
+    pub fn module(&self) -> Option<&ModuleName> {
+        self.module.as_ref()
+    }
+
+    pub fn name(&self) -> &Name {
+        &self.name
+    }
+
+    pub fn level(&self) -> u32 {
+        self.level
+    }
+}

crates/red_knot/src/semantic/flow_graph.rs

@@ -0,0 +1,270 @@
+use super::symbol_table::SymbolId;
+use crate::semantic::{Definition, ExpressionId};
+use ruff_index::{newtype_index, IndexVec};
+use std::iter::FusedIterator;
+use std::ops::Range;
+
+#[newtype_index]
+pub struct FlowNodeId;
+
+#[derive(Debug)]
+pub(crate) enum FlowNode {
+    Start,
+    Definition(DefinitionFlowNode),
+    Branch(BranchFlowNode),
+    Phi(PhiFlowNode),
+    Constraint(ConstraintFlowNode),
+}
+
+/// A point in control flow where a symbol is defined
+#[derive(Debug)]
+pub(crate) struct DefinitionFlowNode {
+    symbol_id: SymbolId,
+    definition: Definition,
+    predecessor: FlowNodeId,
+}
+
+/// A branch in control flow
+#[derive(Debug)]
+pub(crate) struct BranchFlowNode {
+    predecessor: FlowNodeId,
+}
+
+/// A join point where control flow paths come together
+#[derive(Debug)]
+pub(crate) struct PhiFlowNode {
+    first_predecessor: FlowNodeId,
+    second_predecessor: FlowNodeId,
+}
+
+/// A branch test which may apply constraints to a symbol's type
+#[derive(Debug)]
+pub(crate) struct ConstraintFlowNode {
+    predecessor: FlowNodeId,
+    test_expression: ExpressionId,
+}
+
+#[derive(Debug)]
+pub struct FlowGraph {
+    flow_nodes_by_id: IndexVec<FlowNodeId, FlowNode>,
+    expression_map: IndexVec<ExpressionId, FlowNodeId>,
+}
+
+impl FlowGraph {
+    pub fn start() -> FlowNodeId {
+        FlowNodeId::from_usize(0)
+    }
+
+    pub fn for_expr(&self, expr: ExpressionId) -> FlowNodeId {
+        self.expression_map[expr]
+    }
+}
+
+#[derive(Debug)]
+pub(crate) struct FlowGraphBuilder {
+    flow_graph: FlowGraph,
+}
+
+impl FlowGraphBuilder {
+    pub(crate) fn new() -> Self {
+        let mut graph = FlowGraph {
+            flow_nodes_by_id: IndexVec::default(),
+            expression_map: IndexVec::default(),
+        };
+        graph.flow_nodes_by_id.push(FlowNode::Start);
+        Self { flow_graph: graph }
+    }
+
+    pub(crate) fn add(&mut self, node: FlowNode) -> FlowNodeId {
+        self.flow_graph.flow_nodes_by_id.push(node)
+    }
+
+    pub(crate) fn add_definition(
+        &mut self,
+        symbol_id: SymbolId,
+        definition: Definition,
+        predecessor: FlowNodeId,
+    ) -> FlowNodeId {
+        self.add(FlowNode::Definition(DefinitionFlowNode {
+            symbol_id,
+            definition,
+            predecessor,
+        }))
+    }
+
+    pub(crate) fn add_branch(&mut self, predecessor: FlowNodeId) -> FlowNodeId {
+        self.add(FlowNode::Branch(BranchFlowNode { predecessor }))
+    }
+
+    pub(crate) fn add_phi(
+        &mut self,
+        first_predecessor: FlowNodeId,
+        second_predecessor: FlowNodeId,
+    ) -> FlowNodeId {
+        self.add(FlowNode::Phi(PhiFlowNode {
+            first_predecessor,
+            second_predecessor,
+        }))
+    }
+
+    pub(crate) fn add_constraint(
+        &mut self,
+        predecessor: FlowNodeId,
+        test_expression: ExpressionId,
+    ) -> FlowNodeId {
+        self.add(FlowNode::Constraint(ConstraintFlowNode {
+            predecessor,
+            test_expression,
+        }))
+    }
+
+    pub(super) fn record_expr(&mut self, node_id: FlowNodeId) -> ExpressionId {
+        self.flow_graph.expression_map.push(node_id)
+    }
+
+    pub(super) fn finish(mut self) -> FlowGraph {
+        self.flow_graph.flow_nodes_by_id.shrink_to_fit();
+        self.flow_graph.expression_map.shrink_to_fit();
+        self.flow_graph
+    }
+}
+
+/// A definition, and the set of constraints between a use and the definition
+#[derive(Debug, Clone)]
+pub struct ConstrainedDefinition {
+    pub definition: Definition,
+    pub constraints: Vec<ExpressionId>,
+}
+
+/// A flow node and the constraints we passed through to reach it
+#[derive(Debug)]
+struct FlowState {
+    node_id: FlowNodeId,
+    constraints_range: Range<usize>,
+}
+
+#[derive(Debug)]
+pub struct ReachableDefinitionsIterator<'a> {
+    flow_graph: &'a FlowGraph,
+    symbol_id: SymbolId,
+    pending: Vec<FlowState>,
+    constraints: Vec<ExpressionId>,
+}
+
+impl<'a> ReachableDefinitionsIterator<'a> {
+    pub fn new(flow_graph: &'a FlowGraph, symbol_id: SymbolId, start_node_id: FlowNodeId) -> Self {
+        Self {
+            flow_graph,
+            symbol_id,
+            pending: vec![FlowState {
+                node_id: start_node_id,
+                constraints_range: 0..0,
+            }],
+            constraints: vec![],
+        }
+    }
+}
+
+impl<'a> Iterator for ReachableDefinitionsIterator<'a> {
+    type Item = ConstrainedDefinition;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        let FlowState {
+            mut node_id,
+            mut constraints_range,
+        } = self.pending.pop()?;
+        self.constraints.truncate(constraints_range.end + 1);
+        loop {
+            match &self.flow_graph.flow_nodes_by_id[node_id] {
+                FlowNode::Start => {
+                    // constraints on unbound are irrelevant
+                    return Some(ConstrainedDefinition {
+                        definition: Definition::Unbound,
+                        constraints: vec![],
+                    });
+                }
+                FlowNode::Definition(def_node) => {
+                    if def_node.symbol_id == self.symbol_id {
+                        return Some(ConstrainedDefinition {
+                            definition: def_node.definition.clone(),
+                            constraints: self.constraints[constraints_range].to_vec(),
+                        });
+                    }
+                    node_id = def_node.predecessor;
+                }
+                FlowNode::Branch(branch_node) => {
+                    node_id = branch_node.predecessor;
+                }
+                FlowNode::Phi(phi_node) => {
+                    self.pending.push(FlowState {
+                        node_id: phi_node.first_predecessor,
+                        constraints_range: constraints_range.clone(),
+                    });
+                    node_id = phi_node.second_predecessor;
+                }
+                FlowNode::Constraint(constraint_node) => {
+                    node_id = constraint_node.predecessor;
+                    self.constraints.push(constraint_node.test_expression);
+                    constraints_range.end += 1;
+                }
+            }
+        }
+    }
+}
+
+impl<'a> FusedIterator for ReachableDefinitionsIterator<'a> {}
+
+impl std::fmt::Display for FlowGraph {
+    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
+        writeln!(f, "flowchart TD")?;
+        for (id, node) in self.flow_nodes_by_id.iter_enumerated() {
+            write!(f, "  id{}", id.as_u32())?;
+            match node {
+                FlowNode::Start => writeln!(f, r"[\Start/]")?,
+                FlowNode::Definition(def_node) => {
+                    writeln!(f, r"(Define symbol {})", def_node.symbol_id.as_u32())?;
+                    writeln!(
+                        f,
+                        r"  id{}-->id{}",
+                        def_node.predecessor.as_u32(),
+                        id.as_u32()
+                    )?;
+                }
+                FlowNode::Branch(branch_node) => {
+                    writeln!(f, r"{{Branch}}")?;
+                    writeln!(
+                        f,
+                        r"  id{}-->id{}",
+                        branch_node.predecessor.as_u32(),
+                        id.as_u32()
+                    )?;
+                }
+                FlowNode::Phi(phi_node) => {
+                    writeln!(f, r"((Phi))")?;
+                    writeln!(
+                        f,
+                        r"  id{}-->id{}",
+                        phi_node.second_predecessor.as_u32(),
+                        id.as_u32()
+                    )?;
+                    writeln!(
+                        f,
+                        r"  id{}-->id{}",
+                        phi_node.first_predecessor.as_u32(),
+                        id.as_u32()
+                    )?;
+                }
+                FlowNode::Constraint(constraint_node) => {
+                    writeln!(f, r"((Constraint))")?;
+                    writeln!(
+                        f,
+                        r"  id{}-->id{}",
+                        constraint_node.predecessor.as_u32(),
+                        id.as_u32()
+                    )?;
+                }
+            }
+        }
+        Ok(())
+    }
+}

crates/red_knot/src/semantic/symbol_table.rs

@@ -0,0 +1,560 @@
+#![allow(dead_code)]
+
+use std::hash::{Hash, Hasher};
+use std::iter::{Copied, DoubleEndedIterator, FusedIterator};
+use std::num::NonZeroU32;
+
+use bitflags::bitflags;
+use hashbrown::hash_map::{Keys, RawEntryMut};
+use rustc_hash::{FxHashMap, FxHasher};
+
+use ruff_index::{newtype_index, IndexVec};
+
+use crate::ast_ids::NodeKey;
+use crate::module::ModuleName;
+use crate::semantic::{Definition, ExpressionId};
+use crate::Name;
+
+type Map<K, V> = hashbrown::HashMap<K, V, ()>;
+
+#[newtype_index]
+pub struct ScopeId;
+
+impl ScopeId {
+    pub fn scope(self, table: &SymbolTable) -> &Scope {
+        &table.scopes_by_id[self]
+    }
+}
+
+#[newtype_index]
+pub struct SymbolId;
+
+impl SymbolId {
+    pub fn symbol(self, table: &SymbolTable) -> &Symbol {
+        &table.symbols_by_id[self]
+    }
+}
+
+#[derive(Copy, Clone, Debug, PartialEq)]
+pub enum ScopeKind {
+    Module,
+    Annotation,
+    Class,
+    Function,
+}
+
+#[derive(Debug)]
+pub struct Scope {
+    name: Name,
+    kind: ScopeKind,
+    parent: Option<ScopeId>,
+    children: Vec<ScopeId>,
+    /// the definition (e.g. class or function) that created this scope
+    definition: Option<Definition>,
+    /// the symbol (e.g. class or function) that owns this scope
+    defining_symbol: Option<SymbolId>,
+    /// symbol IDs, hashed by symbol name
+    symbols_by_name: Map<SymbolId, ()>,
+}
+
+impl Scope {
+    pub fn name(&self) -> &str {
+        self.name.as_str()
+    }
+
+    pub fn kind(&self) -> ScopeKind {
+        self.kind
+    }
+
+    pub fn definition(&self) -> Option<Definition> {
+        self.definition.clone()
+    }
+
+    pub fn defining_symbol(&self) -> Option<SymbolId> {
+        self.defining_symbol
+    }
+}
+
+#[derive(Debug)]
+pub(crate) enum Kind {
+    FreeVar,
+    CellVar,
+    CellVarAssigned,
+    ExplicitGlobal,
+    ImplicitGlobal,
+}
+
+bitflags! {
+    #[derive(Copy,Clone,Debug)]
+    pub struct SymbolFlags: u8 {
+        const IS_USED         = 1 << 0;
+        const IS_DEFINED      = 1 << 1;
+        /// TODO: This flag is not yet set by anything
+        const MARKED_GLOBAL   = 1 << 2;
+        /// TODO: This flag is not yet set by anything
+        const MARKED_NONLOCAL = 1 << 3;
+    }
+}
+
+#[derive(Debug)]
+pub struct Symbol {
+    name: Name,
+    flags: SymbolFlags,
+    scope_id: ScopeId,
+    // kind: Kind,
+}
+
+impl Symbol {
+    pub fn name(&self) -> &str {
+        self.name.as_str()
+    }
+
+    pub fn scope_id(&self) -> ScopeId {
+        self.scope_id
+    }
+
+    /// Is the symbol used in its containing scope?
+    pub fn is_used(&self) -> bool {
+        self.flags.contains(SymbolFlags::IS_USED)
+    }
+
+    /// Is the symbol defined in its containing scope?
+    pub fn is_defined(&self) -> bool {
+        self.flags.contains(SymbolFlags::IS_DEFINED)
+    }
+
+    // TODO: implement Symbol.kind 2-pass analysis to categorize as: free-var, cell-var,
+    // explicit-global, implicit-global and implement Symbol.kind by modifying the preorder
+    // traversal code
+}
+
+#[derive(Debug, Clone)]
+pub enum Dependency {
+    Module(ModuleName),
+    Relative {
+        level: NonZeroU32,
+        module: Option<ModuleName>,
+    },
+}
+
+/// Table of all symbols in all scopes for a module.
+#[derive(Debug)]
+pub struct SymbolTable {
+    scopes_by_id: IndexVec<ScopeId, Scope>,
+    symbols_by_id: IndexVec<SymbolId, Symbol>,
+    /// the definitions for each symbol
+    defs: FxHashMap<SymbolId, Vec<Definition>>,
+    /// map of AST node (e.g. class/function def) to sub-scope it creates
+    scopes_by_node: FxHashMap<NodeKey, ScopeId>,
+    /// Maps expressions to their enclosing scope.
+    expression_scopes: IndexVec<ExpressionId, ScopeId>,
+    /// dependencies of this module
+    dependencies: Vec<Dependency>,
+}
+
+impl SymbolTable {
+    pub fn dependencies(&self) -> &[Dependency] {
+        &self.dependencies
+    }
+
+    pub const fn root_scope_id() -> ScopeId {
+        ScopeId::from_usize(0)
+    }
+
+    pub fn root_scope(&self) -> &Scope {
+        &self.scopes_by_id[SymbolTable::root_scope_id()]
+    }
+
+    pub fn symbol_ids_for_scope(&self, scope_id: ScopeId) -> Copied<Keys<SymbolId, ()>> {
+        self.scopes_by_id[scope_id].symbols_by_name.keys().copied()
+    }
+
+    pub fn symbols_for_scope(
+        &self,
+        scope_id: ScopeId,
+    ) -> SymbolIterator<Copied<Keys<SymbolId, ()>>> {
+        SymbolIterator {
+            table: self,
+            ids: self.symbol_ids_for_scope(scope_id),
+        }
+    }
+
+    pub fn root_symbol_ids(&self) -> Copied<Keys<SymbolId, ()>> {
+        self.symbol_ids_for_scope(SymbolTable::root_scope_id())
+    }
+
+    pub fn root_symbols(&self) -> SymbolIterator<Copied<Keys<SymbolId, ()>>> {
+        self.symbols_for_scope(SymbolTable::root_scope_id())
+    }
+
+    pub fn child_scope_ids_of(&self, scope_id: ScopeId) -> &[ScopeId] {
+        &self.scopes_by_id[scope_id].children
+    }
+
+    pub fn child_scopes_of(&self, scope_id: ScopeId) -> ScopeIterator<&[ScopeId]> {
+        ScopeIterator {
+            table: self,
+            ids: self.child_scope_ids_of(scope_id),
+        }
+    }
+
+    pub fn root_child_scope_ids(&self) -> &[ScopeId] {
+        self.child_scope_ids_of(SymbolTable::root_scope_id())
+    }
+
+    pub fn root_child_scopes(&self) -> ScopeIterator<&[ScopeId]> {
+        self.child_scopes_of(SymbolTable::root_scope_id())
+    }
+
+    pub fn symbol_id_by_name(&self, scope_id: ScopeId, name: &str) -> Option<SymbolId> {
+        let scope = &self.scopes_by_id[scope_id];
+        let hash = SymbolTable::hash_name(name);
+        let name = Name::new(name);
+        Some(
+            *scope
+                .symbols_by_name
+                .raw_entry()
+                .from_hash(hash, |symid| self.symbols_by_id[*symid].name == name)?
+                .0,
+        )
+    }
+
+    pub fn symbol_by_name(&self, scope_id: ScopeId, name: &str) -> Option<&Symbol> {
+        Some(&self.symbols_by_id[self.symbol_id_by_name(scope_id, name)?])
+    }
+
+    pub fn root_symbol_id_by_name(&self, name: &str) -> Option<SymbolId> {
+        self.symbol_id_by_name(SymbolTable::root_scope_id(), name)
+    }
+
+    pub fn root_symbol_by_name(&self, name: &str) -> Option<&Symbol> {
+        self.symbol_by_name(SymbolTable::root_scope_id(), name)
+    }
+
+    pub fn scope_id_of_symbol(&self, symbol_id: SymbolId) -> ScopeId {
+        self.symbols_by_id[symbol_id].scope_id
+    }
+
+    pub fn scope_of_symbol(&self, symbol_id: SymbolId) -> &Scope {
+        &self.scopes_by_id[self.scope_id_of_symbol(symbol_id)]
+    }
+
+    pub fn scope_id_of_expression(&self, expression: ExpressionId) -> ScopeId {
+        self.expression_scopes[expression]
+    }
+
+    pub fn scope_of_expression(&self, expr_id: ExpressionId) -> &Scope {
+        &self.scopes_by_id[self.scope_id_of_expression(expr_id)]
+    }
+
+    pub fn parent_scopes(
+        &self,
+        scope_id: ScopeId,
+    ) -> ScopeIterator<impl Iterator<Item = ScopeId> + '_> {
+        ScopeIterator {
+            table: self,
+            ids: std::iter::successors(Some(scope_id), |scope| self.scopes_by_id[*scope].parent),
+        }
+    }
+
+    pub fn parent_scope(&self, scope_id: ScopeId) -> Option<ScopeId> {
+        self.scopes_by_id[scope_id].parent
+    }
+
+    pub fn scope_id_for_node(&self, node_key: &NodeKey) -> ScopeId {
+        self.scopes_by_node[node_key]
+    }
+
+    pub fn definitions(&self, symbol_id: SymbolId) -> &[Definition] {
+        self.defs
+            .get(&symbol_id)
+            .map(std::vec::Vec::as_slice)
+            .unwrap_or_default()
+    }
+
+    pub fn all_definitions(&self) -> impl Iterator<Item = (SymbolId, &Definition)> + '_ {
+        self.defs
+            .iter()
+            .flat_map(|(sym_id, defs)| defs.iter().map(move |def| (*sym_id, def)))
+    }
+
+    fn hash_name(name: &str) -> u64 {
+        let mut hasher = FxHasher::default();
+        name.hash(&mut hasher);
+        hasher.finish()
+    }
+}
+
+pub struct SymbolIterator<'a, I> {
+    table: &'a SymbolTable,
+    ids: I,
+}
+
+impl<'a, I> Iterator for SymbolIterator<'a, I>
+where
+    I: Iterator<Item = SymbolId>,
+{
+    type Item = &'a Symbol;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        let id = self.ids.next()?;
+        Some(&self.table.symbols_by_id[id])
+    }
+
+    fn size_hint(&self) -> (usize, Option<usize>) {
+        self.ids.size_hint()
+    }
+}
+
+impl<'a, I> FusedIterator for SymbolIterator<'a, I> where
+    I: Iterator<Item = SymbolId> + FusedIterator
+{
+}
+
+impl<'a, I> DoubleEndedIterator for SymbolIterator<'a, I>
+where
+    I: Iterator<Item = SymbolId> + DoubleEndedIterator,
+{
+    fn next_back(&mut self) -> Option<Self::Item> {
+        let id = self.ids.next_back()?;
+        Some(&self.table.symbols_by_id[id])
+    }
+}
+
+// TODO maybe get rid of this and just do all data access via methods on ScopeId?
+pub struct ScopeIterator<'a, I> {
+    table: &'a SymbolTable,
+    ids: I,
+}
+
+/// iterate (`ScopeId`, `Scope`) pairs for given `ScopeId` iterator
+impl<'a, I> Iterator for ScopeIterator<'a, I>
+where
+    I: Iterator<Item = ScopeId>,
+{
+    type Item = (ScopeId, &'a Scope);
+
+    fn next(&mut self) -> Option<Self::Item> {
+        let id = self.ids.next()?;
+        Some((id, &self.table.scopes_by_id[id]))
+    }
+
+    fn size_hint(&self) -> (usize, Option<usize>) {
+        self.ids.size_hint()
+    }
+}
+
+impl<'a, I> FusedIterator for ScopeIterator<'a, I> where I: Iterator<Item = ScopeId> + FusedIterator {}
+
+impl<'a, I> DoubleEndedIterator for ScopeIterator<'a, I>
+where
+    I: Iterator<Item = ScopeId> + DoubleEndedIterator,
+{
+    fn next_back(&mut self) -> Option<Self::Item> {
+        let id = self.ids.next_back()?;
+        Some((id, &self.table.scopes_by_id[id]))
+    }
+}
+
+#[derive(Debug)]
+pub(super) struct SymbolTableBuilder {
+    symbol_table: SymbolTable,
+}
+
+impl SymbolTableBuilder {
+    pub(super) fn new() -> Self {
+        let mut table = SymbolTable {
+            scopes_by_id: IndexVec::new(),
+            symbols_by_id: IndexVec::new(),
+            defs: FxHashMap::default(),
+            scopes_by_node: FxHashMap::default(),
+            expression_scopes: IndexVec::new(),
+            dependencies: Vec::new(),
+        };
+        table.scopes_by_id.push(Scope {
+            name: Name::new("<module>"),
+            kind: ScopeKind::Module,
+            parent: None,
+            children: Vec::new(),
+            definition: None,
+            defining_symbol: None,
+            symbols_by_name: Map::default(),
+        });
+        Self {
+            symbol_table: table,
+        }
+    }
+
+    pub(super) fn finish(self) -> SymbolTable {
+        let mut symbol_table = self.symbol_table;
+        symbol_table.scopes_by_id.shrink_to_fit();
+        symbol_table.symbols_by_id.shrink_to_fit();
+        symbol_table.defs.shrink_to_fit();
+        symbol_table.scopes_by_node.shrink_to_fit();
+        symbol_table.expression_scopes.shrink_to_fit();
+        symbol_table.dependencies.shrink_to_fit();
+        symbol_table
+    }
+
+    pub(super) fn add_or_update_symbol(
+        &mut self,
+        scope_id: ScopeId,
+        name: &str,
+        flags: SymbolFlags,
+    ) -> SymbolId {
+        let hash = SymbolTable::hash_name(name);
+        let scope = &mut self.symbol_table.scopes_by_id[scope_id];
+        let name = Name::new(name);
+
+        let entry = scope
+            .symbols_by_name
+            .raw_entry_mut()
+            .from_hash(hash, |existing| {
+                self.symbol_table.symbols_by_id[*existing].name == name
+            });
+
+        match entry {
+            RawEntryMut::Occupied(entry) => {
+                if let Some(symbol) = self.symbol_table.symbols_by_id.get_mut(*entry.key()) {
+                    symbol.flags.insert(flags);
+                };
+                *entry.key()
+            }
+            RawEntryMut::Vacant(entry) => {
+                let id = self.symbol_table.symbols_by_id.push(Symbol {
+                    name,
+                    flags,
+                    scope_id,
+                });
+                entry.insert_with_hasher(hash, id, (), |symid| {
+                    SymbolTable::hash_name(&self.symbol_table.symbols_by_id[*symid].name)
+                });
+                id
+            }
+        }
+    }
+
+    pub(super) fn add_definition(&mut self, symbol_id: SymbolId, definition: Definition) {
+        self.symbol_table
+            .defs
+            .entry(symbol_id)
+            .or_default()
+            .push(definition);
+    }
+
+    pub(super) fn add_child_scope(
+        &mut self,
+        parent_scope_id: ScopeId,
+        name: &str,
+        kind: ScopeKind,
+        definition: Option<Definition>,
+        defining_symbol: Option<SymbolId>,
+    ) -> ScopeId {
+        let new_scope_id = self.symbol_table.scopes_by_id.push(Scope {
+            name: Name::new(name),
+            kind,
+            parent: Some(parent_scope_id),
+            children: Vec::new(),
+            definition,
+            defining_symbol,
+            symbols_by_name: Map::default(),
+        });
+        let parent_scope = &mut self.symbol_table.scopes_by_id[parent_scope_id];
+        parent_scope.children.push(new_scope_id);
+        new_scope_id
+    }
+
+    pub(super) fn record_scope_for_node(&mut self, node_key: NodeKey, scope_id: ScopeId) {
+        self.symbol_table.scopes_by_node.insert(node_key, scope_id);
+    }
+
+    pub(super) fn add_dependency(&mut self, dependency: Dependency) {
+        self.symbol_table.dependencies.push(dependency);
+    }
+
+    /// Records the scope for the current expression
+    pub(super) fn record_expression(&mut self, scope: ScopeId) -> ExpressionId {
+        self.symbol_table.expression_scopes.push(scope)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{ScopeKind, SymbolFlags, SymbolTable, SymbolTableBuilder};
+
+    #[test]
+    fn insert_same_name_symbol_twice() {
+        let mut builder = SymbolTableBuilder::new();
+        let root_scope_id = SymbolTable::root_scope_id();
+        let symbol_id_1 =
+            builder.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::IS_DEFINED);
+        let symbol_id_2 = builder.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::IS_USED);
+        let table = builder.finish();
+
+        assert_eq!(symbol_id_1, symbol_id_2);
+        assert!(symbol_id_1.symbol(&table).is_used(), "flags must merge");
+        assert!(symbol_id_1.symbol(&table).is_defined(), "flags must merge");
+    }
+
+    #[test]
+    fn insert_different_named_symbols() {
+        let mut builder = SymbolTableBuilder::new();
+        let root_scope_id = SymbolTable::root_scope_id();
+        let symbol_id_1 = builder.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::empty());
+        let symbol_id_2 = builder.add_or_update_symbol(root_scope_id, "bar", SymbolFlags::empty());
+
+        assert_ne!(symbol_id_1, symbol_id_2);
+    }
+
+    #[test]
+    fn add_child_scope_with_symbol() {
+        let mut builder = SymbolTableBuilder::new();
+        let root_scope_id = SymbolTable::root_scope_id();
+        let foo_symbol_top =
+            builder.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::empty());
+        let c_scope = builder.add_child_scope(root_scope_id, "C", ScopeKind::Class, None, None);
+        let foo_symbol_inner = builder.add_or_update_symbol(c_scope, "foo", SymbolFlags::empty());
+
+        assert_ne!(foo_symbol_top, foo_symbol_inner);
+    }
+
+    #[test]
+    fn scope_from_id() {
+        let table = SymbolTableBuilder::new().finish();
+        let root_scope_id = SymbolTable::root_scope_id();
+        let scope = root_scope_id.scope(&table);
+
+        assert_eq!(scope.name.as_str(), "<module>");
+        assert_eq!(scope.kind, ScopeKind::Module);
+    }
+
+    #[test]
+    fn symbol_from_id() {
+        let mut builder = SymbolTableBuilder::new();
+        let root_scope_id = SymbolTable::root_scope_id();
+        let foo_symbol_id =
+            builder.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::empty());
+        let table = builder.finish();
+        let symbol = foo_symbol_id.symbol(&table);
+
+        assert_eq!(symbol.name(), "foo");
+    }
+
+    #[test]
+    fn bigger_symbol_table() {
+        let mut builder = SymbolTableBuilder::new();
+        let root_scope_id = SymbolTable::root_scope_id();
+        let foo_symbol_id =
+            builder.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::empty());
+        builder.add_or_update_symbol(root_scope_id, "bar", SymbolFlags::empty());
+        builder.add_or_update_symbol(root_scope_id, "baz", SymbolFlags::empty());
+        builder.add_or_update_symbol(root_scope_id, "qux", SymbolFlags::empty());
+        let table = builder.finish();
+
+        let foo_symbol_id_2 = table
+            .root_symbol_id_by_name("foo")
+            .expect("foo symbol to be found");
+
+        assert_eq!(foo_symbol_id_2, foo_symbol_id);
+    }
+}

crates/red_knot/src/semantic/types/infer.rs

@@ -0,0 +1,762 @@
+#![allow(dead_code)]
+
+use ruff_python_ast as ast;
+use ruff_python_ast::AstNode;
+use std::fmt::Debug;
+
+use crate::db::{QueryResult, SemanticDb, SemanticJar};
+
+use crate::module::{resolve_module, ModuleName};
+use crate::parse::parse;
+use crate::semantic::types::{ModuleTypeId, Type};
+use crate::semantic::{
+    resolve_global_symbol, semantic_index, ConstrainedDefinition, Definition, GlobalSymbolId,
+    ImportDefinition, ImportFromDefinition,
+};
+use crate::{FileId, Name};
+
+// FIXME: Figure out proper dead-lock free synchronisation now that this takes `&db` instead of `&mut db`.
+/// Resolve the public-facing type for a symbol (the type seen by other scopes: other modules, or
+/// nested functions). Because calls to nested functions and imports can occur anywhere in control
+/// flow, this type must be conservative and consider all definitions of the symbol that could
+/// possibly be seen by another scope. Currently we take the most conservative approach, which is
+/// the union of all definitions. We may be able to narrow this in future to eliminate definitions
+/// which can't possibly (or at least likely) be seen by any other scope, so that e.g. we could
+/// infer `Literal["1"]` instead of `Literal[1] | Literal["1"]` for `x` in `x = x; x = str(x);`.
+#[tracing::instrument(level = "trace", skip(db))]
+pub fn infer_symbol_public_type(db: &dyn SemanticDb, symbol: GlobalSymbolId) -> QueryResult<Type> {
+    let index = semantic_index(db, symbol.file_id)?;
+    let defs = index.symbol_table().definitions(symbol.symbol_id).to_vec();
+    let jar: &SemanticJar = db.jar()?;
+
+    if let Some(ty) = jar.type_store.get_cached_symbol_public_type(symbol) {
+        return Ok(ty);
+    }
+
+    let ty = infer_type_from_definitions(db, symbol, defs.iter().cloned())?;
+
+    jar.type_store.cache_symbol_public_type(symbol, ty);
+
+    // TODO record dependencies
+    Ok(ty)
+}
+
+/// Infer type of a symbol as union of the given `Definitions`.
+fn infer_type_from_definitions<T>(
+    db: &dyn SemanticDb,
+    symbol: GlobalSymbolId,
+    definitions: T,
+) -> QueryResult<Type>
+where
+    T: Debug + IntoIterator<Item = Definition>,
+{
+    infer_type_from_constrained_definitions(
+        db,
+        symbol,
+        definitions
+            .into_iter()
+            .map(|definition| ConstrainedDefinition {
+                definition,
+                constraints: vec![],
+            }),
+    )
+}
+
+/// Infer type of a symbol as union of the given `ConstrainedDefinitions`.
+fn infer_type_from_constrained_definitions<T>(
+    db: &dyn SemanticDb,
+    symbol: GlobalSymbolId,
+    constrained_definitions: T,
+) -> QueryResult<Type>
+where
+    T: IntoIterator<Item = ConstrainedDefinition>,
+{
+    let jar: &SemanticJar = db.jar()?;
+    let mut tys = constrained_definitions
+        .into_iter()
+        .map(|def| infer_constrained_definition_type(db, symbol, def.clone()))
+        .peekable();
+    if let Some(first) = tys.next() {
+        if tys.peek().is_some() {
+            Ok(jar.type_store.add_union(
+                symbol.file_id,
+                &Iterator::chain(std::iter::once(first), tys).collect::<QueryResult<Vec<_>>>()?,
+            ))
+        } else {
+            first
+        }
+    } else {
+        Ok(Type::Unknown)
+    }
+}
+
+/// Infer type for a ConstrainedDefinition (intersection of the definition type and the
+/// constraints)
+#[tracing::instrument(level = "trace", skip(db))]
+pub fn infer_constrained_definition_type(
+    db: &dyn SemanticDb,
+    symbol: GlobalSymbolId,
+    constrained_definition: ConstrainedDefinition,
+) -> QueryResult<Type> {
+    let ConstrainedDefinition {
+        definition,
+        constraints,
+    } = constrained_definition;
+    let index = semantic_index(db, symbol.file_id)?;
+    let parsed = parse(db.upcast(), symbol.file_id)?;
+    let mut intersected_types = vec![infer_definition_type(db, symbol, definition)?];
+    for constraint in constraints {
+        if let Some(constraint_type) = infer_constraint_type(
+            db,
+            symbol,
+            index.resolve_expression_id(parsed.syntax(), constraint),
+        )? {
+            intersected_types.push(constraint_type);
+        }
+    }
+    let jar: &SemanticJar = db.jar()?;
+    Ok(jar
+        .type_store
+        .add_intersection(symbol.file_id, &intersected_types, &[]))
+}
+
+/// Infer a type for a Definition
+#[tracing::instrument(level = "trace", skip(db))]
+pub fn infer_definition_type(
+    db: &dyn SemanticDb,
+    symbol: GlobalSymbolId,
+    definition: Definition,
+) -> QueryResult<Type> {
+    let jar: &SemanticJar = db.jar()?;
+    let type_store = &jar.type_store;
+    let file_id = symbol.file_id;
+
+    match definition {
+        Definition::Unbound => Ok(Type::Unbound),
+        Definition::Import(ImportDefinition {
+            module: module_name,
+        }) => {
+            if let Some(module) = resolve_module(db, module_name.clone())? {
+                Ok(Type::Module(ModuleTypeId { module, file_id }))
+            } else {
+                Ok(Type::Unknown)
+            }
+        }
+        Definition::ImportFrom(ImportFromDefinition {
+            module,
+            name,
+            level,
+        }) => {
+            // TODO relative imports
+            assert!(matches!(level, 0));
+            let module_name = ModuleName::new(module.as_ref().expect("TODO relative imports"));
+            let Some(module) = resolve_module(db, module_name.clone())? else {
+                return Ok(Type::Unknown);
+            };
+
+            if let Some(remote_symbol) = resolve_global_symbol(db, module, &name)? {
+                infer_symbol_public_type(db, remote_symbol)
+            } else {
+                Ok(Type::Unknown)
+            }
+        }
+        Definition::ClassDef(node_key) => {
+            if let Some(ty) = type_store.get_cached_node_type(file_id, node_key.erased()) {
+                Ok(ty)
+            } else {
+                let parsed = parse(db.upcast(), file_id)?;
+                let ast = parsed.syntax();
+                let index = semantic_index(db, file_id)?;
+                let node = node_key.resolve_unwrap(ast.as_any_node_ref());
+
+                let mut bases = Vec::with_capacity(node.bases().len());
+
+                for base in node.bases() {
+                    bases.push(infer_expr_type(db, file_id, base)?);
+                }
+                let scope_id = index.symbol_table().scope_id_for_node(node_key.erased());
+                let ty = type_store.add_class(file_id, &node.name.id, scope_id, bases);
+                type_store.cache_node_type(file_id, *node_key.erased(), ty);
+                Ok(ty)
+            }
+        }
+        Definition::FunctionDef(node_key) => {
+            if let Some(ty) = type_store.get_cached_node_type(file_id, node_key.erased()) {
+                Ok(ty)
+            } else {
+                let parsed = parse(db.upcast(), file_id)?;
+                let ast = parsed.syntax();
+                let index = semantic_index(db, file_id)?;
+                let node = node_key
+                    .resolve(ast.as_any_node_ref())
+                    .expect("node key should resolve");
+
+                let decorator_tys = node
+                    .decorator_list
+                    .iter()
+                    .map(|decorator| infer_expr_type(db, file_id, &decorator.expression))
+                    .collect::<QueryResult<_>>()?;
+                let scope_id = index.symbol_table().scope_id_for_node(node_key.erased());
+                let ty = type_store.add_function(
+                    file_id,
+                    &node.name.id,
+                    symbol.symbol_id,
+                    scope_id,
+                    decorator_tys,
+                );
+                type_store.cache_node_type(file_id, *node_key.erased(), ty);
+                Ok(ty)
+            }
+        }
+        Definition::Assignment(node_key) => {
+            let parsed = parse(db.upcast(), file_id)?;
+            let ast = parsed.syntax();
+            let node = node_key.resolve_unwrap(ast.as_any_node_ref());
+            // TODO handle unpacking assignment
+            infer_expr_type(db, file_id, &node.value)
+        }
+        Definition::AnnotatedAssignment(node_key) => {
+            let parsed = parse(db.upcast(), file_id)?;
+            let ast = parsed.syntax();
+            let node = node_key.resolve_unwrap(ast.as_any_node_ref());
+            // TODO actually look at the annotation
+            let Some(value) = &node.value else {
+                return Ok(Type::Unknown);
+            };
+            // TODO handle unpacking assignment
+            infer_expr_type(db, file_id, value)
+        }
+        Definition::NamedExpr(node_key) => {
+            let parsed = parse(db.upcast(), file_id)?;
+            let ast = parsed.syntax();
+            let node = node_key.resolve_unwrap(ast.as_any_node_ref());
+            infer_expr_type(db, file_id, &node.value)
+        }
+    }
+}
+
+/// Return the type that the given constraint (an expression from a control-flow test) requires the
+/// given symbol to have. For example, returns the Type "~None" as the constraint type if given the
+/// symbol ID for x and the expression ID for `x is not None`. Returns (Rust) None if the given
+/// expression applies no constraints on the given symbol.
+#[tracing::instrument(level = "trace", skip(db))]
+fn infer_constraint_type(
+    db: &dyn SemanticDb,
+    symbol_id: GlobalSymbolId,
+    // TODO this should preferably take an &ast::Expr instead of AnyNodeRef
+    expression: ast::AnyNodeRef,
+) -> QueryResult<Option<Type>> {
+    let file_id = symbol_id.file_id;
+    let index = semantic_index(db, file_id)?;
+    let jar: &SemanticJar = db.jar()?;
+    let symbol_name = symbol_id.symbol_id.symbol(&index.symbol_table).name();
+    // TODO narrowing attributes
+    // TODO narrowing dict keys
+    // TODO isinstance, ==/!=, type(...), literals, bools...
+    match expression {
+        ast::AnyNodeRef::ExprCompare(ast::ExprCompare {
+            left,
+            ops,
+            comparators,
+            ..
+        }) => {
+            // TODO chained comparisons
+            match left.as_ref() {
+                ast::Expr::Name(ast::ExprName { id, .. }) if id == symbol_name => match ops[0] {
+                    ast::CmpOp::Is | ast::CmpOp::IsNot => {
+                        Ok(match infer_expr_type(db, file_id, &comparators[0])? {
+                            Type::None => Some(Type::None),
+                            _ => None,
+                        }
+                        .map(|ty| {
+                            if matches!(ops[0], ast::CmpOp::IsNot) {
+                                jar.type_store.add_intersection(file_id, &[], &[ty])
+                            } else {
+                                ty
+                            }
+                        }))
+                    }
+                    _ => Ok(None),
+                },
+                _ => Ok(None),
+            }
+        }
+        _ => Ok(None),
+    }
+}
+
+/// Infer type of the given expression.
+fn infer_expr_type(db: &dyn SemanticDb, file_id: FileId, expr: &ast::Expr) -> QueryResult<Type> {
+    // TODO cache the resolution of the type on the node
+    let index = semantic_index(db, file_id)?;
+    match expr {
+        ast::Expr::NoneLiteral(_) => Ok(Type::None),
+        ast::Expr::NumberLiteral(ast::ExprNumberLiteral { value, .. }) => {
+            match value {
+                ast::Number::Int(n) => {
+                    // TODO support big int literals
+                    Ok(n.as_i64().map(Type::IntLiteral).unwrap_or(Type::Unknown))
+                }
+                // TODO builtins.float or builtins.complex
+                _ => Ok(Type::Unknown),
+            }
+        }
+        ast::Expr::Name(name) => {
+            // TODO look up in the correct scope, don't assume global
+            if let Some(symbol_id) = index.symbol_table().root_symbol_id_by_name(&name.id) {
+                infer_type_from_constrained_definitions(
+                    db,
+                    GlobalSymbolId { file_id, symbol_id },
+                    index.reachable_definitions(symbol_id, expr),
+                )
+            } else {
+                Ok(Type::Unknown)
+            }
+        }
+        ast::Expr::Attribute(ast::ExprAttribute { value, attr, .. }) => {
+            let value_type = infer_expr_type(db, file_id, value)?;
+            let attr_name = &Name::new(&attr.id);
+            value_type
+                .get_member(db, attr_name)
+                .map(|ty| ty.unwrap_or(Type::Unknown))
+        }
+        ast::Expr::BinOp(ast::ExprBinOp {
+            left, op, right, ..
+        }) => {
+            let left_ty = infer_expr_type(db, file_id, left)?;
+            let right_ty = infer_expr_type(db, file_id, right)?;
+            // TODO add reverse bin op support if right <: left
+            left_ty.resolve_bin_op(db, *op, right_ty)
+        }
+        ast::Expr::Named(ast::ExprNamed { value, .. }) => infer_expr_type(db, file_id, value),
+        ast::Expr::If(ast::ExprIf { body, orelse, .. }) => {
+            // TODO detect statically known truthy or falsy test
+            let body_ty = infer_expr_type(db, file_id, body)?;
+            let else_ty = infer_expr_type(db, file_id, orelse)?;
+            let jar: &SemanticJar = db.jar()?;
+            Ok(jar.type_store.add_union(file_id, &[body_ty, else_ty]))
+        }
+        _ => todo!("expression type resolution for {:?}", expr),
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    use std::path::PathBuf;
+
+    use crate::db::tests::TestDb;
+    use crate::db::{HasJar, SemanticJar};
+    use crate::module::{
+        resolve_module, set_module_search_paths, ModuleName, ModuleResolutionInputs,
+    };
+    use crate::semantic::{infer_symbol_public_type, resolve_global_symbol, Type};
+    use crate::Name;
+
+    // TODO with virtual filesystem we shouldn't have to write files to disk for these
+    // tests
+
+    struct TestCase {
+        temp_dir: tempfile::TempDir,
+        db: TestDb,
+
+        src: PathBuf,
+    }
+
+    fn create_test() -> std::io::Result<TestCase> {
+        let temp_dir = tempfile::tempdir()?;
+
+        let src = temp_dir.path().join("src");
+        std::fs::create_dir(&src)?;
+        let src = src.canonicalize()?;
+
+        let search_paths = ModuleResolutionInputs {
+            extra_paths: vec![],
+            workspace_root: src.clone(),
+            site_packages: None,
+            custom_typeshed: None,
+        };
+
+        let mut db = TestDb::default();
+        set_module_search_paths(&mut db, search_paths);
+
+        Ok(TestCase { temp_dir, db, src })
+    }
+
+    fn write_to_path(case: &TestCase, relative_path: &str, contents: &str) -> anyhow::Result<()> {
+        let path = case.src.join(relative_path);
+        std::fs::write(path, contents)?;
+        Ok(())
+    }
+
+    fn get_public_type(
+        case: &TestCase,
+        module_name: &str,
+        variable_name: &str,
+    ) -> anyhow::Result<Type> {
+        let db = &case.db;
+        let module = resolve_module(db, ModuleName::new(module_name))?.expect("Module to exist");
+        let symbol = resolve_global_symbol(db, module, variable_name)?.expect("symbol to exist");
+
+        Ok(infer_symbol_public_type(db, symbol)?)
+    }
+
+    fn assert_public_type(
+        case: &TestCase,
+        module_name: &str,
+        variable_name: &str,
+        type_name: &str,
+    ) -> anyhow::Result<()> {
+        let ty = get_public_type(case, module_name, variable_name)?;
+
+        let jar = HasJar::<SemanticJar>::jar(&case.db)?;
+        assert_eq!(format!("{}", ty.display(&jar.type_store)), type_name);
+        Ok(())
+    }
+
+    #[test]
+    fn follow_import_to_class() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(&case, "a.py", "from b import C as D; E = D")?;
+        write_to_path(&case, "b.py", "class C: pass")?;
+
+        assert_public_type(&case, "a", "E", "Literal[C]")
+    }
+
+    #[test]
+    fn resolve_base_class_by_name() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "mod.py",
+            "
+                class Base: pass
+                class Sub(Base): pass
+            ",
+        )?;
+
+        let ty = get_public_type(&case, "mod", "Sub")?;
+
+        let Type::Class(class_id) = ty else {
+            panic!("Sub is not a Class")
+        };
+        let jar = HasJar::<SemanticJar>::jar(&case.db)?;
+        let base_names: Vec<_> = jar
+            .type_store
+            .get_class(class_id)
+            .bases()
+            .iter()
+            .map(|base_ty| format!("{}", base_ty.display(&jar.type_store)))
+            .collect();
+
+        assert_eq!(base_names, vec!["Literal[Base]"]);
+
+        Ok(())
+    }
+
+    #[test]
+    fn resolve_method() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "mod.py",
+            "
+                class C:
+                    def f(self): pass
+            ",
+        )?;
+
+        let ty = get_public_type(&case, "mod", "C")?;
+
+        let Type::Class(class_id) = ty else {
+            panic!("C is not a Class");
+        };
+
+        let member_ty = class_id
+            .get_own_class_member(&case.db, &Name::new("f"))
+            .expect("C.f to resolve");
+
+        let Some(Type::Function(func_id)) = member_ty else {
+            panic!("C.f is not a Function");
+        };
+
+        let jar = HasJar::<SemanticJar>::jar(&case.db)?;
+        let function = jar.type_store.get_function(func_id);
+        assert_eq!(function.name(), "f");
+
+        Ok(())
+    }
+
+    #[test]
+    fn resolve_module_member() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(&case, "a.py", "import b; D = b.C")?;
+        write_to_path(&case, "b.py", "class C: pass")?;
+
+        assert_public_type(&case, "a", "D", "Literal[C]")
+    }
+
+    #[test]
+    fn resolve_literal() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(&case, "a.py", "x = 1")?;
+
+        assert_public_type(&case, "a", "x", "Literal[1]")
+    }
+
+    #[test]
+    fn resolve_union() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                if flag:
+                    x = 1
+                else:
+                    x = 2
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[1, 2]")
+    }
+
+    #[test]
+    fn resolve_visible_def() -> anyhow::Result<()> {
+        let case = create_test()?;
+        write_to_path(&case, "a.py", "y = 1; y = 2; x = y")?;
+        assert_public_type(&case, "a", "x", "Literal[2]")
+    }
+
+    #[test]
+    fn join_paths() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                y = 1
+                y = 2
+                if flag:
+                    y = 3
+                x = y
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[2, 3]")
+    }
+
+    #[test]
+    fn maybe_unbound() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                if flag:
+                    y = 1
+                x = y
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[1] | Unbound")
+    }
+
+    #[test]
+    fn if_elif_else() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                y = 1
+                y = 2
+                if flag:
+                    y = 3
+                elif flag2:
+                    y = 4
+                else:
+                    r = y
+                    y = 5
+                    s = y
+                x = y
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[3, 4, 5]")?;
+        assert_public_type(&case, "a", "r", "Literal[2]")?;
+        assert_public_type(&case, "a", "s", "Literal[5]")
+    }
+
+    #[test]
+    fn if_elif() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                y = 1
+                y = 2
+                if flag:
+                    y = 3
+                elif flag2:
+                    y = 4
+                x = y
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[2, 3, 4]")
+    }
+
+    #[test]
+    fn literal_int_arithmetic() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                a = 2 + 1
+                b = a - 4
+                c = a * b
+                d = c / 3
+                e = 5 % 3
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "a", "Literal[3]")?;
+        assert_public_type(&case, "a", "b", "Literal[-1]")?;
+        assert_public_type(&case, "a", "c", "Literal[-3]")?;
+        assert_public_type(&case, "a", "d", "Literal[-1]")?;
+        assert_public_type(&case, "a", "e", "Literal[2]")
+    }
+
+    #[test]
+    fn walrus() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                x = (y := 1) + 1
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[2]")?;
+        assert_public_type(&case, "a", "y", "Literal[1]")
+    }
+
+    #[test]
+    fn ifexpr() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                x = 1 if flag else 2
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[1, 2]")
+    }
+
+    #[test]
+    fn ifexpr_walrus() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                y = z = 0
+                x = (y := 1) if flag else (z := 2)
+                a = y
+                b = z
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[1, 2]")?;
+        assert_public_type(&case, "a", "a", "Literal[0, 1]")?;
+        assert_public_type(&case, "a", "b", "Literal[0, 2]")
+    }
+
+    #[test]
+    fn ifexpr_walrus_2() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                y = 0
+                (y := 1) if flag else (y := 2)
+                a = y
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "a", "Literal[1, 2]")
+    }
+
+    #[test]
+    fn ifexpr_nested() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                x = 1 if flag else 2 if flag2 else 3
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[1, 2, 3]")
+    }
+
+    #[test]
+    fn none() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                x = 1 if flag else None
+            ",
+        )?;
+
+        assert_public_type(&case, "a", "x", "Literal[1] | None")
+    }
+
+    #[test]
+    fn narrow_none() -> anyhow::Result<()> {
+        let case = create_test()?;
+
+        write_to_path(
+            &case,
+            "a.py",
+            "
+                x = 1 if flag else None
+                y = 0
+                if x is not None:
+                    y = x
+                z = y
+            ",
+        )?;
+
+        // TODO normalization of unions and intersections: this type is technically correct but
+        // begging for normalization
+        assert_public_type(&case, "a", "z", "Literal[0] | Literal[1] | None & ~None")
+    }
+}

crates/red_knot/src/source.rs

@@ -1,18 +1,17 @@
-use crate::cache::KeyValueCache;
-use crate::db::{HasJar, SourceDb, SourceJar};
-use ruff_notebook::Notebook;
-use ruff_python_ast::PySourceType;
 use std::ops::{Deref, DerefMut};
 use std::sync::Arc;
 
+use ruff_notebook::Notebook;
+use ruff_python_ast::PySourceType;
+
+use crate::cache::KeyValueCache;
+use crate::db::{QueryResult, SourceDb};
 use crate::files::FileId;
 
 #[tracing::instrument(level = "debug", skip(db))]
-pub(crate) fn source_text<Db>(db: &Db, file_id: FileId) -> Source
-where
-    Db: SourceDb + HasJar<SourceJar>,
-{
-    let sources = &db.jar().sources;
+pub(crate) fn source_text(db: &dyn SourceDb, file_id: FileId) -> QueryResult<Source> {
+    let jar = db.jar()?;
+    let sources = &jar.sources;
 
     sources.get(&file_id, |file_id| {
         let path = db.file_path(*file_id);
@@ -43,7 +42,7 @@ where
             }
         };
 
-        Source { kind }
+        Ok(Source { kind })
     })
 }
 
@@ -54,6 +53,16 @@ pub enum SourceKind {
     IpyNotebook(Arc<Notebook>),
 }
 
+impl<'a> From<&'a SourceKind> for PySourceType {
+    fn from(value: &'a SourceKind) -> Self {
+        match value {
+            SourceKind::Python(_) => PySourceType::Python,
+            SourceKind::Stub(_) => PySourceType::Stub,
+            SourceKind::IpyNotebook(_) => PySourceType::Ipynb,
+        }
+    }
+}
+
 #[derive(Debug, Clone, PartialEq)]
 pub struct Source {
     kind: SourceKind,

crates/red_knot/src/symbols.rs

@@ -1,816 +0,0 @@
-#![allow(dead_code)]
-
-use std::hash::{Hash, Hasher};
-use std::iter::{Copied, DoubleEndedIterator, FusedIterator};
-use std::ops::{Deref, DerefMut};
-use std::sync::Arc;
-
-use hashbrown::hash_map::{Keys, RawEntryMut};
-use rustc_hash::{FxHashMap, FxHasher};
-
-use ruff_index::{newtype_index, IndexVec};
-use ruff_python_ast as ast;
-use ruff_python_ast::visitor::preorder::PreorderVisitor;
-
-use crate::ast_ids::TypedNodeKey;
-use crate::cache::KeyValueCache;
-use crate::db::{HasJar, SemanticDb, SemanticJar};
-use crate::files::FileId;
-use crate::module::{Module, ModuleName};
-use crate::Name;
-
-#[allow(unreachable_pub)]
-#[tracing::instrument(level = "debug", skip(db))]
-pub fn symbol_table<Db>(db: &Db, file_id: FileId) -> Arc<SymbolTable>
-where
-    Db: SemanticDb + HasJar<SemanticJar>,
-{
-    let jar = db.jar();
-
-    jar.symbol_tables.get(&file_id, |_| {
-        let parsed = db.parse(file_id);
-        Arc::from(SymbolTable::from_ast(parsed.ast()))
-    })
-}
-
-type Map<K, V> = hashbrown::HashMap<K, V, ()>;
-
-#[newtype_index]
-pub(crate) struct ScopeId;
-
-impl ScopeId {
-    pub(crate) fn scope(self, table: &SymbolTable) -> &Scope {
-        &table.scopes_by_id[self]
-    }
-}
-
-#[newtype_index]
-pub struct SymbolId;
-
-impl SymbolId {
-    pub(crate) fn symbol(self, table: &SymbolTable) -> &Symbol {
-        &table.symbols_by_id[self]
-    }
-}
-
-#[derive(Copy, Clone, Debug, PartialEq)]
-pub(crate) enum ScopeKind {
-    Module,
-    Annotation,
-    Class,
-    Function,
-}
-
-#[derive(Debug)]
-pub(crate) struct Scope {
-    name: Name,
-    kind: ScopeKind,
-    child_scopes: Vec<ScopeId>,
-    // symbol IDs, hashed by symbol name
-    symbols_by_name: Map<SymbolId, ()>,
-}
-
-impl Scope {
-    pub(crate) fn name(&self) -> &str {
-        self.name.as_str()
-    }
-
-    pub(crate) fn kind(&self) -> ScopeKind {
-        self.kind
-    }
-}
-
-#[derive(Debug)]
-pub(crate) struct Symbol {
-    name: Name,
-}
-
-impl Symbol {
-    pub(crate) fn name(&self) -> &str {
-        self.name.as_str()
-    }
-}
-
-// TODO storing TypedNodeKey for definitions means we have to search to find them again in the AST;
-// this is at best O(log n). If looking up definitions is a bottleneck we should look for
-// alternatives here.
-#[derive(Debug)]
-pub(crate) enum Definition {
-    // For the import cases, we don't need reference to any arbitrary AST subtrees (annotations,
-    // RHS), and referencing just the import statement node is imprecise (a single import statement
-    // can assign many symbols, we'd have to re-search for the one we care about), so we just copy
-    // the small amount of information we need from the AST.
-    Import(ImportDefinition),
-    ImportFrom(ImportFromDefinition),
-    ClassDef(TypedNodeKey<ast::StmtClassDef>),
-    FunctionDef(TypedNodeKey<ast::StmtFunctionDef>),
-    Assignment(TypedNodeKey<ast::StmtAssign>),
-    AnnotatedAssignment(TypedNodeKey<ast::StmtAnnAssign>),
-    // TODO with statements, except handlers, function args...
-}
-
-#[derive(Debug)]
-pub(crate) struct ImportDefinition {
-    pub(crate) module: Name,
-}
-
-#[derive(Debug)]
-pub(crate) struct ImportFromDefinition {
-    pub(crate) module: Option<Name>,
-    pub(crate) name: Name,
-    pub(crate) level: u32,
-}
-
-#[derive(Debug, Clone)]
-pub(crate) enum Dependency {
-    Module(Name),
-    Relative { level: u32, module: Option<Name> },
-}
-
-impl Dependency {
-    pub(crate) fn module_name<Db>(&self, db: &Db, relative_to: Option<Module>) -> Option<ModuleName>
-    where
-        Db: SemanticDb + HasJar<SemanticJar>,
-    {
-        match self {
-            Dependency::Module(name) => Some(ModuleName::new(name.as_str())),
-            Dependency::Relative { level, module } => {
-                relative_to?.relative_name(db, *level, module.as_deref())
-            }
-        }
-    }
-}
-
-/// Table of all symbols in all scopes for a module.
-#[derive(Debug)]
-pub struct SymbolTable {
-    scopes_by_id: IndexVec<ScopeId, Scope>,
-    symbols_by_id: IndexVec<SymbolId, Symbol>,
-    defs: FxHashMap<SymbolId, Vec<Definition>>,
-    dependencies: Vec<Dependency>,
-}
-
-impl SymbolTable {
-    pub(crate) fn from_ast(module: &ast::ModModule) -> Self {
-        let root_scope_id = SymbolTable::root_scope_id();
-        let mut builder = SymbolTableBuilder {
-            table: SymbolTable::new(),
-            scopes: vec![root_scope_id],
-        };
-        builder.visit_body(&module.body);
-        builder.table
-    }
-
-    pub(crate) fn new() -> Self {
-        let mut table = SymbolTable {
-            scopes_by_id: IndexVec::new(),
-            symbols_by_id: IndexVec::new(),
-            defs: FxHashMap::default(),
-            dependencies: Vec::new(),
-        };
-        table.scopes_by_id.push(Scope {
-            name: Name::new("<module>"),
-            kind: ScopeKind::Module,
-            child_scopes: Vec::new(),
-            symbols_by_name: Map::default(),
-        });
-        table
-    }
-
-    pub(crate) fn dependencies(&self) -> &[Dependency] {
-        &self.dependencies
-    }
-
-    pub(crate) const fn root_scope_id() -> ScopeId {
-        ScopeId::from_usize(0)
-    }
-
-    pub(crate) fn root_scope(&self) -> &Scope {
-        &self.scopes_by_id[SymbolTable::root_scope_id()]
-    }
-
-    pub(crate) fn symbol_ids_for_scope(&self, scope_id: ScopeId) -> Copied<Keys<SymbolId, ()>> {
-        self.scopes_by_id[scope_id].symbols_by_name.keys().copied()
-    }
-
-    pub(crate) fn symbols_for_scope(
-        &self,
-        scope_id: ScopeId,
-    ) -> SymbolIterator<Copied<Keys<SymbolId, ()>>> {
-        SymbolIterator {
-            table: self,
-            ids: self.symbol_ids_for_scope(scope_id),
-        }
-    }
-
-    pub(crate) fn root_symbol_ids(&self) -> Copied<Keys<SymbolId, ()>> {
-        self.symbol_ids_for_scope(SymbolTable::root_scope_id())
-    }
-
-    pub(crate) fn root_symbols(&self) -> SymbolIterator<Copied<Keys<SymbolId, ()>>> {
-        self.symbols_for_scope(SymbolTable::root_scope_id())
-    }
-
-    pub(crate) fn child_scope_ids_of(&self, scope_id: ScopeId) -> &[ScopeId] {
-        &self.scopes_by_id[scope_id].child_scopes
-    }
-
-    pub(crate) fn child_scopes_of(&self, scope_id: ScopeId) -> ScopeIterator<&[ScopeId]> {
-        ScopeIterator {
-            table: self,
-            ids: self.child_scope_ids_of(scope_id),
-        }
-    }
-
-    pub(crate) fn root_child_scope_ids(&self) -> &[ScopeId] {
-        self.child_scope_ids_of(SymbolTable::root_scope_id())
-    }
-
-    pub(crate) fn root_child_scopes(&self) -> ScopeIterator<&[ScopeId]> {
-        self.child_scopes_of(SymbolTable::root_scope_id())
-    }
-
-    pub(crate) fn symbol_id_by_name(&self, scope_id: ScopeId, name: &str) -> Option<SymbolId> {
-        let scope = &self.scopes_by_id[scope_id];
-        let hash = SymbolTable::hash_name(name);
-        let name = Name::new(name);
-        scope
-            .symbols_by_name
-            .raw_entry()
-            .from_hash(hash, |symid| self.symbols_by_id[*symid].name == name)
-            .map(|(symbol_id, ())| *symbol_id)
-    }
-
-    pub(crate) fn symbol_by_name(&self, scope_id: ScopeId, name: &str) -> Option<&Symbol> {
-        Some(&self.symbols_by_id[self.symbol_id_by_name(scope_id, name)?])
-    }
-
-    pub(crate) fn root_symbol_id_by_name(&self, name: &str) -> Option<SymbolId> {
-        self.symbol_id_by_name(SymbolTable::root_scope_id(), name)
-    }
-
-    pub(crate) fn root_symbol_by_name(&self, name: &str) -> Option<&Symbol> {
-        self.symbol_by_name(SymbolTable::root_scope_id(), name)
-    }
-
-    pub(crate) fn defs(&self, symbol_id: SymbolId) -> &[Definition] {
-        self.defs
-            .get(&symbol_id)
-            .map(std::vec::Vec::as_slice)
-            .unwrap_or_default()
-    }
-
-    fn add_symbol_to_scope(&mut self, scope_id: ScopeId, name: &str) -> SymbolId {
-        let hash = SymbolTable::hash_name(name);
-        let scope = &mut self.scopes_by_id[scope_id];
-        let name = Name::new(name);
-
-        let entry = scope
-            .symbols_by_name
-            .raw_entry_mut()
-            .from_hash(hash, |existing| self.symbols_by_id[*existing].name == name);
-
-        match entry {
-            RawEntryMut::Occupied(entry) => *entry.key(),
-            RawEntryMut::Vacant(entry) => {
-                let id = self.symbols_by_id.push(Symbol { name });
-                entry.insert_with_hasher(hash, id, (), |_| hash);
-                id
-            }
-        }
-    }
-
-    fn add_child_scope(
-        &mut self,
-        parent_scope_id: ScopeId,
-        name: &str,
-        kind: ScopeKind,
-    ) -> ScopeId {
-        let new_scope_id = self.scopes_by_id.push(Scope {
-            name: Name::new(name),
-            kind,
-            child_scopes: Vec::new(),
-            symbols_by_name: Map::default(),
-        });
-        let parent_scope = &mut self.scopes_by_id[parent_scope_id];
-        parent_scope.child_scopes.push(new_scope_id);
-        new_scope_id
-    }
-
-    fn hash_name(name: &str) -> u64 {
-        let mut hasher = FxHasher::default();
-        name.hash(&mut hasher);
-        hasher.finish()
-    }
-}
-
-pub(crate) struct SymbolIterator<'a, I> {
-    table: &'a SymbolTable,
-    ids: I,
-}
-
-impl<'a, I> Iterator for SymbolIterator<'a, I>
-where
-    I: Iterator<Item = SymbolId>,
-{
-    type Item = &'a Symbol;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        let id = self.ids.next()?;
-        Some(&self.table.symbols_by_id[id])
-    }
-
-    fn size_hint(&self) -> (usize, Option<usize>) {
-        self.ids.size_hint()
-    }
-}
-
-impl<'a, I> FusedIterator for SymbolIterator<'a, I> where
-    I: Iterator<Item = SymbolId> + FusedIterator
-{
-}
-
-impl<'a, I> DoubleEndedIterator for SymbolIterator<'a, I>
-where
-    I: Iterator<Item = SymbolId> + DoubleEndedIterator,
-{
-    fn next_back(&mut self) -> Option<Self::Item> {
-        let id = self.ids.next_back()?;
-        Some(&self.table.symbols_by_id[id])
-    }
-}
-
-pub(crate) struct ScopeIterator<'a, I> {
-    table: &'a SymbolTable,
-    ids: I,
-}
-
-impl<'a, I> Iterator for ScopeIterator<'a, I>
-where
-    I: Iterator<Item = ScopeId>,
-{
-    type Item = &'a Scope;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        let id = self.ids.next()?;
-        Some(&self.table.scopes_by_id[id])
-    }
-
-    fn size_hint(&self) -> (usize, Option<usize>) {
-        self.ids.size_hint()
-    }
-}
-
-impl<'a, I> FusedIterator for ScopeIterator<'a, I> where I: Iterator<Item = ScopeId> + FusedIterator {}
-
-impl<'a, I> DoubleEndedIterator for ScopeIterator<'a, I>
-where
-    I: Iterator<Item = ScopeId> + DoubleEndedIterator,
-{
-    fn next_back(&mut self) -> Option<Self::Item> {
-        let id = self.ids.next_back()?;
-        Some(&self.table.scopes_by_id[id])
-    }
-}
-
-struct SymbolTableBuilder {
-    table: SymbolTable,
-    scopes: Vec<ScopeId>,
-}
-
-impl SymbolTableBuilder {
-    fn add_symbol(&mut self, identifier: &str) -> SymbolId {
-        self.table.add_symbol_to_scope(self.cur_scope(), identifier)
-    }
-
-    fn add_symbol_with_def(&mut self, identifier: &str, definition: Definition) -> SymbolId {
-        let symbol_id = self.add_symbol(identifier);
-        self.table
-            .defs
-            .entry(symbol_id)
-            .or_default()
-            .push(definition);
-        symbol_id
-    }
-
-    fn push_scope(&mut self, child_of: ScopeId, name: &str, kind: ScopeKind) -> ScopeId {
-        let scope_id = self.table.add_child_scope(child_of, name, kind);
-        self.scopes.push(scope_id);
-        scope_id
-    }
-
-    fn pop_scope(&mut self) -> ScopeId {
-        self.scopes
-            .pop()
-            .expect("Scope stack should never be empty")
-    }
-
-    fn cur_scope(&self) -> ScopeId {
-        *self
-            .scopes
-            .last()
-            .expect("Scope stack should never be empty")
-    }
-
-    fn with_type_params(
-        &mut self,
-        name: &str,
-        params: &Option<Box<ast::TypeParams>>,
-        nested: impl FnOnce(&mut Self),
-    ) {
-        if let Some(type_params) = params {
-            self.push_scope(self.cur_scope(), name, ScopeKind::Annotation);
-            for type_param in &type_params.type_params {
-                let name = match type_param {
-                    ast::TypeParam::TypeVar(ast::TypeParamTypeVar { name, .. }) => name,
-                    ast::TypeParam::ParamSpec(ast::TypeParamParamSpec { name, .. }) => name,
-                    ast::TypeParam::TypeVarTuple(ast::TypeParamTypeVarTuple { name, .. }) => name,
-                };
-                self.add_symbol(name);
-            }
-        }
-        nested(self);
-        if params.is_some() {
-            self.pop_scope();
-        }
-    }
-}
-
-impl PreorderVisitor<'_> for SymbolTableBuilder {
-    fn visit_expr(&mut self, expr: &ast::Expr) {
-        if let ast::Expr::Name(ast::ExprName { id, .. }) = expr {
-            self.add_symbol(id);
-        }
-        ast::visitor::preorder::walk_expr(self, expr);
-    }
-
-    fn visit_stmt(&mut self, stmt: &ast::Stmt) {
-        // TODO need to capture more definition statements here
-        match stmt {
-            ast::Stmt::ClassDef(node) => {
-                let def = Definition::ClassDef(TypedNodeKey::from_node(node));
-                self.add_symbol_with_def(&node.name, def);
-                self.with_type_params(&node.name, &node.type_params, |builder| {
-                    builder.push_scope(builder.cur_scope(), &node.name, ScopeKind::Class);
-                    ast::visitor::preorder::walk_stmt(builder, stmt);
-                    builder.pop_scope();
-                });
-            }
-            ast::Stmt::FunctionDef(node) => {
-                let def = Definition::FunctionDef(TypedNodeKey::from_node(node));
-                self.add_symbol_with_def(&node.name, def);
-                self.with_type_params(&node.name, &node.type_params, |builder| {
-                    builder.push_scope(builder.cur_scope(), &node.name, ScopeKind::Function);
-                    ast::visitor::preorder::walk_stmt(builder, stmt);
-                    builder.pop_scope();
-                });
-            }
-            ast::Stmt::Import(ast::StmtImport { names, .. }) => {
-                for alias in names {
-                    let symbol_name = if let Some(asname) = &alias.asname {
-                        asname.id.as_str()
-                    } else {
-                        alias.name.id.split('.').next().unwrap()
-                    };
-
-                    let module = Name::new(&alias.name.id);
-
-                    let def = Definition::Import(ImportDefinition {
-                        module: module.clone(),
-                    });
-                    self.add_symbol_with_def(symbol_name, def);
-                    self.table.dependencies.push(Dependency::Module(module));
-                }
-            }
-            ast::Stmt::ImportFrom(ast::StmtImportFrom {
-                module,
-                names,
-                level,
-                ..
-            }) => {
-                let module = module.as_ref().map(|m| Name::new(&m.id));
-
-                for alias in names {
-                    let symbol_name = if let Some(asname) = &alias.asname {
-                        asname.id.as_str()
-                    } else {
-                        alias.name.id.as_str()
-                    };
-                    let def = Definition::ImportFrom(ImportFromDefinition {
-                        module: module.clone(),
-                        name: Name::new(&alias.name.id),
-                        level: *level,
-                    });
-                    self.add_symbol_with_def(symbol_name, def);
-                }
-
-                let dependency = if let Some(module) = module {
-                    if *level == 0 {
-                        Dependency::Module(module)
-                    } else {
-                        Dependency::Relative {
-                            level: *level,
-                            module: Some(module),
-                        }
-                    }
-                } else {
-                    Dependency::Relative {
-                        level: *level,
-                        module,
-                    }
-                };
-
-                self.table.dependencies.push(dependency);
-            }
-            _ => {
-                ast::visitor::preorder::walk_stmt(self, stmt);
-            }
-        }
-    }
-}
-
-#[derive(Debug, Default)]
-pub struct SymbolTablesStorage(KeyValueCache<FileId, Arc<SymbolTable>>);
-
-impl Deref for SymbolTablesStorage {
-    type Target = KeyValueCache<FileId, Arc<SymbolTable>>;
-
-    fn deref(&self) -> &Self::Target {
-        &self.0
-    }
-}
-
-impl DerefMut for SymbolTablesStorage {
-    fn deref_mut(&mut self) -> &mut Self::Target {
-        &mut self.0
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use textwrap::dedent;
-
-    use crate::parse::Parsed;
-    use crate::symbols::ScopeKind;
-
-    use super::{SymbolId, SymbolIterator, SymbolTable};
-
-    mod from_ast {
-        use super::*;
-
-        fn parse(code: &str) -> Parsed {
-            Parsed::from_text(&dedent(code))
-        }
-
-        fn names<I>(it: SymbolIterator<I>) -> Vec<&str>
-        where
-            I: Iterator<Item = SymbolId>,
-        {
-            let mut symbols: Vec<_> = it.map(|sym| sym.name.as_str()).collect();
-            symbols.sort_unstable();
-            symbols
-        }
-
-        #[test]
-        fn empty() {
-            let parsed = parse("");
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()).len(), 0);
-        }
-
-        #[test]
-        fn simple() {
-            let parsed = parse("x");
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["x"]);
-            assert_eq!(
-                table.defs(table.root_symbol_id_by_name("x").unwrap()).len(),
-                0
-            );
-        }
-
-        #[test]
-        fn annotation_only() {
-            let parsed = parse("x: int");
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["int", "x"]);
-            // TODO record definition
-        }
-
-        #[test]
-        fn import() {
-            let parsed = parse("import foo");
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["foo"]);
-            assert_eq!(
-                table
-                    .defs(table.root_symbol_id_by_name("foo").unwrap())
-                    .len(),
-                1
-            );
-        }
-
-        #[test]
-        fn import_sub() {
-            let parsed = parse("import foo.bar");
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["foo"]);
-        }
-
-        #[test]
-        fn import_as() {
-            let parsed = parse("import foo.bar as baz");
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["baz"]);
-        }
-
-        #[test]
-        fn import_from() {
-            let parsed = parse("from bar import foo");
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["foo"]);
-            assert_eq!(
-                table
-                    .defs(table.root_symbol_id_by_name("foo").unwrap())
-                    .len(),
-                1
-            );
-        }
-
-        #[test]
-        fn class_scope() {
-            let parsed = parse(
-                "
-                class C:
-                    x = 1
-                y = 2
-                ",
-            );
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["C", "y"]);
-            let scopes = table.root_child_scope_ids();
-            assert_eq!(scopes.len(), 1);
-            let c_scope = scopes[0].scope(&table);
-            assert_eq!(c_scope.kind(), ScopeKind::Class);
-            assert_eq!(c_scope.name(), "C");
-            assert_eq!(names(table.symbols_for_scope(scopes[0])), vec!["x"]);
-            assert_eq!(
-                table.defs(table.root_symbol_id_by_name("C").unwrap()).len(),
-                1
-            );
-        }
-
-        #[test]
-        fn func_scope() {
-            let parsed = parse(
-                "
-                def func():
-                    x = 1
-                y = 2
-                ",
-            );
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["func", "y"]);
-            let scopes = table.root_child_scope_ids();
-            assert_eq!(scopes.len(), 1);
-            let func_scope = scopes[0].scope(&table);
-            assert_eq!(func_scope.kind(), ScopeKind::Function);
-            assert_eq!(func_scope.name(), "func");
-            assert_eq!(names(table.symbols_for_scope(scopes[0])), vec!["x"]);
-            assert_eq!(
-                table
-                    .defs(table.root_symbol_id_by_name("func").unwrap())
-                    .len(),
-                1
-            );
-        }
-
-        #[test]
-        fn dupes() {
-            let parsed = parse(
-                "
-                def func():
-                    x = 1
-                def func():
-                    y = 2
-                ",
-            );
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["func"]);
-            let scopes = table.root_child_scope_ids();
-            assert_eq!(scopes.len(), 2);
-            let func_scope_1 = scopes[0].scope(&table);
-            let func_scope_2 = scopes[1].scope(&table);
-            assert_eq!(func_scope_1.kind(), ScopeKind::Function);
-            assert_eq!(func_scope_1.name(), "func");
-            assert_eq!(func_scope_2.kind(), ScopeKind::Function);
-            assert_eq!(func_scope_2.name(), "func");
-            assert_eq!(names(table.symbols_for_scope(scopes[0])), vec!["x"]);
-            assert_eq!(names(table.symbols_for_scope(scopes[1])), vec!["y"]);
-            assert_eq!(
-                table
-                    .defs(table.root_symbol_id_by_name("func").unwrap())
-                    .len(),
-                2
-            );
-        }
-
-        #[test]
-        fn generic_func() {
-            let parsed = parse(
-                "
-                def func[T]():
-                    x = 1
-                ",
-            );
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["func"]);
-            let scopes = table.root_child_scope_ids();
-            assert_eq!(scopes.len(), 1);
-            let ann_scope_id = scopes[0];
-            let ann_scope = ann_scope_id.scope(&table);
-            assert_eq!(ann_scope.kind(), ScopeKind::Annotation);
-            assert_eq!(ann_scope.name(), "func");
-            assert_eq!(names(table.symbols_for_scope(ann_scope_id)), vec!["T"]);
-            let scopes = table.child_scope_ids_of(ann_scope_id);
-            assert_eq!(scopes.len(), 1);
-            let func_scope_id = scopes[0];
-            let func_scope = func_scope_id.scope(&table);
-            assert_eq!(func_scope.kind(), ScopeKind::Function);
-            assert_eq!(func_scope.name(), "func");
-            assert_eq!(names(table.symbols_for_scope(func_scope_id)), vec!["x"]);
-        }
-
-        #[test]
-        fn generic_class() {
-            let parsed = parse(
-                "
-                class C[T]:
-                    x = 1
-                ",
-            );
-            let table = SymbolTable::from_ast(parsed.ast());
-            assert_eq!(names(table.root_symbols()), vec!["C"]);
-            let scopes = table.root_child_scope_ids();
-            assert_eq!(scopes.len(), 1);
-            let ann_scope_id = scopes[0];
-            let ann_scope = ann_scope_id.scope(&table);
-            assert_eq!(ann_scope.kind(), ScopeKind::Annotation);
-            assert_eq!(ann_scope.name(), "C");
-            assert_eq!(names(table.symbols_for_scope(ann_scope_id)), vec!["T"]);
-            let scopes = table.child_scope_ids_of(ann_scope_id);
-            assert_eq!(scopes.len(), 1);
-            let func_scope_id = scopes[0];
-            let func_scope = func_scope_id.scope(&table);
-            assert_eq!(func_scope.kind(), ScopeKind::Class);
-            assert_eq!(func_scope.name(), "C");
-            assert_eq!(names(table.symbols_for_scope(func_scope_id)), vec!["x"]);
-        }
-    }
-
-    #[test]
-    fn insert_same_name_symbol_twice() {
-        let mut table = SymbolTable::new();
-        let root_scope_id = SymbolTable::root_scope_id();
-        let symbol_id_1 = table.add_symbol_to_scope(root_scope_id, "foo");
-        let symbol_id_2 = table.add_symbol_to_scope(root_scope_id, "foo");
-        assert_eq!(symbol_id_1, symbol_id_2);
-    }
-
-    #[test]
-    fn insert_different_named_symbols() {
-        let mut table = SymbolTable::new();
-        let root_scope_id = SymbolTable::root_scope_id();
-        let symbol_id_1 = table.add_symbol_to_scope(root_scope_id, "foo");
-        let symbol_id_2 = table.add_symbol_to_scope(root_scope_id, "bar");
-        assert_ne!(symbol_id_1, symbol_id_2);
-    }
-
-    #[test]
-    fn add_child_scope_with_symbol() {
-        let mut table = SymbolTable::new();
-        let root_scope_id = SymbolTable::root_scope_id();
-        let foo_symbol_top = table.add_symbol_to_scope(root_scope_id, "foo");
-        let c_scope = table.add_child_scope(root_scope_id, "C", ScopeKind::Class);
-        let foo_symbol_inner = table.add_symbol_to_scope(c_scope, "foo");
-        assert_ne!(foo_symbol_top, foo_symbol_inner);
-    }
-
-    #[test]
-    fn scope_from_id() {
-        let table = SymbolTable::new();
-        let root_scope_id = SymbolTable::root_scope_id();
-        let scope = root_scope_id.scope(&table);
-        assert_eq!(scope.name.as_str(), "<module>");
-        assert_eq!(scope.kind, ScopeKind::Module);
-    }
-
-    #[test]
-    fn symbol_from_id() {
-        let mut table = SymbolTable::new();
-        let root_scope_id = SymbolTable::root_scope_id();
-        let foo_symbol_id = table.add_symbol_to_scope(root_scope_id, "foo");
-        let symbol = foo_symbol_id.symbol(&table);
-        assert_eq!(symbol.name.as_str(), "foo");
-    }
-}

crates/red_knot/src/types.rs

@@ -1,527 +0,0 @@
-#![allow(dead_code)]
-use crate::ast_ids::NodeKey;
-use crate::files::FileId;
-use crate::symbols::SymbolId;
-use crate::{FxDashMap, FxIndexSet, Name};
-use ruff_index::{newtype_index, IndexVec};
-use rustc_hash::FxHashMap;
-
-pub(crate) mod infer;
-
-pub(crate) use infer::infer_symbol_type;
-
-/// unique ID for a type
-#[derive(Copy, Clone, Debug, PartialEq, Eq, Hash)]
-pub enum Type {
-    /// the dynamic or gradual type: a statically-unknown set of values
-    Any,
-    /// the empty set of values
-    Never,
-    /// unknown type (no annotation)
-    /// equivalent to Any, or to object in strict mode
-    Unknown,
-    /// name is not bound to any value
-    Unbound,
-    /// a specific function object
-    Function(FunctionTypeId),
-    /// a specific class object
-    Class(ClassTypeId),
-    /// the set of Python objects with the given class in their __class__'s method resolution order
-    Instance(ClassTypeId),
-    Union(UnionTypeId),
-    Intersection(IntersectionTypeId),
-    // TODO protocols, callable types, overloads, generics, type vars
-}
-
-impl Type {
-    fn display<'a>(&'a self, store: &'a TypeStore) -> DisplayType<'a> {
-        DisplayType { ty: self, store }
-    }
-}
-
-// TODO: currently calling `get_function` et al and holding on to the `FunctionTypeRef` will lock a
-// shard of this dashmap, for as long as you hold the reference. This may be a problem. We could
-// switch to having all the arenas hold Arc, or we could see if we can split up ModuleTypeStore,
-// and/or give it inner mutability and finer-grained internal locking.
-#[derive(Debug, Default)]
-pub struct TypeStore {
-    modules: FxDashMap<FileId, ModuleTypeStore>,
-}
-
-impl TypeStore {
-    pub fn remove_module(&mut self, file_id: FileId) {
-        self.modules.remove(&file_id);
-    }
-
-    pub fn cache_symbol_type(&mut self, file_id: FileId, symbol_id: SymbolId, ty: Type) {
-        self.add_or_get_module(file_id)
-            .symbol_types
-            .insert(symbol_id, ty);
-    }
-
-    pub fn cache_node_type(&mut self, file_id: FileId, node_key: NodeKey, ty: Type) {
-        self.add_or_get_module(file_id)
-            .node_types
-            .insert(node_key, ty);
-    }
-
-    pub fn get_cached_symbol_type(&self, file_id: FileId, symbol_id: SymbolId) -> Option<Type> {
-        self.try_get_module(file_id)?
-            .symbol_types
-            .get(&symbol_id)
-            .copied()
-    }
-
-    pub fn get_cached_node_type(&self, file_id: FileId, node_key: &NodeKey) -> Option<Type> {
-        self.try_get_module(file_id)?
-            .node_types
-            .get(node_key)
-            .copied()
-    }
-
-    fn add_or_get_module(&mut self, file_id: FileId) -> ModuleStoreRefMut {
-        self.modules
-            .entry(file_id)
-            .or_insert_with(|| ModuleTypeStore::new(file_id))
-    }
-
-    fn get_module(&self, file_id: FileId) -> ModuleStoreRef {
-        self.try_get_module(file_id).expect("module should exist")
-    }
-
-    fn try_get_module(&self, file_id: FileId) -> Option<ModuleStoreRef> {
-        self.modules.get(&file_id)
-    }
-
-    fn add_function(&mut self, file_id: FileId, name: &str) -> FunctionTypeId {
-        self.add_or_get_module(file_id).add_function(name)
-    }
-
-    fn add_class(&mut self, file_id: FileId, name: &str) -> ClassTypeId {
-        self.add_or_get_module(file_id).add_class(name)
-    }
-
-    fn add_union(&mut self, file_id: FileId, elems: &[Type]) -> UnionTypeId {
-        self.add_or_get_module(file_id).add_union(elems)
-    }
-
-    fn add_intersection(
-        &mut self,
-        file_id: FileId,
-        positive: &[Type],
-        negative: &[Type],
-    ) -> IntersectionTypeId {
-        self.add_or_get_module(file_id)
-            .add_intersection(positive, negative)
-    }
-
-    fn get_function(&self, id: FunctionTypeId) -> FunctionTypeRef {
-        FunctionTypeRef {
-            module_store: self.get_module(id.file_id),
-            function_id: id.func_id,
-        }
-    }
-
-    fn get_class(&self, id: ClassTypeId) -> ClassTypeRef {
-        ClassTypeRef {
-            module_store: self.get_module(id.file_id),
-            class_id: id.class_id,
-        }
-    }
-
-    fn get_union(&self, id: UnionTypeId) -> UnionTypeRef {
-        UnionTypeRef {
-            module_store: self.get_module(id.file_id),
-            union_id: id.union_id,
-        }
-    }
-
-    fn get_intersection(&self, id: IntersectionTypeId) -> IntersectionTypeRef {
-        IntersectionTypeRef {
-            module_store: self.get_module(id.file_id),
-            intersection_id: id.intersection_id,
-        }
-    }
-}
-
-type ModuleStoreRef<'a> = dashmap::mapref::one::Ref<
-    'a,
-    FileId,
-    ModuleTypeStore,
-    std::hash::BuildHasherDefault<rustc_hash::FxHasher>,
->;
-
-type ModuleStoreRefMut<'a> = dashmap::mapref::one::RefMut<
-    'a,
-    FileId,
-    ModuleTypeStore,
-    std::hash::BuildHasherDefault<rustc_hash::FxHasher>,
->;
-
-#[derive(Debug)]
-pub(crate) struct FunctionTypeRef<'a> {
-    module_store: ModuleStoreRef<'a>,
-    function_id: ModuleFunctionTypeId,
-}
-
-impl<'a> std::ops::Deref for FunctionTypeRef<'a> {
-    type Target = FunctionType;
-
-    fn deref(&self) -> &Self::Target {
-        self.module_store.get_function(self.function_id)
-    }
-}
-
-#[derive(Debug)]
-pub(crate) struct ClassTypeRef<'a> {
-    module_store: ModuleStoreRef<'a>,
-    class_id: ModuleClassTypeId,
-}
-
-impl<'a> std::ops::Deref for ClassTypeRef<'a> {
-    type Target = ClassType;
-
-    fn deref(&self) -> &Self::Target {
-        self.module_store.get_class(self.class_id)
-    }
-}
-
-#[derive(Debug)]
-pub(crate) struct UnionTypeRef<'a> {
-    module_store: ModuleStoreRef<'a>,
-    union_id: ModuleUnionTypeId,
-}
-
-impl<'a> std::ops::Deref for UnionTypeRef<'a> {
-    type Target = UnionType;
-
-    fn deref(&self) -> &Self::Target {
-        self.module_store.get_union(self.union_id)
-    }
-}
-
-#[derive(Debug)]
-pub(crate) struct IntersectionTypeRef<'a> {
-    module_store: ModuleStoreRef<'a>,
-    intersection_id: ModuleIntersectionTypeId,
-}
-
-impl<'a> std::ops::Deref for IntersectionTypeRef<'a> {
-    type Target = IntersectionType;
-
-    fn deref(&self) -> &Self::Target {
-        self.module_store.get_intersection(self.intersection_id)
-    }
-}
-
-#[derive(Copy, Clone, Debug, Hash, Eq, PartialEq)]
-pub struct FunctionTypeId {
-    file_id: FileId,
-    func_id: ModuleFunctionTypeId,
-}
-
-#[derive(Copy, Clone, Debug, Hash, Eq, PartialEq)]
-pub struct ClassTypeId {
-    file_id: FileId,
-    class_id: ModuleClassTypeId,
-}
-
-#[derive(Copy, Clone, Debug, Hash, Eq, PartialEq)]
-pub struct UnionTypeId {
-    file_id: FileId,
-    union_id: ModuleUnionTypeId,
-}
-
-#[derive(Copy, Clone, Debug, Hash, Eq, PartialEq)]
-pub struct IntersectionTypeId {
-    file_id: FileId,
-    intersection_id: ModuleIntersectionTypeId,
-}
-
-#[newtype_index]
-struct ModuleFunctionTypeId;
-
-#[newtype_index]
-struct ModuleClassTypeId;
-
-#[newtype_index]
-struct ModuleUnionTypeId;
-
-#[newtype_index]
-struct ModuleIntersectionTypeId;
-
-#[derive(Debug)]
-struct ModuleTypeStore {
-    file_id: FileId,
-    /// arena of all function types defined in this module
-    functions: IndexVec<ModuleFunctionTypeId, FunctionType>,
-    /// arena of all class types defined in this module
-    classes: IndexVec<ModuleClassTypeId, ClassType>,
-    /// arenda of all union types created in this module
-    unions: IndexVec<ModuleUnionTypeId, UnionType>,
-    /// arena of all intersection types created in this module
-    intersections: IndexVec<ModuleIntersectionTypeId, IntersectionType>,
-    /// cached types of symbols in this module
-    symbol_types: FxHashMap<SymbolId, Type>,
-    /// cached types of AST nodes in this module
-    node_types: FxHashMap<NodeKey, Type>,
-}
-
-impl ModuleTypeStore {
-    fn new(file_id: FileId) -> Self {
-        Self {
-            file_id,
-            functions: IndexVec::default(),
-            classes: IndexVec::default(),
-            unions: IndexVec::default(),
-            intersections: IndexVec::default(),
-            symbol_types: FxHashMap::default(),
-            node_types: FxHashMap::default(),
-        }
-    }
-
-    fn add_function(&mut self, name: &str) -> FunctionTypeId {
-        let func_id = self.functions.push(FunctionType {
-            name: Name::new(name),
-        });
-        FunctionTypeId {
-            file_id: self.file_id,
-            func_id,
-        }
-    }
-
-    fn add_class(&mut self, name: &str) -> ClassTypeId {
-        let class_id = self.classes.push(ClassType {
-            name: Name::new(name),
-        });
-        ClassTypeId {
-            file_id: self.file_id,
-            class_id,
-        }
-    }
-
-    fn add_union(&mut self, elems: &[Type]) -> UnionTypeId {
-        let union_id = self.unions.push(UnionType {
-            elements: elems.iter().copied().collect(),
-        });
-        UnionTypeId {
-            file_id: self.file_id,
-            union_id,
-        }
-    }
-
-    fn add_intersection(&mut self, positive: &[Type], negative: &[Type]) -> IntersectionTypeId {
-        let intersection_id = self.intersections.push(IntersectionType {
-            positive: positive.iter().copied().collect(),
-            negative: negative.iter().copied().collect(),
-        });
-        IntersectionTypeId {
-            file_id: self.file_id,
-            intersection_id,
-        }
-    }
-
-    fn get_function(&self, func_id: ModuleFunctionTypeId) -> &FunctionType {
-        &self.functions[func_id]
-    }
-
-    fn get_class(&self, class_id: ModuleClassTypeId) -> &ClassType {
-        &self.classes[class_id]
-    }
-
-    fn get_union(&self, union_id: ModuleUnionTypeId) -> &UnionType {
-        &self.unions[union_id]
-    }
-
-    fn get_intersection(&self, intersection_id: ModuleIntersectionTypeId) -> &IntersectionType {
-        &self.intersections[intersection_id]
-    }
-}
-
-#[derive(Copy, Clone, Debug)]
-struct DisplayType<'a> {
-    ty: &'a Type,
-    store: &'a TypeStore,
-}
-
-impl std::fmt::Display for DisplayType<'_> {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        match self.ty {
-            Type::Any => f.write_str("Any"),
-            Type::Never => f.write_str("Never"),
-            Type::Unknown => f.write_str("Unknown"),
-            Type::Unbound => f.write_str("Unbound"),
-            // TODO functions and classes should display using a fully qualified name
-            Type::Class(class_id) => {
-                f.write_str("Literal[")?;
-                f.write_str(self.store.get_class(*class_id).name())?;
-                f.write_str("]")
-            }
-            Type::Instance(class_id) => f.write_str(self.store.get_class(*class_id).name()),
-            Type::Function(func_id) => f.write_str(self.store.get_function(*func_id).name()),
-            Type::Union(union_id) => self
-                .store
-                .get_module(union_id.file_id)
-                .get_union(union_id.union_id)
-                .display(f, self.store),
-            Type::Intersection(int_id) => self
-                .store
-                .get_module(int_id.file_id)
-                .get_intersection(int_id.intersection_id)
-                .display(f, self.store),
-        }
-    }
-}
-
-#[derive(Debug)]
-pub(crate) struct ClassType {
-    name: Name,
-}
-
-impl ClassType {
-    fn name(&self) -> &str {
-        self.name.as_str()
-    }
-}
-
-#[derive(Debug)]
-pub(crate) struct FunctionType {
-    name: Name,
-}
-
-impl FunctionType {
-    fn name(&self) -> &str {
-        self.name.as_str()
-    }
-}
-
-#[derive(Debug)]
-pub(crate) struct UnionType {
-    // the union type includes values in any of these types
-    elements: FxIndexSet<Type>,
-}
-
-impl UnionType {
-    fn display(&self, f: &mut std::fmt::Formatter<'_>, store: &TypeStore) -> std::fmt::Result {
-        f.write_str("(")?;
-        let mut first = true;
-        for ty in &self.elements {
-            if !first {
-                f.write_str(" | ")?;
-            };
-            first = false;
-            write!(f, "{}", ty.display(store))?;
-        }
-        f.write_str(")")
-    }
-}
-
-// Negation types aren't expressible in annotations, and are most likely to arise from type
-// narrowing along with intersections (e.g. `if not isinstance(...)`), so we represent them
-// directly in intersections rather than as a separate type. This sacrifices some efficiency in the
-// case where a Not appears outside an intersection (unclear when that could even happen, but we'd
-// have to represent it as a single-element intersection if it did) in exchange for better
-// efficiency in the not-within-intersection case.
-#[derive(Debug)]
-pub(crate) struct IntersectionType {
-    // the intersection type includes only values in all of these types
-    positive: FxIndexSet<Type>,
-    // negated elements of the intersection, e.g.
-    negative: FxIndexSet<Type>,
-}
-
-impl IntersectionType {
-    fn display(&self, f: &mut std::fmt::Formatter<'_>, store: &TypeStore) -> std::fmt::Result {
-        f.write_str("(")?;
-        let mut first = true;
-        for (neg, ty) in self
-            .positive
-            .iter()
-            .map(|ty| (false, ty))
-            .chain(self.negative.iter().map(|ty| (true, ty)))
-        {
-            if !first {
-                f.write_str(" & ")?;
-            };
-            first = false;
-            if neg {
-                f.write_str("~")?;
-            };
-            write!(f, "{}", ty.display(store))?;
-        }
-        f.write_str(")")
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use crate::files::Files;
-    use crate::types::{Type, TypeStore};
-    use crate::FxIndexSet;
-    use std::path::Path;
-
-    #[test]
-    fn add_class() {
-        let mut store = TypeStore::default();
-        let files = Files::default();
-        let file_id = files.intern(Path::new("/foo"));
-        let id = store.add_class(file_id, "C");
-        assert_eq!(store.get_class(id).name(), "C");
-        let inst = Type::Instance(id);
-        assert_eq!(format!("{}", inst.display(&store)), "C");
-    }
-
-    #[test]
-    fn add_function() {
-        let mut store = TypeStore::default();
-        let files = Files::default();
-        let file_id = files.intern(Path::new("/foo"));
-        let id = store.add_function(file_id, "func");
-        assert_eq!(store.get_function(id).name(), "func");
-        let func = Type::Function(id);
-        assert_eq!(format!("{}", func.display(&store)), "func");
-    }
-
-    #[test]
-    fn add_union() {
-        let mut store = TypeStore::default();
-        let files = Files::default();
-        let file_id = files.intern(Path::new("/foo"));
-        let c1 = store.add_class(file_id, "C1");
-        let c2 = store.add_class(file_id, "C2");
-        let elems = vec![Type::Instance(c1), Type::Instance(c2)];
-        let id = store.add_union(file_id, &elems);
-        assert_eq!(
-            store.get_union(id).elements,
-            elems.into_iter().collect::<FxIndexSet<_>>()
-        );
-        let union = Type::Union(id);
-        assert_eq!(format!("{}", union.display(&store)), "(C1 | C2)");
-    }
-
-    #[test]
-    fn add_intersection() {
-        let mut store = TypeStore::default();
-        let files = Files::default();
-        let file_id = files.intern(Path::new("/foo"));
-        let c1 = store.add_class(file_id, "C1");
-        let c2 = store.add_class(file_id, "C2");
-        let c3 = store.add_class(file_id, "C3");
-        let pos = vec![Type::Instance(c1), Type::Instance(c2)];
-        let neg = vec![Type::Instance(c3)];
-        let id = store.add_intersection(file_id, &pos, &neg);
-        assert_eq!(
-            store.get_intersection(id).positive,
-            pos.into_iter().collect::<FxIndexSet<_>>()
-        );
-        assert_eq!(
-            store.get_intersection(id).negative,
-            neg.into_iter().collect::<FxIndexSet<_>>()
-        );
-        let intersection = Type::Intersection(id);
-        assert_eq!(
-            format!("{}", intersection.display(&store)),
-            "(C1 & C2 & ~C3)"
-        );
-    }
-}

crates/red_knot/src/types/infer.rs

@@ -1,141 +0,0 @@
-#![allow(dead_code)]
-use crate::db::{HasJar, SemanticDb, SemanticJar};
-use crate::module::ModuleName;
-use crate::symbols::{Definition, ImportFromDefinition, SymbolId};
-use crate::types::Type;
-use crate::FileId;
-use ruff_python_ast::AstNode;
-
-// TODO this should not take a &mut db, it should be a query, not a mutation. This means we'll need
-// to use interior mutability in TypeStore instead, and avoid races in populating the cache.
-#[tracing::instrument(level = "trace", skip(db))]
-pub fn infer_symbol_type<Db>(db: &mut Db, file_id: FileId, symbol_id: SymbolId) -> Type
-where
-    Db: SemanticDb + HasJar<SemanticJar>,
-{
-    let symbols = db.symbol_table(file_id);
-    let defs = symbols.defs(symbol_id);
-
-    if let Some(ty) = db
-        .jar()
-        .type_store
-        .get_cached_symbol_type(file_id, symbol_id)
-    {
-        return ty;
-    }
-
-    // TODO handle multiple defs, conditional defs...
-    assert_eq!(defs.len(), 1);
-
-    let ty = match &defs[0] {
-        Definition::ImportFrom(ImportFromDefinition {
-            module,
-            name,
-            level,
-        }) => {
-            // TODO relative imports
-            assert!(matches!(level, 0));
-            let module_name = ModuleName::new(module.as_ref().expect("TODO relative imports"));
-            if let Some(module) = db.resolve_module(module_name) {
-                let remote_file_id = module.path(db).file();
-                let remote_symbols = db.symbol_table(remote_file_id);
-                if let Some(remote_symbol_id) = remote_symbols.root_symbol_id_by_name(name) {
-                    db.infer_symbol_type(remote_file_id, remote_symbol_id)
-                } else {
-                    Type::Unknown
-                }
-            } else {
-                Type::Unknown
-            }
-        }
-        Definition::ClassDef(node_key) => {
-            if let Some(ty) = db
-                .jar()
-                .type_store
-                .get_cached_node_type(file_id, node_key.erased())
-            {
-                ty
-            } else {
-                let parsed = db.parse(file_id);
-                let ast = parsed.ast();
-                let node = node_key.resolve_unwrap(ast.as_any_node_ref());
-
-                let store = &mut db.jar_mut().type_store;
-                let ty = Type::Class(store.add_class(file_id, &node.name.id));
-                store.cache_node_type(file_id, *node_key.erased(), ty);
-                ty
-            }
-        }
-        _ => todo!("other kinds of definitions"),
-    };
-
-    db.jar_mut()
-        .type_store
-        .cache_symbol_type(file_id, symbol_id, ty);
-    // TODO record dependencies
-    ty
-}
-
-#[cfg(test)]
-mod tests {
-    use crate::db::tests::TestDb;
-    use crate::db::{HasJar, SemanticDb, SemanticJar};
-    use crate::module::{ModuleName, ModuleSearchPath, ModuleSearchPathKind};
-    use crate::types::Type;
-
-    // TODO with virtual filesystem we shouldn't have to write files to disk for these
-    // tests
-
-    struct TestCase {
-        temp_dir: tempfile::TempDir,
-        db: TestDb,
-
-        src: ModuleSearchPath,
-    }
-
-    fn create_test() -> std::io::Result<TestCase> {
-        let temp_dir = tempfile::tempdir()?;
-
-        let src = temp_dir.path().join("src");
-        std::fs::create_dir(&src)?;
-        let src = ModuleSearchPath::new(src.canonicalize()?, ModuleSearchPathKind::FirstParty);
-
-        let roots = vec![src.clone()];
-
-        let mut db = TestDb::default();
-        db.set_module_search_paths(roots);
-
-        Ok(TestCase { temp_dir, db, src })
-    }
-
-    #[test]
-    fn follow_import_to_class() -> std::io::Result<()> {
-        let TestCase {
-            src,
-            mut db,
-            temp_dir: _temp_dir,
-        } = create_test()?;
-
-        let a_path = src.path().join("a.py");
-        let b_path = src.path().join("b.py");
-        std::fs::write(a_path, "from b import C as D")?;
-        std::fs::write(b_path, "class C: pass")?;
-        let a_file = db
-            .resolve_module(ModuleName::new("a"))
-            .expect("module should be found")
-            .path(&db)
-            .file();
-        let a_syms = db.symbol_table(a_file);
-        let d_sym = a_syms
-            .root_symbol_id_by_name("D")
-            .expect("D symbol should be found");
-
-        let ty = db.infer_symbol_type(a_file, d_sym);
-
-        let jar = HasJar::<SemanticJar>::jar(&db);
-
-        assert!(matches!(ty, Type::Class(_)));
-        assert_eq!(format!("{}", ty.display(&jar.type_store)), "Literal[C]");
-        Ok(())
-    }
-}

crates/red_knot/src/watch.rs

@@ -1,38 +1,38 @@
-use anyhow::Context;
 use std::path::Path;
 
-use crate::files::Files;
-use crate::program::{FileChange, FileChangeKind};
+use anyhow::Context;
 use notify::event::{CreateKind, RemoveKind};
 use notify::{recommended_watcher, Event, EventKind, RecommendedWatcher, RecursiveMode, Watcher};
 
+use crate::program::{FileChangeKind, FileWatcherChange};
+
 pub struct FileWatcher {
     watcher: RecommendedWatcher,
 }
 
 pub trait EventHandler: Send + 'static {
-    fn handle(&self, changes: Vec<FileChange>);
+    fn handle(&self, changes: Vec<FileWatcherChange>);
 }
 
 impl<F> EventHandler for F
 where
-    F: Fn(Vec<FileChange>) + Send + 'static,
+    F: Fn(Vec<FileWatcherChange>) + Send + 'static,
 {
-    fn handle(&self, changes: Vec<FileChange>) {
+    fn handle(&self, changes: Vec<FileWatcherChange>) {
         let f = self;
         f(changes);
     }
 }
 
 impl FileWatcher {
-    pub fn new<E>(handler: E, files: Files) -> anyhow::Result<Self>
+    pub fn new<E>(handler: E) -> anyhow::Result<Self>
     where
         E: EventHandler,
     {
-        Self::from_handler(Box::new(handler), files)
+        Self::from_handler(Box::new(handler))
     }
 
-    fn from_handler(handler: Box<dyn EventHandler>, files: Files) -> anyhow::Result<Self> {
+    fn from_handler(handler: Box<dyn EventHandler>) -> anyhow::Result<Self> {
         let watcher = recommended_watcher(move |changes: notify::Result<Event>| {
             match changes {
                 Ok(event) => {
@@ -50,8 +50,7 @@ impl FileWatcher {
 
                     for path in event.paths {
                         if path.is_file() {
-                            let id = files.intern(&path);
-                            changes.push(FileChange::new(id, change_kind));
+                            changes.push(FileWatcherChange::new(path, change_kind));
                         }
                     }
 

crates/red_knot_module_resolver/Cargo.toml

@@ -0,0 +1,35 @@
+[package]
+name = "red_knot_module_resolver"
+version = "0.0.0"
+publish = false
+authors = { workspace = true }
+edition = { workspace = true }
+rust-version = { workspace = true }
+homepage = { workspace = true }
+documentation = { workspace = true }
+repository = { workspace = true }
+license = { workspace = true }
+
+[dependencies]
+ruff_db = { workspace = true }
+ruff_python_stdlib = { workspace = true }
+
+rustc-hash = { workspace = true }
+salsa = { workspace = true }
+smol_str = { workspace = true }
+tracing = { workspace = true }
+zip = { workspace = true }
+
+[build-dependencies]
+path-slash = { workspace = true }
+walkdir = { workspace = true }
+zip = { workspace = true }
+
+[dev-dependencies]
+anyhow = { workspace = true }
+insta = { workspace = true }
+tempfile = { workspace = true }
+walkdir = { workspace = true }
+
+[lints]
+workspace = true

crates/red_knot_module_resolver/README.md

@@ -0,0 +1,9 @@
+# Red Knot
+
+A work-in-progress multifile module resolver for Ruff.
+
+## Vendored types for the stdlib
+
+This crate vendors [typeshed](https://github.com/python/typeshed)'s stubs for the standard library. The vendored stubs can be found in `crates/red_knot_module_resolver/vendor/typeshed`. The file `crates/red_knot_module_resolver/vendor/typeshed/source_commit.txt` tells you the typeshed commit that our vendored stdlib stubs currently correspond to.
+
+The typeshed stubs are updated every two weeks via an automated PR using the `sync_typeshed.yaml` workflow in the `.github/workflows` directory. This workflow can also be triggered at any time via [workflow dispatch](https://docs.github.com/en/actions/using-workflows/manually-running-a-workflow#running-a-workflow).

crates/red_knot_module_resolver/build.rs

@@ -0,0 +1,74 @@
+//! Build script to package our vendored typeshed files
+//! into a zip archive that can be included in the Ruff binary.
+//!
+//! This script should be automatically run at build time
+//! whenever the script itself changes, or whenever any files
+//! in `crates/red_knot_module_resolver/vendor/typeshed` change.
+
+use std::fs::File;
+use std::path::Path;
+
+use path_slash::PathExt;
+use zip::result::ZipResult;
+use zip::write::{FileOptions, ZipWriter};
+use zip::CompressionMethod;
+
+const TYPESHED_SOURCE_DIR: &str = "vendor/typeshed";
+const TYPESHED_ZIP_LOCATION: &str = "/zipped_typeshed.zip";
+
+/// Recursively zip the contents of an entire directory.
+///
+/// This routine is adapted from a recipe at
+/// <https://github.com/zip-rs/zip-old/blob/5d0f198124946b7be4e5969719a7f29f363118cd/examples/write_dir.rs>
+fn zip_dir(directory_path: &str, writer: File) -> ZipResult<File> {
+    let mut zip = ZipWriter::new(writer);
+
+    let options = FileOptions::default()
+        .compression_method(CompressionMethod::Zstd)
+        .unix_permissions(0o644);
+
+    for entry in walkdir::WalkDir::new(directory_path) {
+        let dir_entry = entry.unwrap();
+        let absolute_path = dir_entry.path();
+        let normalized_relative_path = absolute_path
+            .strip_prefix(Path::new(directory_path))
+            .unwrap()
+            .to_slash()
+            .expect("Unexpected non-utf8 typeshed path!");
+
+        // Write file or directory explicitly
+        // Some unzip tools unzip files with directory paths correctly, some do not!
+        if absolute_path.is_file() {
+            println!("adding file {absolute_path:?} as {normalized_relative_path:?} ...");
+            zip.start_file(normalized_relative_path, options)?;
+            let mut f = File::open(absolute_path)?;
+            std::io::copy(&mut f, &mut zip).unwrap();
+        } else if !normalized_relative_path.is_empty() {
+            // Only if not root! Avoids path spec / warning
+            // and mapname conversion failed error on unzip
+            println!("adding dir {absolute_path:?} as {normalized_relative_path:?} ...");
+            zip.add_directory(normalized_relative_path, options)?;
+        }
+    }
+    zip.finish()
+}
+
+fn main() {
+    println!("cargo:rerun-if-changed={TYPESHED_SOURCE_DIR}");
+    assert!(
+        Path::new(TYPESHED_SOURCE_DIR).is_dir(),
+        "Where is typeshed?"
+    );
+    let out_dir = std::env::var("OUT_DIR").unwrap();
+
+    // N.B. Deliberately using `format!()` instead of `Path::join()` here,
+    // so that we use `/` as a path separator on all platforms.
+    // That enables us to load the typeshed zip at compile time in `module.rs`
+    // (otherwise we'd have to dynamically determine the exact path to the typeshed zip
+    // based on the default path separator for the specific platform we're on,
+    // which can't be done at compile time.)
+    let zipped_typeshed_location = format!("{out_dir}{TYPESHED_ZIP_LOCATION}");
+
+    let zipped_typeshed = File::create(zipped_typeshed_location).unwrap();
+    zip_dir(TYPESHED_SOURCE_DIR, zipped_typeshed).unwrap();
+}

crates/red_knot_module_resolver/src/db.rs

@@ -0,0 +1,156 @@
+use ruff_db::Upcast;
+
+use crate::resolver::{
+    file_to_module,
+    internal::{ModuleNameIngredient, ModuleResolverSearchPaths},
+    resolve_module_query,
+};
+
+#[salsa::jar(db=Db)]
+pub struct Jar(
+    ModuleNameIngredient<'_>,
+    ModuleResolverSearchPaths,
+    resolve_module_query,
+    file_to_module,
+);
+
+pub trait Db: salsa::DbWithJar<Jar> + ruff_db::Db + Upcast<dyn ruff_db::Db> {}
+
+pub(crate) mod tests {
+    use std::sync;
+
+    use salsa::DebugWithDb;
+
+    use ruff_db::file_system::{FileSystem, MemoryFileSystem, OsFileSystem};
+    use ruff_db::vfs::Vfs;
+
+    use super::*;
+
+    #[salsa::db(Jar, ruff_db::Jar)]
+    pub(crate) struct TestDb {
+        storage: salsa::Storage<Self>,
+        file_system: TestFileSystem,
+        events: sync::Arc<sync::Mutex<Vec<salsa::Event>>>,
+        vfs: Vfs,
+    }
+
+    impl TestDb {
+        #[allow(unused)]
+        pub(crate) fn new() -> Self {
+            Self {
+                storage: salsa::Storage::default(),
+                file_system: TestFileSystem::Memory(MemoryFileSystem::default()),
+                events: sync::Arc::default(),
+                vfs: Vfs::with_stubbed_vendored(),
+            }
+        }
+
+        /// Returns the memory file system.
+        ///
+        /// ## Panics
+        /// If this test db isn't using a memory file system.
+        #[allow(unused)]
+        pub(crate) fn memory_file_system(&self) -> &MemoryFileSystem {
+            if let TestFileSystem::Memory(fs) = &self.file_system {
+                fs
+            } else {
+                panic!("The test db is not using a memory file system");
+            }
+        }
+
+        /// Uses the real file system instead of the memory file system.
+        ///
+        /// This useful for testing advanced file system features like permissions, symlinks, etc.
+        ///
+        /// Note that any files written to the memory file system won't be copied over.
+        #[allow(unused)]
+        pub(crate) fn with_os_file_system(&mut self) {
+            self.file_system = TestFileSystem::Os(OsFileSystem);
+        }
+
+        #[allow(unused)]
+        pub(crate) fn vfs_mut(&mut self) -> &mut Vfs {
+            &mut self.vfs
+        }
+
+        /// Takes the salsa events.
+        ///
+        /// ## Panics
+        /// If there are any pending salsa snapshots.
+        #[allow(unused)]
+        pub(crate) fn take_salsa_events(&mut self) -> Vec<salsa::Event> {
+            let inner = sync::Arc::get_mut(&mut self.events).expect("no pending salsa snapshots");
+
+            let events = inner.get_mut().unwrap();
+            std::mem::take(&mut *events)
+        }
+
+        /// Clears the salsa events.
+        ///
+        /// ## Panics
+        /// If there are any pending salsa snapshots.
+        #[allow(unused)]
+        pub(crate) fn clear_salsa_events(&mut self) {
+            self.take_salsa_events();
+        }
+    }
+
+    impl Upcast<dyn ruff_db::Db> for TestDb {
+        fn upcast(&self) -> &(dyn ruff_db::Db + 'static) {
+            self
+        }
+    }
+
+    impl ruff_db::Db for TestDb {
+        fn file_system(&self) -> &dyn ruff_db::file_system::FileSystem {
+            self.file_system.inner()
+        }
+
+        fn vfs(&self) -> &ruff_db::vfs::Vfs {
+            &self.vfs
+        }
+    }
+
+    impl Db for TestDb {}
+
+    impl salsa::Database for TestDb {
+        fn salsa_event(&self, event: salsa::Event) {
+            tracing::trace!("event: {:?}", event.debug(self));
+            let mut events = self.events.lock().unwrap();
+            events.push(event);
+        }
+    }
+
+    impl salsa::ParallelDatabase for TestDb {
+        fn snapshot(&self) -> salsa::Snapshot<Self> {
+            salsa::Snapshot::new(Self {
+                storage: self.storage.snapshot(),
+                file_system: self.file_system.snapshot(),
+                events: self.events.clone(),
+                vfs: self.vfs.snapshot(),
+            })
+        }
+    }
+
+    enum TestFileSystem {
+        Memory(MemoryFileSystem),
+        #[allow(unused)]
+        Os(OsFileSystem),
+    }
+
+    impl TestFileSystem {
+        fn inner(&self) -> &dyn FileSystem {
+            match self {
+                Self::Memory(inner) => inner,
+                Self::Os(inner) => inner,
+            }
+        }
+
+        fn snapshot(&self) -> Self {
+            match self {
+                Self::Memory(inner) => Self::Memory(inner.snapshot()),
+                Self::Os(inner) => Self::Os(inner.snapshot()),
+            }
+        }
+    }
+}

crates/red_knot_module_resolver/src/lib.rs

@@ -0,0 +1,9 @@
+mod db;
+mod module;
+mod resolver;
+mod typeshed;
+
+pub use db::{Db, Jar};
+pub use module::{ModuleKind, ModuleName};
+pub use resolver::{resolve_module, set_module_resolution_settings, ModuleResolutionSettings};
+pub use typeshed::versions::TypeshedVersions;

crates/red_knot_module_resolver/src/module.rs

@@ -0,0 +1,346 @@
+use std::fmt::Formatter;
+use std::ops::Deref;
+use std::sync::Arc;
+
+use ruff_db::file_system::FileSystemPath;
+use ruff_db::vfs::{VfsFile, VfsPath};
+use ruff_python_stdlib::identifiers::is_identifier;
+
+use crate::Db;
+
+/// A module name, e.g. `foo.bar`.
+///
+/// Always normalized to the absolute form (never a relative module name, i.e., never `.foo`).
+#[derive(Clone, Debug, Eq, PartialEq, Hash, PartialOrd, Ord)]
+pub struct ModuleName(smol_str::SmolStr);
+
+impl ModuleName {
+    /// Creates a new module name for `name`. Returns `Some` if `name` is a valid, absolute
+    /// module name and `None` otherwise.
+    ///
+    /// The module name is invalid if:
+    ///
+    /// * The name is empty
+    /// * The name is relative
+    /// * The name ends with a `.`
+    /// * The name contains a sequence of multiple dots
+    /// * A component of a name (the part between two dots) isn't a valid python identifier.
+    #[inline]
+    pub fn new(name: &str) -> Option<Self> {
+        Self::new_from_smol(smol_str::SmolStr::new(name))
+    }
+
+    /// Creates a new module name for `name` where `name` is a static string.
+    /// Returns `Some` if `name` is a valid, absolute module name and `None` otherwise.
+    ///
+    /// The module name is invalid if:
+    ///
+    /// * The name is empty
+    /// * The name is relative
+    /// * The name ends with a `.`
+    /// * The name contains a sequence of multiple dots
+    /// * A component of a name (the part between two dots) isn't a valid python identifier.
+    ///
+    /// ## Examples
+    ///
+    /// ```
+    /// use red_knot_module_resolver::ModuleName;
+    ///
+    /// assert_eq!(ModuleName::new_static("foo.bar").as_deref(), Some("foo.bar"));
+    /// assert_eq!(ModuleName::new_static(""), None);
+    /// assert_eq!(ModuleName::new_static("..foo"), None);
+    /// assert_eq!(ModuleName::new_static(".foo"), None);
+    /// assert_eq!(ModuleName::new_static("foo."), None);
+    /// assert_eq!(ModuleName::new_static("foo..bar"), None);
+    /// assert_eq!(ModuleName::new_static("2000"), None);
+    /// ```
+    #[inline]
+    pub fn new_static(name: &'static str) -> Option<Self> {
+        Self::new_from_smol(smol_str::SmolStr::new_static(name))
+    }
+
+    fn new_from_smol(name: smol_str::SmolStr) -> Option<Self> {
+        if name.is_empty() {
+            return None;
+        }
+
+        if name.split('.').all(is_identifier) {
+            Some(Self(name))
+        } else {
+            None
+        }
+    }
+
+    /// An iterator over the components of the module name:
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use red_knot_module_resolver::ModuleName;
+    ///
+    /// assert_eq!(ModuleName::new_static("foo.bar.baz").unwrap().components().collect::<Vec<_>>(), vec!["foo", "bar", "baz"]);
+    /// ```
+    pub fn components(&self) -> impl DoubleEndedIterator<Item = &str> {
+        self.0.split('.')
+    }
+
+    /// The name of this module's immediate parent, if it has a parent.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use red_knot_module_resolver::ModuleName;
+    ///
+    /// assert_eq!(ModuleName::new_static("foo.bar").unwrap().parent(), Some(ModuleName::new_static("foo").unwrap()));
+    /// assert_eq!(ModuleName::new_static("foo.bar.baz").unwrap().parent(), Some(ModuleName::new_static("foo.bar").unwrap()));
+    /// assert_eq!(ModuleName::new_static("root").unwrap().parent(), None);
+    /// ```
+    pub fn parent(&self) -> Option<ModuleName> {
+        let (parent, _) = self.0.rsplit_once('.')?;
+
+        Some(Self(smol_str::SmolStr::new(parent)))
+    }
+
+    /// Returns `true` if the name starts with `other`.
+    ///
+    /// This is equivalent to checking if `self` is a sub-module of `other`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use red_knot_module_resolver::ModuleName;
+    ///
+    /// assert!(ModuleName::new_static("foo.bar").unwrap().starts_with(&ModuleName::new_static("foo").unwrap()));
+    ///
+    /// assert!(!ModuleName::new_static("foo.bar").unwrap().starts_with(&ModuleName::new_static("bar").unwrap()));
+    /// assert!(!ModuleName::new_static("foo_bar").unwrap().starts_with(&ModuleName::new_static("foo").unwrap()));
+    /// ```
+    pub fn starts_with(&self, other: &ModuleName) -> bool {
+        let mut self_components = self.components();
+        let other_components = other.components();
+
+        for other_component in other_components {
+            if self_components.next() != Some(other_component) {
+                return false;
+            }
+        }
+
+        true
+    }
+
+    #[inline]
+    pub fn as_str(&self) -> &str {
+        &self.0
+    }
+
+    pub(crate) fn from_relative_path(path: &FileSystemPath) -> Option<Self> {
+        let path = if path.ends_with("__init__.py") || path.ends_with("__init__.pyi") {
+            path.parent()?
+        } else {
+            path
+        };
+
+        let name = if let Some(parent) = path.parent() {
+            let mut name = String::with_capacity(path.as_str().len());
+
+            for component in parent.components() {
+                name.push_str(component.as_os_str().to_str()?);
+                name.push('.');
+            }
+
+            // SAFETY: Unwrap is safe here or `parent` would have returned `None`.
+            name.push_str(path.file_stem().unwrap());
+
+            smol_str::SmolStr::from(name)
+        } else {
+            smol_str::SmolStr::new(path.file_stem()?)
+        };
+
+        Some(Self(name))
+    }
+}
+
+impl Deref for ModuleName {
+    type Target = str;
+
+    #[inline]
+    fn deref(&self) -> &Self::Target {
+        self.as_str()
+    }
+}
+
+impl PartialEq<str> for ModuleName {
+    fn eq(&self, other: &str) -> bool {
+        self.as_str() == other
+    }
+}
+
+impl PartialEq<ModuleName> for str {
+    fn eq(&self, other: &ModuleName) -> bool {
+        self == other.as_str()
+    }
+}
+
+impl std::fmt::Display for ModuleName {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        f.write_str(&self.0)
+    }
+}
+
+/// Representation of a Python module.
+#[derive(Clone, PartialEq, Eq)]
+pub struct Module {
+    inner: Arc<ModuleInner>,
+}
+
+impl Module {
+    pub(crate) fn new(
+        name: ModuleName,
+        kind: ModuleKind,
+        search_path: ModuleSearchPath,
+        file: VfsFile,
+    ) -> Self {
+        Self {
+            inner: Arc::new(ModuleInner {
+                name,
+                kind,
+                search_path,
+                file,
+            }),
+        }
+    }
+
+    /// The absolute name of the module (e.g. `foo.bar`)
+    pub fn name(&self) -> &ModuleName {
+        &self.inner.name
+    }
+
+    /// The file to the source code that defines this module
+    pub fn file(&self) -> VfsFile {
+        self.inner.file
+    }
+
+    /// The search path from which the module was resolved.
+    pub fn search_path(&self) -> &ModuleSearchPath {
+        &self.inner.search_path
+    }
+
+    /// Determine whether this module is a single-file module or a package
+    pub fn kind(&self) -> ModuleKind {
+        self.inner.kind
+    }
+}
+
+impl std::fmt::Debug for Module {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("Module")
+            .field("name", &self.name())
+            .field("kind", &self.kind())
+            .field("file", &self.file())
+            .field("search_path", &self.search_path())
+            .finish()
+    }
+}
+
+impl salsa::DebugWithDb<dyn Db> for Module {
+    fn fmt(&self, f: &mut Formatter<'_>, db: &dyn Db) -> std::fmt::Result {
+        f.debug_struct("Module")
+            .field("name", &self.name())
+            .field("kind", &self.kind())
+            .field("file", &self.file().debug(db.upcast()))
+            .field("search_path", &self.search_path())
+            .finish()
+    }
+}
+
+#[derive(PartialEq, Eq)]
+struct ModuleInner {
+    name: ModuleName,
+    kind: ModuleKind,
+    search_path: ModuleSearchPath,
+    file: VfsFile,
+}
+
+#[derive(Copy, Clone, Debug, Eq, PartialEq, Hash)]
+pub enum ModuleKind {
+    /// A single-file module (e.g. `foo.py` or `foo.pyi`)
+    Module,
+
+    /// A python package (`foo/__init__.py` or `foo/__init__.pyi`)
+    Package,
+}
+
+/// A search path in which to search modules.
+/// Corresponds to a path in [`sys.path`](https://docs.python.org/3/library/sys_path_init.html) at runtime.
+///
+/// Cloning a search path is cheap because it's an `Arc`.
+#[derive(Clone, PartialEq, Eq)]
+pub struct ModuleSearchPath {
+    inner: Arc<ModuleSearchPathInner>,
+}
+
+impl ModuleSearchPath {
+    pub fn new<P>(path: P, kind: ModuleSearchPathKind) -> Self
+    where
+        P: Into<VfsPath>,
+    {
+        Self {
+            inner: Arc::new(ModuleSearchPathInner {
+                path: path.into(),
+                kind,
+            }),
+        }
+    }
+
+    /// Determine whether this is a first-party, third-party or standard-library search path
+    pub fn kind(&self) -> ModuleSearchPathKind {
+        self.inner.kind
+    }
+
+    /// Return the location of the search path on the file system
+    pub fn path(&self) -> &VfsPath {
+        &self.inner.path
+    }
+}
+
+impl std::fmt::Debug for ModuleSearchPath {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("ModuleSearchPath")
+            .field("path", &self.inner.path)
+            .field("kind", &self.kind())
+            .finish()
+    }
+}
+
+#[derive(Eq, PartialEq)]
+struct ModuleSearchPathInner {
+    path: VfsPath,
+    kind: ModuleSearchPathKind,
+}
+
+/// Enumeration of the different kinds of search paths type checkers are expected to support.
+///
+/// N.B. Although we don't implement `Ord` for this enum, they are ordered in terms of the
+/// priority that we want to give these modules when resolving them.
+/// This is roughly [the order given in the typing spec], but typeshed's stubs
+/// for the standard library are moved higher up to match Python's semantics at runtime.
+///
+/// [the order given in the typing spec]: https://typing.readthedocs.io/en/latest/spec/distributing.html#import-resolution-ordering
+#[derive(Copy, Clone, Debug, Eq, PartialEq, Hash)]
+pub enum ModuleSearchPathKind {
+    /// "Extra" paths provided by the user in a config file, env var or CLI flag.
+    /// E.g. mypy's `MYPYPATH` env var, or pyright's `stubPath` configuration setting
+    Extra,
+
+    /// Files in the project we're directly being invoked on
+    FirstParty,
+
+    /// The `stdlib` directory of typeshed (either vendored or custom)
+    StandardLibrary,
+
+    /// Stubs or runtime modules installed in site-packages
+    SitePackagesThirdParty,
+
+    /// Vendored third-party stubs from typeshed
+    VendoredThirdParty,
+}

crates/red_knot_module_resolver/src/resolver.rs

@@ -0,0 +1,938 @@
+use salsa::DebugWithDb;
+use std::ops::Deref;
+
+use ruff_db::file_system::{FileSystem, FileSystemPath, FileSystemPathBuf};
+use ruff_db::vfs::{system_path_to_file, vfs_path_to_file, VfsFile, VfsPath};
+
+use crate::module::{Module, ModuleKind, ModuleName, ModuleSearchPath, ModuleSearchPathKind};
+use crate::resolver::internal::ModuleResolverSearchPaths;
+use crate::Db;
+
+const TYPESHED_STDLIB_DIRECTORY: &str = "stdlib";
+
+/// Configures the module search paths for the module resolver.
+///
+/// Must be called before calling any other module resolution functions.
+pub fn set_module_resolution_settings(db: &mut dyn Db, config: ModuleResolutionSettings) {
+    // There's no concurrency issue here because we hold a `&mut dyn Db` reference. No other
+    // thread can mutate the `Db` while we're in this call, so using `try_get` to test if
+    // the settings have already been set is safe.
+    if let Some(existing) = ModuleResolverSearchPaths::try_get(db) {
+        existing
+            .set_search_paths(db)
+            .to(config.into_ordered_search_paths());
+    } else {
+        ModuleResolverSearchPaths::new(db, config.into_ordered_search_paths());
+    }
+}
+
+/// Resolves a module name to a module.
+pub fn resolve_module(db: &dyn Db, module_name: ModuleName) -> Option<Module> {
+    let interned_name = internal::ModuleNameIngredient::new(db, module_name);
+
+    resolve_module_query(db, interned_name)
+}
+
+/// Salsa query that resolves an interned [`ModuleNameIngredient`] to a module.
+///
+/// This query should not be called directly. Instead, use [`resolve_module`]. It only exists
+/// because Salsa requires the module name to be an ingredient.
+#[salsa::tracked]
+pub(crate) fn resolve_module_query<'db>(
+    db: &'db dyn Db,
+    module_name: internal::ModuleNameIngredient<'db>,
+) -> Option<Module> {
+    let _ = tracing::trace_span!("resolve_module", module_name = ?module_name.debug(db)).enter();
+
+    let name = module_name.name(db);
+
+    let (search_path, module_file, kind) = resolve_name(db, name)?;
+
+    let module = Module::new(name.clone(), kind, search_path, module_file);
+
+    Some(module)
+}
+
+/// Resolves the module for the given path.
+///
+/// Returns `None` if the path is not a module locatable via `sys.path`.
+#[tracing::instrument(level = "debug", skip(db))]
+pub fn path_to_module(db: &dyn Db, path: &VfsPath) -> Option<Module> {
+    // It's not entirely clear on first sight why this method calls `file_to_module` instead of
+    // it being the other way round, considering that the first thing that `file_to_module` does
+    // is to retrieve the file's path.
+    //
+    // The reason is that `file_to_module` is a tracked Salsa query and salsa queries require that
+    // all arguments are Salsa ingredients (something stored in Salsa). `Path`s aren't salsa ingredients but
+    // `VfsFile` is. So what we do here is to retrieve the `path`'s `VfsFile` so that we can make
+    // use of Salsa's caching and invalidation.
+    let file = vfs_path_to_file(db.upcast(), path)?;
+    file_to_module(db, file)
+}
+
+/// Resolves the module for the file with the given id.
+///
+/// Returns `None` if the file is not a module locatable via `sys.path`.
+#[salsa::tracked]
+#[allow(unused)]
+pub(crate) fn file_to_module(db: &dyn Db, file: VfsFile) -> Option<Module> {
+    let _ = tracing::trace_span!("file_to_module", file = ?file.debug(db.upcast())).enter();
+
+    let path = file.path(db.upcast());
+
+    let search_paths = module_search_paths(db);
+
+    let relative_path = search_paths
+        .iter()
+        .find_map(|root| match (root.path(), path) {
+            (VfsPath::FileSystem(root_path), VfsPath::FileSystem(path)) => {
+                let relative_path = path.strip_prefix(root_path).ok()?;
+                Some(relative_path)
+            }
+            (VfsPath::Vendored(_), VfsPath::Vendored(_)) => {
+                todo!("Add support for vendored modules")
+            }
+            (VfsPath::Vendored(_), VfsPath::FileSystem(_))
+            | (VfsPath::FileSystem(_), VfsPath::Vendored(_)) => None,
+        })?;
+
+    let module_name = ModuleName::from_relative_path(relative_path)?;
+
+    // Resolve the module name to see if Python would resolve the name to the same path.
+    // If it doesn't, then that means that multiple modules have the same name in different
+    // root paths, but that the module corresponding to `path` is in a lower priority search path,
+    // in which case we ignore it.
+    let module = resolve_module(db, module_name)?;
+
+    if file == module.file() {
+        Some(module)
+    } else {
+        // This path is for a module with the same name but with a different precedence. For example:
+        // ```
+        // src/foo.py
+        // src/foo/__init__.py
+        // ```
+        // The module name of `src/foo.py` is `foo`, but the module loaded by Python is `src/foo/__init__.py`.
+        // That means we need to ignore `src/foo.py` even though it resolves to the same module name.
+        None
+    }
+}
+
+/// Configures the search paths that are used to resolve modules.
+#[derive(Eq, PartialEq, Debug)]
+pub struct ModuleResolutionSettings {
+    /// List of user-provided paths that should take first priority in the module resolution.
+    /// Examples in other type checkers are mypy's MYPYPATH environment variable,
+    /// or pyright's stubPath configuration setting.
+    pub extra_paths: Vec<FileSystemPathBuf>,
+
+    /// The root of the workspace, used for finding first-party modules.
+    pub workspace_root: FileSystemPathBuf,
+
+    /// The path to the user's `site-packages` directory, where third-party packages from ``PyPI`` are installed.
+    pub site_packages: Option<FileSystemPathBuf>,
+
+    /// Optional path to standard-library typeshed stubs.
+    /// Currently this has to be a directory that exists on disk.
+    ///
+    /// (TODO: fall back to vendored stubs if no custom directory is provided.)
+    pub custom_typeshed: Option<FileSystemPathBuf>,
+}
+
+impl ModuleResolutionSettings {
+    /// Implementation of PEP 561's module resolution order
+    /// (with some small, deliberate, differences)
+    fn into_ordered_search_paths(self) -> OrderedSearchPaths {
+        let ModuleResolutionSettings {
+            extra_paths,
+            workspace_root,
+            site_packages,
+            custom_typeshed,
+        } = self;
+
+        let mut paths: Vec<_> = extra_paths
+            .into_iter()
+            .map(|path| ModuleSearchPath::new(path, ModuleSearchPathKind::Extra))
+            .collect();
+
+        paths.push(ModuleSearchPath::new(
+            workspace_root,
+            ModuleSearchPathKind::FirstParty,
+        ));
+
+        // TODO fallback to vendored typeshed stubs if no custom typeshed directory is provided by the user
+        if let Some(custom_typeshed) = custom_typeshed {
+            paths.push(ModuleSearchPath::new(
+                custom_typeshed.join(TYPESHED_STDLIB_DIRECTORY),
+                ModuleSearchPathKind::StandardLibrary,
+            ));
+        }
+
+        // TODO vendor typeshed's third-party stubs as well as the stdlib and fallback to them as a final step
+        if let Some(site_packages) = site_packages {
+            paths.push(ModuleSearchPath::new(
+                site_packages,
+                ModuleSearchPathKind::SitePackagesThirdParty,
+            ));
+        }
+
+        OrderedSearchPaths(paths)
+    }
+}
+
+/// A resolved module resolution order, implementing PEP 561
+/// (with some small, deliberate differences)
+#[derive(Clone, Debug, Default, Eq, PartialEq)]
+pub(crate) struct OrderedSearchPaths(Vec<ModuleSearchPath>);
+
+impl Deref for OrderedSearchPaths {
+    type Target = [ModuleSearchPath];
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+// The singleton methods generated by salsa are all `pub` instead of `pub(crate)` which triggers
+// `unreachable_pub`. Work around this by creating a module and allow `unreachable_pub` for it.
+// Salsa also generates uses to `_db` variables for `interned` which triggers `clippy::used_underscore_binding`. Suppress that too
+// TODO(micha): Contribute a fix for this upstream where the singleton methods have the same visibility as the struct.
+#[allow(unreachable_pub, clippy::used_underscore_binding)]
+pub(crate) mod internal {
+    use crate::module::ModuleName;
+    use crate::resolver::OrderedSearchPaths;
+
+    #[salsa::input(singleton)]
+    pub(crate) struct ModuleResolverSearchPaths {
+        #[return_ref]
+        pub(super) search_paths: OrderedSearchPaths,
+    }
+
+    /// A thin wrapper around `ModuleName` to make it a Salsa ingredient.
+    ///
+    /// This is needed because Salsa requires that all query arguments are salsa ingredients.
+    #[salsa::interned]
+    pub(crate) struct ModuleNameIngredient<'db> {
+        #[return_ref]
+        pub(super) name: ModuleName,
+    }
+}
+
+fn module_search_paths(db: &dyn Db) -> &[ModuleSearchPath] {
+    ModuleResolverSearchPaths::get(db).search_paths(db)
+}
+
+/// Given a module name and a list of search paths in which to lookup modules,
+/// attempt to resolve the module name
+fn resolve_name(db: &dyn Db, name: &ModuleName) -> Option<(ModuleSearchPath, VfsFile, ModuleKind)> {
+    let search_paths = module_search_paths(db);
+
+    for search_path in search_paths {
+        let mut components = name.components();
+        let module_name = components.next_back()?;
+
+        let VfsPath::FileSystem(fs_search_path) = search_path.path() else {
+            todo!("Vendored search paths are not yet supported");
+        };
+
+        match resolve_package(db.file_system(), fs_search_path, components) {
+            Ok(resolved_package) => {
+                let mut package_path = resolved_package.path;
+
+                package_path.push(module_name);
+
+                // Must be a `__init__.pyi` or `__init__.py` or it isn't a package.
+                let kind = if db.file_system().is_directory(&package_path) {
+                    package_path.push("__init__");
+                    ModuleKind::Package
+                } else {
+                    ModuleKind::Module
+                };
+
+                // TODO Implement full https://peps.python.org/pep-0561/#type-checker-module-resolution-order resolution
+                let stub = package_path.with_extension("pyi");
+
+                if let Some(stub) = system_path_to_file(db.upcast(), &stub) {
+                    return Some((search_path.clone(), stub, kind));
+                }
+
+                let module = package_path.with_extension("py");
+
+                if let Some(module) = system_path_to_file(db.upcast(), &module) {
+                    return Some((search_path.clone(), module, kind));
+                }
+
+                // For regular packages, don't search the next search path. All files of that
+                // package must be in the same location
+                if resolved_package.kind.is_regular_package() {
+                    return None;
+                }
+            }
+            Err(parent_kind) => {
+                if parent_kind.is_regular_package() {
+                    // For regular packages, don't search the next search path.
+                    return None;
+                }
+            }
+        }
+    }
+
+    None
+}
+
+fn resolve_package<'a, I>(
+    fs: &dyn FileSystem,
+    module_search_path: &FileSystemPath,
+    components: I,
+) -> Result<ResolvedPackage, PackageKind>
+where
+    I: Iterator<Item = &'a str>,
+{
+    let mut package_path = module_search_path.to_path_buf();
+
+    // `true` if inside a folder that is a namespace package (has no `__init__.py`).
+    // Namespace packages are special because they can be spread across multiple search paths.
+    // https://peps.python.org/pep-0420/
+    let mut in_namespace_package = false;
+
+    // `true` if resolving a sub-package. For example, `true` when resolving `bar` of `foo.bar`.
+    let mut in_sub_package = false;
+
+    // For `foo.bar.baz`, test that `foo` and `baz` both contain a `__init__.py`.
+    for folder in components {
+        package_path.push(folder);
+
+        let has_init_py = fs.is_file(&package_path.join("__init__.py"))
+            || fs.is_file(&package_path.join("__init__.pyi"));
+
+        if has_init_py {
+            in_namespace_package = false;
+        } else if fs.is_directory(&package_path) {
+            // A directory without an `__init__.py` is a namespace package, continue with the next folder.
+            in_namespace_package = true;
+        } else if in_namespace_package {
+            // Package not found but it is part of a namespace package.
+            return Err(PackageKind::Namespace);
+        } else if in_sub_package {
+            // A regular sub package wasn't found.
+            return Err(PackageKind::Regular);
+        } else {
+            // We couldn't find `foo` for `foo.bar.baz`, search the next search path.
+            return Err(PackageKind::Root);
+        }
+
+        in_sub_package = true;
+    }
+
+    let kind = if in_namespace_package {
+        PackageKind::Namespace
+    } else if in_sub_package {
+        PackageKind::Regular
+    } else {
+        PackageKind::Root
+    };
+
+    Ok(ResolvedPackage {
+        kind,
+        path: package_path,
+    })
+}
+
+#[derive(Debug)]
+struct ResolvedPackage {
+    path: FileSystemPathBuf,
+    kind: PackageKind,
+}
+
+#[derive(Copy, Clone, Eq, PartialEq, Debug)]
+enum PackageKind {
+    /// A root package or module. E.g. `foo` in `foo.bar.baz` or just `foo`.
+    Root,
+
+    /// A regular sub-package where the parent contains an `__init__.py`.
+    ///
+    /// For example, `bar` in `foo.bar` when the `foo` directory contains an `__init__.py`.
+    Regular,
+
+    /// A sub-package in a namespace package. A namespace package is a package without an `__init__.py`.
+    ///
+    /// For example, `bar` in `foo.bar` if the `foo` directory contains no `__init__.py`.
+    Namespace,
+}
+
+impl PackageKind {
+    const fn is_regular_package(self) -> bool {
+        matches!(self, PackageKind::Regular)
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    use ruff_db::file_system::{FileSystemPath, FileSystemPathBuf};
+    use ruff_db::vfs::{system_path_to_file, VfsFile, VfsPath};
+
+    use crate::db::tests::TestDb;
+    use crate::module::{ModuleKind, ModuleName};
+
+    use super::{
+        path_to_module, resolve_module, set_module_resolution_settings, ModuleResolutionSettings,
+        TYPESHED_STDLIB_DIRECTORY,
+    };
+
+    struct TestCase {
+        db: TestDb,
+
+        src: FileSystemPathBuf,
+        custom_typeshed: FileSystemPathBuf,
+        site_packages: FileSystemPathBuf,
+    }
+
+    fn create_resolver() -> std::io::Result<TestCase> {
+        let mut db = TestDb::new();
+
+        let src = FileSystemPath::new("src").to_path_buf();
+        let site_packages = FileSystemPath::new("site_packages").to_path_buf();
+        let custom_typeshed = FileSystemPath::new("typeshed").to_path_buf();
+
+        let fs = db.memory_file_system();
+
+        fs.create_directory_all(&src)?;
+        fs.create_directory_all(&site_packages)?;
+        fs.create_directory_all(&custom_typeshed)?;
+
+        let settings = ModuleResolutionSettings {
+            extra_paths: vec![],
+            workspace_root: src.clone(),
+            site_packages: Some(site_packages.clone()),
+            custom_typeshed: Some(custom_typeshed.clone()),
+        };
+
+        set_module_resolution_settings(&mut db, settings);
+
+        Ok(TestCase {
+            db,
+            src,
+            custom_typeshed,
+            site_packages,
+        })
+    }
+
+    #[test]
+    fn first_party_module() -> anyhow::Result<()> {
+        let TestCase { db, src, .. } = create_resolver()?;
+
+        let foo_module_name = ModuleName::new_static("foo").unwrap();
+        let foo_path = src.join("foo.py");
+        db.memory_file_system()
+            .write_file(&foo_path, "print('Hello, world!')")?;
+
+        let foo_module = resolve_module(&db, foo_module_name.clone()).unwrap();
+
+        assert_eq!(
+            Some(&foo_module),
+            resolve_module(&db, foo_module_name.clone()).as_ref()
+        );
+
+        assert_eq!("foo", foo_module.name());
+        assert_eq!(&src, foo_module.search_path().path());
+        assert_eq!(ModuleKind::Module, foo_module.kind());
+        assert_eq!(&foo_path, foo_module.file().path(&db));
+
+        assert_eq!(
+            Some(foo_module),
+            path_to_module(&db, &VfsPath::FileSystem(foo_path))
+        );
+
+        Ok(())
+    }
+
+    #[test]
+    fn stdlib() -> anyhow::Result<()> {
+        let TestCase {
+            db,
+            custom_typeshed,
+            ..
+        } = create_resolver()?;
+
+        let stdlib_dir = custom_typeshed.join(TYPESHED_STDLIB_DIRECTORY);
+        let functools_path = stdlib_dir.join("functools.py");
+        db.memory_file_system()
+            .write_file(&functools_path, "def update_wrapper(): ...")?;
+
+        let functools_module_name = ModuleName::new_static("functools").unwrap();
+        let functools_module = resolve_module(&db, functools_module_name.clone()).unwrap();
+
+        assert_eq!(
+            Some(&functools_module),
+            resolve_module(&db, functools_module_name).as_ref()
+        );
+
+        assert_eq!(&stdlib_dir, functools_module.search_path().path());
+        assert_eq!(ModuleKind::Module, functools_module.kind());
+        assert_eq!(&functools_path.clone(), functools_module.file().path(&db));
+
+        assert_eq!(
+            Some(functools_module),
+            path_to_module(&db, &VfsPath::FileSystem(functools_path))
+        );
+
+        Ok(())
+    }
+
+    #[test]
+    fn first_party_precedence_over_stdlib() -> anyhow::Result<()> {
+        let TestCase {
+            db,
+            src,
+            custom_typeshed,
+            ..
+        } = create_resolver()?;
+
+        let stdlib_dir = custom_typeshed.join(TYPESHED_STDLIB_DIRECTORY);
+        let stdlib_functools_path = stdlib_dir.join("functools.py");
+        let first_party_functools_path = src.join("functools.py");
+
+        db.memory_file_system().write_files([
+            (&stdlib_functools_path, "def update_wrapper(): ..."),
+            (&first_party_functools_path, "def update_wrapper(): ..."),
+        ])?;
+
+        let functools_module_name = ModuleName::new_static("functools").unwrap();
+        let functools_module = resolve_module(&db, functools_module_name.clone()).unwrap();
+
+        assert_eq!(
+            Some(&functools_module),
+            resolve_module(&db, functools_module_name).as_ref()
+        );
+        assert_eq!(&src, functools_module.search_path().path());
+        assert_eq!(ModuleKind::Module, functools_module.kind());
+        assert_eq!(
+            &first_party_functools_path.clone(),
+            functools_module.file().path(&db)
+        );
+
+        assert_eq!(
+            Some(functools_module),
+            path_to_module(&db, &VfsPath::FileSystem(first_party_functools_path))
+        );
+
+        Ok(())
+    }
+
+    // TODO: Port typeshed test case. Porting isn't possible at the moment because the vendored zip
+    //   is part of the red knot crate
+    // #[test]
+    // fn typeshed_zip_created_at_build_time() -> anyhow::Result<()> {
+    //     // The file path here is hardcoded in this crate's `build.rs` script.
+    //     // Luckily this crate will fail to build if this file isn't available at build time.
+    //     const TYPESHED_ZIP_BYTES: &[u8] =
+    //         include_bytes!(concat!(env!("OUT_DIR"), "/zipped_typeshed.zip"));
+    //     assert!(!TYPESHED_ZIP_BYTES.is_empty());
+    //     let mut typeshed_zip_archive = ZipArchive::new(Cursor::new(TYPESHED_ZIP_BYTES))?;
+    //
+    //     let path_to_functools = Path::new("stdlib").join("functools.pyi");
+    //     let mut functools_module_stub = typeshed_zip_archive
+    //         .by_name(path_to_functools.to_str().unwrap())
+    //         .unwrap();
+    //     assert!(functools_module_stub.is_file());
+    //
+    //     let mut functools_module_stub_source = String::new();
+    //     functools_module_stub.read_to_string(&mut functools_module_stub_source)?;
+    //
+    //     assert!(functools_module_stub_source.contains("def update_wrapper("));
+    //     Ok(())
+    // }
+
+    #[test]
+    fn resolve_package() -> anyhow::Result<()> {
+        let TestCase { src, db, .. } = create_resolver()?;
+
+        let foo_dir = src.join("foo");
+        let foo_path = foo_dir.join("__init__.py");
+
+        db.memory_file_system()
+            .write_file(&foo_path, "print('Hello, world!')")?;
+
+        let foo_module = resolve_module(&db, ModuleName::new_static("foo").unwrap()).unwrap();
+
+        assert_eq!("foo", foo_module.name());
+        assert_eq!(&src, foo_module.search_path().path());
+        assert_eq!(&foo_path, foo_module.file().path(&db));
+
+        assert_eq!(
+            Some(&foo_module),
+            path_to_module(&db, &VfsPath::FileSystem(foo_path)).as_ref()
+        );
+
+        // Resolving by directory doesn't resolve to the init file.
+        assert_eq!(None, path_to_module(&db, &VfsPath::FileSystem(foo_dir)));
+
+        Ok(())
+    }
+
+    #[test]
+    fn package_priority_over_module() -> anyhow::Result<()> {
+        let TestCase { db, src, .. } = create_resolver()?;
+
+        let foo_dir = src.join("foo");
+        let foo_init = foo_dir.join("__init__.py");
+
+        db.memory_file_system()
+            .write_file(&foo_init, "print('Hello, world!')")?;
+
+        let foo_py = src.join("foo.py");
+        db.memory_file_system()
+            .write_file(&foo_py, "print('Hello, world!')")?;
+
+        let foo_module = resolve_module(&db, ModuleName::new_static("foo").unwrap()).unwrap();
+
+        assert_eq!(&src, foo_module.search_path().path());
+        assert_eq!(&foo_init, foo_module.file().path(&db));
+        assert_eq!(ModuleKind::Package, foo_module.kind());
+
+        assert_eq!(
+            Some(foo_module),
+            path_to_module(&db, &VfsPath::FileSystem(foo_init))
+        );
+        assert_eq!(None, path_to_module(&db, &VfsPath::FileSystem(foo_py)));
+
+        Ok(())
+    }
+
+    #[test]
+    fn typing_stub_over_module() -> anyhow::Result<()> {
+        let TestCase { db, src, .. } = create_resolver()?;
+
+        let foo_stub = src.join("foo.pyi");
+        let foo_py = src.join("foo.py");
+        db.memory_file_system()
+            .write_files([(&foo_stub, "x: int"), (&foo_py, "print('Hello, world!')")])?;
+
+        let foo = resolve_module(&db, ModuleName::new_static("foo").unwrap()).unwrap();
+
+        assert_eq!(&src, foo.search_path().path());
+        assert_eq!(&foo_stub, foo.file().path(&db));
+
+        assert_eq!(
+            Some(foo),
+            path_to_module(&db, &VfsPath::FileSystem(foo_stub))
+        );
+        assert_eq!(None, path_to_module(&db, &VfsPath::FileSystem(foo_py)));
+
+        Ok(())
+    }
+
+    #[test]
+    fn sub_packages() -> anyhow::Result<()> {
+        let TestCase { db, src, .. } = create_resolver()?;
+
+        let foo = src.join("foo");
+        let bar = foo.join("bar");
+        let baz = bar.join("baz.py");
+
+        db.memory_file_system().write_files([
+            (&foo.join("__init__.py"), ""),
+            (&bar.join("__init__.py"), ""),
+            (&baz, "print('Hello, world!')"),
+        ])?;
+
+        let baz_module =
+            resolve_module(&db, ModuleName::new_static("foo.bar.baz").unwrap()).unwrap();
+
+        assert_eq!(&src, baz_module.search_path().path());
+        assert_eq!(&baz, baz_module.file().path(&db));
+
+        assert_eq!(
+            Some(baz_module),
+            path_to_module(&db, &VfsPath::FileSystem(baz))
+        );
+
+        Ok(())
+    }
+
+    #[test]
+    fn namespace_package() -> anyhow::Result<()> {
+        let TestCase {
+            db,
+            src,
+            site_packages,
+            ..
+        } = create_resolver()?;
+
+        // From [PEP420](https://peps.python.org/pep-0420/#nested-namespace-packages).
+        // But uses `src` for `project1` and `site_packages2` for `project2`.
+        // ```
+        // src
+        //   parent
+        //     child
+        //       one.py
+        // site_packages
+        //   parent
+        //     child
+        //       two.py
+        // ```
+
+        let parent1 = src.join("parent");
+        let child1 = parent1.join("child");
+        let one = child1.join("one.py");
+
+        let parent2 = site_packages.join("parent");
+        let child2 = parent2.join("child");
+        let two = child2.join("two.py");
+
+        db.memory_file_system().write_files([
+            (&one, "print('Hello, world!')"),
+            (&two, "print('Hello, world!')"),
+        ])?;
+
+        let one_module =
+            resolve_module(&db, ModuleName::new_static("parent.child.one").unwrap()).unwrap();
+
+        assert_eq!(
+            Some(one_module),
+            path_to_module(&db, &VfsPath::FileSystem(one))
+        );
+
+        let two_module =
+            resolve_module(&db, ModuleName::new_static("parent.child.two").unwrap()).unwrap();
+        assert_eq!(
+            Some(two_module),
+            path_to_module(&db, &VfsPath::FileSystem(two))
+        );
+
+        Ok(())
+    }
+
+    #[test]
+    fn regular_package_in_namespace_package() -> anyhow::Result<()> {
+        let TestCase {
+            db,
+            src,
+            site_packages,
+            ..
+        } = create_resolver()?;
+
+        // Adopted test case from the [PEP420 examples](https://peps.python.org/pep-0420/#nested-namespace-packages).
+        // The `src/parent/child` package is a regular package. Therefore, `site_packages/parent/child/two.py` should not be resolved.
+        // ```
+        // src
+        //   parent
+        //     child
+        //       one.py
+        // site_packages
+        //   parent
+        //     child
+        //       two.py
+        // ```
+
+        let parent1 = src.join("parent");
+        let child1 = parent1.join("child");
+        let one = child1.join("one.py");
+
+        let parent2 = site_packages.join("parent");
+        let child2 = parent2.join("child");
+        let two = child2.join("two.py");
+
+        db.memory_file_system().write_files([
+            (&child1.join("__init__.py"), "print('Hello, world!')"),
+            (&one, "print('Hello, world!')"),
+            (&two, "print('Hello, world!')"),
+        ])?;
+
+        let one_module =
+            resolve_module(&db, ModuleName::new_static("parent.child.one").unwrap()).unwrap();
+
+        assert_eq!(
+            Some(one_module),
+            path_to_module(&db, &VfsPath::FileSystem(one))
+        );
+
+        assert_eq!(
+            None,
+            resolve_module(&db, ModuleName::new_static("parent.child.two").unwrap())
+        );
+        Ok(())
+    }
+
+    #[test]
+    fn module_search_path_priority() -> anyhow::Result<()> {
+        let TestCase {
+            db,
+            src,
+            site_packages,
+            ..
+        } = create_resolver()?;
+
+        let foo_src = src.join("foo.py");
+        let foo_site_packages = site_packages.join("foo.py");
+
+        db.memory_file_system()
+            .write_files([(&foo_src, ""), (&foo_site_packages, "")])?;
+
+        let foo_module = resolve_module(&db, ModuleName::new_static("foo").unwrap()).unwrap();
+
+        assert_eq!(&src, foo_module.search_path().path());
+        assert_eq!(&foo_src, foo_module.file().path(&db));
+
+        assert_eq!(
+            Some(foo_module),
+            path_to_module(&db, &VfsPath::FileSystem(foo_src))
+        );
+        assert_eq!(
+            None,
+            path_to_module(&db, &VfsPath::FileSystem(foo_site_packages))
+        );
+
+        Ok(())
+    }
+
+    #[test]
+    #[cfg(target_family = "unix")]
+    fn symlink() -> anyhow::Result<()> {
+        let TestCase {
+            mut db,
+            src,
+            site_packages,
+            custom_typeshed,
+        } = create_resolver()?;
+
+        db.with_os_file_system();
+
+        let temp_dir = tempfile::tempdir()?;
+        let root = FileSystemPath::from_std_path(temp_dir.path()).unwrap();
+
+        let src = root.join(src);
+        let site_packages = root.join(site_packages);
+        let custom_typeshed = root.join(custom_typeshed);
+
+        let foo = src.join("foo.py");
+        let bar = src.join("bar.py");
+
+        std::fs::create_dir_all(src.as_std_path())?;
+        std::fs::create_dir_all(site_packages.as_std_path())?;
+        std::fs::create_dir_all(custom_typeshed.as_std_path())?;
+
+        std::fs::write(foo.as_std_path(), "")?;
+        std::os::unix::fs::symlink(foo.as_std_path(), bar.as_std_path())?;
+
+        let settings = ModuleResolutionSettings {
+            extra_paths: vec![],
+            workspace_root: src.clone(),
+            site_packages: Some(site_packages),
+            custom_typeshed: Some(custom_typeshed),
+        };
+
+        set_module_resolution_settings(&mut db, settings);
+
+        let foo_module = resolve_module(&db, ModuleName::new_static("foo").unwrap()).unwrap();
+        let bar_module = resolve_module(&db, ModuleName::new_static("bar").unwrap()).unwrap();
+
+        assert_ne!(foo_module, bar_module);
+
+        assert_eq!(&src, foo_module.search_path().path());
+        assert_eq!(&foo, foo_module.file().path(&db));
+
+        // `foo` and `bar` shouldn't resolve to the same file
+
+        assert_eq!(&src, bar_module.search_path().path());
+        assert_eq!(&bar, bar_module.file().path(&db));
+        assert_eq!(&foo, foo_module.file().path(&db));
+
+        assert_ne!(&foo_module, &bar_module);
+
+        assert_eq!(
+            Some(foo_module),
+            path_to_module(&db, &VfsPath::FileSystem(foo))
+        );
+        assert_eq!(
+            Some(bar_module),
+            path_to_module(&db, &VfsPath::FileSystem(bar))
+        );
+
+        Ok(())
+    }
+
+    #[test]
+    fn deleting_an_unrealted_file_doesnt_change_module_resolution() -> anyhow::Result<()> {
+        let TestCase { mut db, src, .. } = create_resolver()?;
+
+        let foo_path = src.join("foo.py");
+        let bar_path = src.join("bar.py");
+
+        db.memory_file_system()
+            .write_files([(&foo_path, "x = 1"), (&bar_path, "y = 2")])?;
+
+        let foo_module_name = ModuleName::new_static("foo").unwrap();
+        let foo_module = resolve_module(&db, foo_module_name.clone()).unwrap();
+
+        let bar = system_path_to_file(&db, &bar_path).expect("bar.py to exist");
+
+        db.clear_salsa_events();
+
+        // Delete `bar.py`
+        db.memory_file_system().remove_file(&bar_path)?;
+        bar.touch(&mut db);
+
+        // Re-query the foo module. The foo module should still be cached because `bar.py` isn't relevant
+        // for resolving `foo`.
+
+        let foo_module2 = resolve_module(&db, foo_module_name);
+
+        assert!(!db
+            .take_salsa_events()
+            .iter()
+            .any(|event| { matches!(event.kind, salsa::EventKind::WillExecute { .. }) }));
+
+        assert_eq!(Some(foo_module), foo_module2);
+
+        Ok(())
+    }
+
+    #[test]
+    fn adding_a_file_on_which_the_module_resolution_depends_on_invalidates_the_query(
+    ) -> anyhow::Result<()> {
+        let TestCase { mut db, src, .. } = create_resolver()?;
+        let foo_path = src.join("foo.py");
+
+        let foo_module_name = ModuleName::new_static("foo").unwrap();
+        assert_eq!(resolve_module(&db, foo_module_name.clone()), None);
+
+        // Now write the foo file
+        db.memory_file_system().write_file(&foo_path, "x = 1")?;
+        VfsFile::touch_path(&mut db, &VfsPath::FileSystem(foo_path.clone()));
+        let foo_file = system_path_to_file(&db, &foo_path).expect("foo.py to exist");
+
+        let foo_module = resolve_module(&db, foo_module_name).expect("Foo module to resolve");
+        assert_eq!(foo_file, foo_module.file());
+
+        Ok(())
+    }
+
+    #[test]
+    fn removing_a_file_that_the_module_resolution_depends_on_invalidates_the_query(
+    ) -> anyhow::Result<()> {
+        let TestCase { mut db, src, .. } = create_resolver()?;
+        let foo_path = src.join("foo.py");
+        let foo_init_path = src.join("foo/__init__.py");
+
+        db.memory_file_system()
+            .write_files([(&foo_path, "x = 1"), (&foo_init_path, "x = 2")])?;
+
+        let foo_module_name = ModuleName::new_static("foo").unwrap();
+        let foo_module = resolve_module(&db, foo_module_name.clone()).expect("foo module to exist");
+
+        assert_eq!(&foo_init_path, foo_module.file().path(&db));
+
+        // Delete `foo/__init__.py` and the `foo` folder. `foo` should now resolve to `foo.py`
+        db.memory_file_system().remove_file(&foo_init_path)?;
+        db.memory_file_system()
+            .remove_directory(foo_init_path.parent().unwrap())?;
+        VfsFile::touch_path(&mut db, &VfsPath::FileSystem(foo_init_path.clone()));
+
+        let foo_module = resolve_module(&db, foo_module_name).expect("Foo module to resolve");
+        assert_eq!(&foo_path, foo_module.file().path(&db));
+
+        Ok(())
+    }
+}

crates/red_knot_module_resolver/src/typeshed.rs

@@ -0,0 +1,91 @@
+pub(crate) mod versions;
+
+#[cfg(test)]
+mod tests {
+    use std::io::{self, Read};
+    use std::path::Path;
+
+    use ruff_db::vendored::VendoredFileSystem;
+    use ruff_db::vfs::VendoredPath;
+
+    // The file path here is hardcoded in this crate's `build.rs` script.
+    // Luckily this crate will fail to build if this file isn't available at build time.
+    const TYPESHED_ZIP_BYTES: &[u8] =
+        include_bytes!(concat!(env!("OUT_DIR"), "/zipped_typeshed.zip"));
+
+    #[test]
+    fn typeshed_zip_created_at_build_time() {
+        let mut typeshed_zip_archive =
+            zip::ZipArchive::new(io::Cursor::new(TYPESHED_ZIP_BYTES)).unwrap();
+
+        let mut functools_module_stub = typeshed_zip_archive
+            .by_name("stdlib/functools.pyi")
+            .unwrap();
+        assert!(functools_module_stub.is_file());
+
+        let mut functools_module_stub_source = String::new();
+        functools_module_stub
+            .read_to_string(&mut functools_module_stub_source)
+            .unwrap();
+
+        assert!(functools_module_stub_source.contains("def update_wrapper("));
+    }
+
+    #[test]
+    fn typeshed_vfs_consistent_with_vendored_stubs() {
+        let vendored_typeshed_dir = Path::new("vendor/typeshed").canonicalize().unwrap();
+        let vendored_typeshed_stubs = VendoredFileSystem::new(TYPESHED_ZIP_BYTES).unwrap();
+
+        let mut empty_iterator = true;
+        for entry in walkdir::WalkDir::new(&vendored_typeshed_dir).min_depth(1) {
+            empty_iterator = false;
+            let entry = entry.unwrap();
+            let absolute_path = entry.path();
+            let file_type = entry.file_type();
+
+            let relative_path = absolute_path
+                .strip_prefix(&vendored_typeshed_dir)
+                .unwrap_or_else(|_| {
+                    panic!("Expected {absolute_path:?} to be a child of {vendored_typeshed_dir:?}")
+                });
+
+            let vendored_path = <&VendoredPath>::try_from(relative_path)
+                .unwrap_or_else(|_| panic!("Expected {relative_path:?} to be valid UTF-8"));
+
+            assert!(
+                vendored_typeshed_stubs.exists(vendored_path),
+                "Expected {vendored_path:?} to exist in the `VendoredFileSystem`!
+
+                Vendored file system:
+
+                {vendored_typeshed_stubs:#?}
+                "
+            );
+
+            let vendored_path_kind = vendored_typeshed_stubs
+                .metadata(vendored_path)
+                .unwrap_or_else(|| {
+                    panic!(
+                        "Expected metadata for {vendored_path:?} to be retrievable from the `VendoredFileSystem!
+
+                        Vendored file system:
+
+                        {vendored_typeshed_stubs:#?}
+                        "
+                    )
+                })
+                .kind();
+
+            assert_eq!(
+                vendored_path_kind.is_directory(),
+                file_type.is_dir(),
+                "{vendored_path:?} had type {vendored_path_kind:?}, inconsistent with fs path {relative_path:?}: {file_type:?}"
+            );
+        }
+
+        assert!(
+            !empty_iterator,
+            "Expected there to be at least one file or directory in the vendored typeshed stubs!"
+        );
+    }
+}

crates/red_knot_module_resolver/src/typeshed/versions.rs

@@ -0,0 +1,590 @@
+use std::collections::BTreeMap;
+use std::fmt;
+use std::num::{NonZeroU16, NonZeroUsize};
+use std::ops::{RangeFrom, RangeInclusive};
+use std::str::FromStr;
+
+use rustc_hash::FxHashMap;
+
+use crate::module::ModuleName;
+
+#[derive(Debug, PartialEq, Eq)]
+pub struct TypeshedVersionsParseError {
+    line_number: NonZeroU16,
+    reason: TypeshedVersionsParseErrorKind,
+}
+
+impl fmt::Display for TypeshedVersionsParseError {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        let TypeshedVersionsParseError {
+            line_number,
+            reason,
+        } = self;
+        write!(
+            f,
+            "Error while parsing line {line_number} of typeshed's VERSIONS file: {reason}"
+        )
+    }
+}
+
+impl std::error::Error for TypeshedVersionsParseError {
+    fn source(&self) -> Option<&(dyn std::error::Error + 'static)> {
+        if let TypeshedVersionsParseErrorKind::IntegerParsingFailure { err, .. } = &self.reason {
+            Some(err)
+        } else {
+            None
+        }
+    }
+}
+
+#[derive(Debug, PartialEq, Eq)]
+pub enum TypeshedVersionsParseErrorKind {
+    TooManyLines(NonZeroUsize),
+    UnexpectedNumberOfColons,
+    InvalidModuleName(String),
+    UnexpectedNumberOfHyphens,
+    UnexpectedNumberOfPeriods(String),
+    IntegerParsingFailure {
+        version: String,
+        err: std::num::ParseIntError,
+    },
+}
+
+impl fmt::Display for TypeshedVersionsParseErrorKind {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            Self::TooManyLines(num_lines) => write!(
+                f,
+                "File has too many lines ({num_lines}); maximum allowed is {}",
+                NonZeroU16::MAX
+            ),
+            Self::UnexpectedNumberOfColons => {
+                f.write_str("Expected every non-comment line to have exactly one colon")
+            }
+            Self::InvalidModuleName(name) => write!(
+                f,
+                "Expected all components of '{name}' to be valid Python identifiers"
+            ),
+            Self::UnexpectedNumberOfHyphens => {
+                f.write_str("Expected every non-comment line to have exactly one '-' character")
+            }
+            Self::UnexpectedNumberOfPeriods(format) => write!(
+                f,
+                "Expected all versions to be in the form {{MAJOR}}.{{MINOR}}; got '{format}'"
+            ),
+            Self::IntegerParsingFailure { version, err } => write!(
+                f,
+                "Failed to convert '{version}' to a pair of integers due to {err}",
+            ),
+        }
+    }
+}
+
+#[derive(Debug, PartialEq, Eq)]
+pub struct TypeshedVersions(FxHashMap<ModuleName, PyVersionRange>);
+
+impl TypeshedVersions {
+    pub fn len(&self) -> usize {
+        self.0.len()
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.0.is_empty()
+    }
+
+    pub fn contains_module(&self, module_name: &ModuleName) -> bool {
+        self.0.contains_key(module_name)
+    }
+
+    pub fn module_exists_on_version(
+        &self,
+        module: ModuleName,
+        version: impl Into<PyVersion>,
+    ) -> bool {
+        let version = version.into();
+        let mut module: Option<ModuleName> = Some(module);
+        while let Some(module_to_try) = module {
+            if let Some(range) = self.0.get(&module_to_try) {
+                return range.contains(version);
+            }
+            module = module_to_try.parent();
+        }
+        false
+    }
+}
+
+impl FromStr for TypeshedVersions {
+    type Err = TypeshedVersionsParseError;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        let mut map = FxHashMap::default();
+
+        for (line_index, line) in s.lines().enumerate() {
+            // humans expect line numbers to be 1-indexed
+            let line_number = NonZeroUsize::new(line_index.saturating_add(1)).unwrap();
+
+            let Ok(line_number) = NonZeroU16::try_from(line_number) else {
+                return Err(TypeshedVersionsParseError {
+                    line_number: NonZeroU16::MAX,
+                    reason: TypeshedVersionsParseErrorKind::TooManyLines(line_number),
+                });
+            };
+
+            let Some(content) = line.split('#').map(str::trim).next() else {
+                continue;
+            };
+            if content.is_empty() {
+                continue;
+            }
+
+            let mut parts = content.split(':').map(str::trim);
+            let (Some(module_name), Some(rest), None) = (parts.next(), parts.next(), parts.next())
+            else {
+                return Err(TypeshedVersionsParseError {
+                    line_number,
+                    reason: TypeshedVersionsParseErrorKind::UnexpectedNumberOfColons,
+                });
+            };
+
+            let Some(module_name) = ModuleName::new(module_name) else {
+                return Err(TypeshedVersionsParseError {
+                    line_number,
+                    reason: TypeshedVersionsParseErrorKind::InvalidModuleName(
+                        module_name.to_string(),
+                    ),
+                });
+            };
+
+            match PyVersionRange::from_str(rest) {
+                Ok(version) => map.insert(module_name, version),
+                Err(reason) => {
+                    return Err(TypeshedVersionsParseError {
+                        line_number,
+                        reason,
+                    })
+                }
+            };
+        }
+
+        Ok(Self(map))
+    }
+}
+
+impl fmt::Display for TypeshedVersions {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        let sorted_items: BTreeMap<&ModuleName, &PyVersionRange> = self.0.iter().collect();
+        for (module_name, range) in sorted_items {
+            writeln!(f, "{module_name}: {range}")?;
+        }
+        Ok(())
+    }
+}
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+enum PyVersionRange {
+    AvailableFrom(RangeFrom<PyVersion>),
+    AvailableWithin(RangeInclusive<PyVersion>),
+}
+
+impl PyVersionRange {
+    fn contains(&self, version: PyVersion) -> bool {
+        match self {
+            Self::AvailableFrom(inner) => inner.contains(&version),
+            Self::AvailableWithin(inner) => inner.contains(&version),
+        }
+    }
+}
+
+impl FromStr for PyVersionRange {
+    type Err = TypeshedVersionsParseErrorKind;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        let mut parts = s.split('-').map(str::trim);
+        match (parts.next(), parts.next(), parts.next()) {
+            (Some(lower), Some(""), None) => Ok(Self::AvailableFrom((lower.parse()?)..)),
+            (Some(lower), Some(upper), None) => {
+                Ok(Self::AvailableWithin((lower.parse()?)..=(upper.parse()?)))
+            }
+            _ => Err(TypeshedVersionsParseErrorKind::UnexpectedNumberOfHyphens),
+        }
+    }
+}
+
+impl fmt::Display for PyVersionRange {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            Self::AvailableFrom(range_from) => write!(f, "{}-", range_from.start),
+            Self::AvailableWithin(range_inclusive) => {
+                write!(f, "{}-{}", range_inclusive.start(), range_inclusive.end())
+            }
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, Eq, PartialEq, Ord, PartialOrd, Hash)]
+pub struct PyVersion {
+    major: u8,
+    minor: u8,
+}
+
+impl FromStr for PyVersion {
+    type Err = TypeshedVersionsParseErrorKind;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        let mut parts = s.split('.').map(str::trim);
+        let (Some(major), Some(minor), None) = (parts.next(), parts.next(), parts.next()) else {
+            return Err(TypeshedVersionsParseErrorKind::UnexpectedNumberOfPeriods(
+                s.to_string(),
+            ));
+        };
+        let major = match u8::from_str(major) {
+            Ok(major) => major,
+            Err(err) => {
+                return Err(TypeshedVersionsParseErrorKind::IntegerParsingFailure {
+                    version: s.to_string(),
+                    err,
+                })
+            }
+        };
+        let minor = match u8::from_str(minor) {
+            Ok(minor) => minor,
+            Err(err) => {
+                return Err(TypeshedVersionsParseErrorKind::IntegerParsingFailure {
+                    version: s.to_string(),
+                    err,
+                })
+            }
+        };
+        Ok(Self { major, minor })
+    }
+}
+
+impl fmt::Display for PyVersion {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        let PyVersion { major, minor } = self;
+        write!(f, "{major}.{minor}")
+    }
+}
+
+// TODO: unify with the PythonVersion enum in the linter/formatter crates?
+#[derive(Copy, Clone, Hash, Debug, PartialEq, Eq, PartialOrd, Ord, Default)]
+pub enum SupportedPyVersion {
+    Py37,
+    #[default]
+    Py38,
+    Py39,
+    Py310,
+    Py311,
+    Py312,
+    Py313,
+}
+
+impl From<SupportedPyVersion> for PyVersion {
+    fn from(value: SupportedPyVersion) -> Self {
+        match value {
+            SupportedPyVersion::Py37 => PyVersion { major: 3, minor: 7 },
+            SupportedPyVersion::Py38 => PyVersion { major: 3, minor: 8 },
+            SupportedPyVersion::Py39 => PyVersion { major: 3, minor: 9 },
+            SupportedPyVersion::Py310 => PyVersion {
+                major: 3,
+                minor: 10,
+            },
+            SupportedPyVersion::Py311 => PyVersion {
+                major: 3,
+                minor: 11,
+            },
+            SupportedPyVersion::Py312 => PyVersion {
+                major: 3,
+                minor: 12,
+            },
+            SupportedPyVersion::Py313 => PyVersion {
+                major: 3,
+                minor: 13,
+            },
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::num::{IntErrorKind, NonZeroU16};
+    use std::path::Path;
+
+    use super::*;
+
+    use insta::assert_snapshot;
+
+    const TYPESHED_STDLIB_DIR: &str = "stdlib";
+
+    #[allow(unsafe_code)]
+    const ONE: NonZeroU16 = unsafe { NonZeroU16::new_unchecked(1) };
+
+    #[test]
+    fn can_parse_vendored_versions_file() {
+        let versions_data = include_str!(concat!(
+            env!("CARGO_MANIFEST_DIR"),
+            "/vendor/typeshed/stdlib/VERSIONS"
+        ));
+
+        let versions = TypeshedVersions::from_str(versions_data).unwrap();
+        assert!(versions.len() > 100);
+        assert!(versions.len() < 1000);
+
+        let asyncio = ModuleName::new_static("asyncio").unwrap();
+        let asyncio_staggered = ModuleName::new_static("asyncio.staggered").unwrap();
+        let audioop = ModuleName::new_static("audioop").unwrap();
+
+        assert!(versions.contains_module(&asyncio));
+        assert!(versions.module_exists_on_version(asyncio, SupportedPyVersion::Py310));
+
+        assert!(versions.contains_module(&asyncio_staggered));
+        assert!(
+            versions.module_exists_on_version(asyncio_staggered.clone(), SupportedPyVersion::Py38)
+        );
+        assert!(!versions.module_exists_on_version(asyncio_staggered, SupportedPyVersion::Py37));
+
+        assert!(versions.contains_module(&audioop));
+        assert!(versions.module_exists_on_version(audioop.clone(), SupportedPyVersion::Py312));
+        assert!(!versions.module_exists_on_version(audioop, SupportedPyVersion::Py313));
+    }
+
+    #[test]
+    fn typeshed_versions_consistent_with_vendored_stubs() {
+        const VERSIONS_DATA: &str = include_str!("../../vendor/typeshed/stdlib/VERSIONS");
+        let vendored_typeshed_dir = Path::new("vendor/typeshed").canonicalize().unwrap();
+        let vendored_typeshed_versions = TypeshedVersions::from_str(VERSIONS_DATA).unwrap();
+
+        let mut empty_iterator = true;
+
+        let stdlib_stubs_path = vendored_typeshed_dir.join(TYPESHED_STDLIB_DIR);
+
+        for entry in std::fs::read_dir(&stdlib_stubs_path).unwrap() {
+            empty_iterator = false;
+            let entry = entry.unwrap();
+            let absolute_path = entry.path();
+
+            let relative_path = absolute_path
+                .strip_prefix(&stdlib_stubs_path)
+                .unwrap_or_else(|_| panic!("Expected path to be a child of {stdlib_stubs_path:?} but found {absolute_path:?}"));
+
+            let relative_path_str = relative_path.as_os_str().to_str().unwrap_or_else(|| {
+                panic!("Expected all typeshed paths to be valid UTF-8; got {relative_path:?}")
+            });
+            if relative_path_str == "VERSIONS" {
+                continue;
+            }
+
+            let top_level_module = if let Some(extension) = relative_path.extension() {
+                // It was a file; strip off the file extension to get the module name:
+                let extension = extension
+                    .to_str()
+                    .unwrap_or_else(||panic!("Expected all file extensions to be UTF-8; was not true for {relative_path:?}"));
+
+                relative_path_str
+                    .strip_suffix(extension)
+                    .and_then(|string| string.strip_suffix('.')).unwrap_or_else(|| {
+                        panic!("Expected path {relative_path_str:?} to end with computed extension {extension:?}")
+                })
+            } else {
+                // It was a directory; no need to do anything to get the module name
+                relative_path_str
+            };
+
+            let top_level_module = ModuleName::new(top_level_module)
+                .unwrap_or_else(|| panic!("{top_level_module:?} was not a valid module name!"));
+
+            assert!(vendored_typeshed_versions.contains_module(&top_level_module));
+        }
+
+        assert!(
+            !empty_iterator,
+            "Expected there to be at least one file or directory in the vendored typeshed stubs"
+        );
+    }
+
+    #[test]
+    fn can_parse_mock_versions_file() {
+        const VERSIONS: &str = "\
+# a comment
+    # some more comment
+# yet more comment
+
+
+# and some more comment
+
+bar: 2.7-3.10
+
+# more comment
+bar.baz: 3.1-3.9
+foo: 3.8-   # trailing comment
+";
+        let parsed_versions = TypeshedVersions::from_str(VERSIONS).unwrap();
+        assert_eq!(parsed_versions.len(), 3);
+        assert_snapshot!(parsed_versions.to_string(), @r###"
+        bar: 2.7-3.10
+        bar.baz: 3.1-3.9
+        foo: 3.8-
+        "###
+        );
+
+        let foo = ModuleName::new_static("foo").unwrap();
+        let bar = ModuleName::new_static("bar").unwrap();
+        let bar_baz = ModuleName::new_static("bar.baz").unwrap();
+        let spam = ModuleName::new_static("spam").unwrap();
+
+        assert!(parsed_versions.contains_module(&foo));
+        assert!(!parsed_versions.module_exists_on_version(foo.clone(), SupportedPyVersion::Py37));
+        assert!(parsed_versions.module_exists_on_version(foo.clone(), SupportedPyVersion::Py38));
+        assert!(parsed_versions.module_exists_on_version(foo, SupportedPyVersion::Py311));
+
+        assert!(parsed_versions.contains_module(&bar));
+        assert!(parsed_versions.module_exists_on_version(bar.clone(), SupportedPyVersion::Py37));
+        assert!(parsed_versions.module_exists_on_version(bar.clone(), SupportedPyVersion::Py310));
+        assert!(!parsed_versions.module_exists_on_version(bar, SupportedPyVersion::Py311));
+
+        assert!(parsed_versions.contains_module(&bar_baz));
+        assert!(parsed_versions.module_exists_on_version(bar_baz.clone(), SupportedPyVersion::Py37));
+        assert!(parsed_versions.module_exists_on_version(bar_baz.clone(), SupportedPyVersion::Py39));
+        assert!(!parsed_versions.module_exists_on_version(bar_baz, SupportedPyVersion::Py310));
+
+        assert!(!parsed_versions.contains_module(&spam));
+        assert!(!parsed_versions.module_exists_on_version(spam.clone(), SupportedPyVersion::Py37));
+        assert!(!parsed_versions.module_exists_on_version(spam, SupportedPyVersion::Py313));
+    }
+
+    #[test]
+    fn invalid_huge_versions_file() {
+        let offset = 100;
+        let too_many = u16::MAX as usize + offset;
+
+        let mut massive_versions_file = String::new();
+        for i in 0..too_many {
+            massive_versions_file.push_str(&format!("x{i}: 3.8-\n"));
+        }
+
+        assert_eq!(
+            TypeshedVersions::from_str(&massive_versions_file),
+            Err(TypeshedVersionsParseError {
+                line_number: NonZeroU16::MAX,
+                reason: TypeshedVersionsParseErrorKind::TooManyLines(
+                    NonZeroUsize::new(too_many + 1 - offset).unwrap()
+                )
+            })
+        );
+    }
+
+    #[test]
+    fn invalid_typeshed_versions_bad_colon_number() {
+        assert_eq!(
+            TypeshedVersions::from_str("foo 3.7"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::UnexpectedNumberOfColons
+            })
+        );
+        assert_eq!(
+            TypeshedVersions::from_str("foo:: 3.7"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::UnexpectedNumberOfColons
+            })
+        );
+    }
+
+    #[test]
+    fn invalid_typeshed_versions_non_identifier_modules() {
+        assert_eq!(
+            TypeshedVersions::from_str("not!an!identifier!: 3.7"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::InvalidModuleName(
+                    "not!an!identifier!".to_string()
+                )
+            })
+        );
+        assert_eq!(
+            TypeshedVersions::from_str("(also_not).(an_identifier): 3.7"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::InvalidModuleName(
+                    "(also_not).(an_identifier)".to_string()
+                )
+            })
+        );
+    }
+
+    #[test]
+    fn invalid_typeshed_versions_bad_hyphen_number() {
+        assert_eq!(
+            TypeshedVersions::from_str("foo: 3.8"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::UnexpectedNumberOfHyphens
+            })
+        );
+        assert_eq!(
+            TypeshedVersions::from_str("foo: 3.8--"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::UnexpectedNumberOfHyphens
+            })
+        );
+        assert_eq!(
+            TypeshedVersions::from_str("foo: 3.8--3.9"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::UnexpectedNumberOfHyphens
+            })
+        );
+    }
+
+    #[test]
+    fn invalid_typeshed_versions_bad_period_number() {
+        assert_eq!(
+            TypeshedVersions::from_str("foo: 38-"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::UnexpectedNumberOfPeriods("38".to_string())
+            })
+        );
+        assert_eq!(
+            TypeshedVersions::from_str("foo: 3..8-"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::UnexpectedNumberOfPeriods(
+                    "3..8".to_string()
+                )
+            })
+        );
+        assert_eq!(
+            TypeshedVersions::from_str("foo: 3.8-3..11"),
+            Err(TypeshedVersionsParseError {
+                line_number: ONE,
+                reason: TypeshedVersionsParseErrorKind::UnexpectedNumberOfPeriods(
+                    "3..11".to_string()
+                )
+            })
+        );
+    }
+
+    #[test]
+    fn invalid_typeshed_versions_non_digits() {
+        let err = TypeshedVersions::from_str("foo: 1.two-").unwrap_err();
+        assert_eq!(err.line_number, ONE);
+        let TypeshedVersionsParseErrorKind::IntegerParsingFailure { version, err } = err.reason
+        else {
+            panic!()
+        };
+        assert_eq!(version, "1.two".to_string());
+        assert_eq!(*err.kind(), IntErrorKind::InvalidDigit);
+
+        let err = TypeshedVersions::from_str("foo: 3.8-four.9").unwrap_err();
+        assert_eq!(err.line_number, ONE);
+        let TypeshedVersionsParseErrorKind::IntegerParsingFailure { version, err } = err.reason
+        else {
+            panic!()
+        };
+        assert_eq!(version, "four.9".to_string());
+        assert_eq!(*err.kind(), IntErrorKind::InvalidDigit);
+    }
+}

crates/red_knot_module_resolver/vendor/typeshed/LICENSE

@@ -0,0 +1,237 @@
+The "typeshed" project is licensed under the terms of the Apache license, as
+reproduced below.
+
+= = = = =
+
+Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "{}"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright {yyyy} {name of copyright owner}
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
+
+= = = = =
+
+Parts of typeshed are licensed under different licenses (like the MIT
+license), reproduced below.
+
+= = = = =
+
+The MIT License
+
+Copyright (c) 2015 Jukka Lehtosalo and contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+the rights to use, copy, modify, merge, publish, distribute, sublicense,
+and/or sell copies of the Software, and to permit persons to whom the
+Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+DEALINGS IN THE SOFTWARE.
+
+= = = = =

crates/red_knot_module_resolver/vendor/typeshed/README.md

@@ -0,0 +1,124 @@
+# typeshed
+
+[![Tests](https://github.com/python/typeshed/actions/workflows/tests.yml/badge.svg)](https://github.com/python/typeshed/actions/workflows/tests.yml)
+[![Chat at https://gitter.im/python/typing](https://badges.gitter.im/python/typing.svg)](https://gitter.im/python/typing?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge)
+[![Pull Requests Welcome](https://img.shields.io/badge/pull%20requests-welcome-brightgreen.svg)](https://github.com/python/typeshed/blob/main/CONTRIBUTING.md)
+
+## About
+
+Typeshed contains external type annotations for the Python standard library
+and Python builtins, as well as third party packages as contributed by
+people external to those projects.
+
+This data can e.g. be used for static analysis, type checking, type inference,
+and autocompletion.
+
+For information on how to use typeshed, read below.  Information for
+contributors can be found in [CONTRIBUTING.md](CONTRIBUTING.md).  **Please read
+it before submitting pull requests; do not report issues with annotations to
+the project the stubs are for, but instead report them here to typeshed.**
+
+Further documentation on stub files, typeshed, and Python's typing system in
+general, can also be found at https://typing.readthedocs.io/en/latest/.
+
+Typeshed supports Python versions 3.8 and up.
+
+## Using
+
+If you're just using a type checker ([mypy](https://github.com/python/mypy/),
+[pyright](https://github.com/microsoft/pyright),
+[pytype](https://github.com/google/pytype/), PyCharm, ...), as opposed to
+developing it, you don't need to interact with the typeshed repo at
+all: a copy of standard library part of typeshed is bundled with type checkers.
+And type stubs for third party packages and modules you are using can
+be installed from PyPI. For example, if you are using `html5lib` and `requests`,
+you can install the type stubs using
+
+```bash
+$ pip install types-html5lib types-requests
+```
+
+These PyPI packages follow [PEP 561](http://www.python.org/dev/peps/pep-0561/)
+and are automatically released (up to once a day) by
+[typeshed internal machinery](https://github.com/typeshed-internal/stub_uploader).
+
+Type checkers should be able to use these stub packages when installed. For more
+details, see the documentation for your type checker.
+
+### Package versioning for third-party stubs
+
+Version numbers of third-party stub packages consist of at least four parts.
+All parts of the stub version, except for the last part, correspond to the
+version of the runtime package being stubbed. For example, if the `types-foo`
+package has version `1.2.0.20240309`, this guarantees that the `types-foo` package
+contains stubs targeted against `foo==1.2.*` and tested against the latest
+version of `foo` matching that specifier. In this example, the final element
+of the version number (20240309) indicates that the stub package was pushed on
+March 9, 2024.
+
+At typeshed, we try to keep breaking changes to a minimum. However, due to the
+nature of stubs, any version bump can introduce changes that might make your
+code fail to type check.
+
+There are several strategies available for specifying the version of a stubs
+package you're using, each with its own tradeoffs:
+
+1. Use the same bounds that you use for the package being stubbed. For example,
+   if you use `requests>=2.30.0,<2.32`, you can use
+   `types-requests>=2.30.0,<2.32`. This ensures that the stubs are compatible
+   with the package you are using, but it carries a small risk of breaking
+   type checking due to changes in the stubs.
+
+   Another risk of this strategy is that stubs often lag behind
+   the package being stubbed. You might want to force the package being stubbed
+   to a certain minimum version because it fixes a critical bug, but if
+   correspondingly updated stubs have not been released, your type
+   checking results may not be fully accurate.
+2. Pin the stubs to a known good version and update the pin from time to time
+   (either manually, or using a tool such as dependabot or renovate).
+
+   For example, if you use `types-requests==2.31.0.1`, you can have confidence
+   that upgrading dependencies will not break type checking. However, you will
+   miss out on improvements in the stubs that could potentially improve type
+   checking until you update the pin. This strategy also has the risk that the
+   stubs you are using might become incompatible with the package being stubbed.
+3. Don't pin the stubs. This is the option that demands the least work from
+   you when it comes to updating version pins, and has the advantage that you
+   will automatically benefit from improved stubs whenever a new version of the
+   stubs package is released. However, it carries the risk that the stubs
+   become incompatible with the package being stubbed.
+
+   For example, if a new major version of the package is released, there's a
+   chance the stubs might be updated to reflect the new version of the runtime
+   package before you update the package being stubbed.
+
+You can also switch between the different strategies as needed. For example,
+you could default to strategy (1), but fall back to strategy (2) when
+a problem arises that can't easily be fixed.
+
+### The `_typeshed` package
+
+typeshed includes a package `_typeshed` as part of the standard library.
+This package and its submodules contain utility types, but are not
+available at runtime. For more information about how to use this package,
+[see the `stdlib/_typeshed` directory](https://github.com/python/typeshed/tree/main/stdlib/_typeshed).
+
+## Discussion
+
+If you've run into behavior in the type checker that suggests the type
+stubs for a given library are incorrect or incomplete,
+we want to hear from you!
+
+Our main forum for discussion is the project's [GitHub issue
+tracker](https://github.com/python/typeshed/issues).  This is the right
+place to start a discussion of any of the above or most any other
+topic concerning the project.
+
+If you have general questions about typing with Python, or you need
+a review of your type annotations or stubs outside of typeshed, head over to
+[our discussion forum](https://github.com/python/typing/discussions).
+For less formal discussion, try the typing chat room on
+[gitter.im](https://gitter.im/python/typing).  Some typeshed maintainers
+are almost always present; feel free to find us there and we're happy
+to chat.  Substantive technical discussion will be directed to the
+issue tracker.

crates/red_knot_module_resolver/vendor/typeshed/source_commit.txt

@@ -0,0 +1 @@
+114409d49b43ba62a179ebb856fa70a5161f751e

crates/red_knot_module_resolver/vendor/typeshed/stdlib/VERSIONS

@@ -0,0 +1,310 @@
+# The structure of this file is as follows:
+# - Blank lines and comments starting with `#` are ignored.
+# - Lines contain the name of a module, followed by a colon,
+#   a space, and a version range (for example: `symbol: 3.0-3.9`).
+#
+# Version ranges may be of the form "X.Y-A.B" or "X.Y-". The
+# first form means that a module was introduced in version X.Y and last
+# available in version A.B. The second form means that the module was
+# introduced in version X.Y and is still available in the latest
+# version of Python.
+#
+# If a submodule is not listed separately, it has the same lifetime as
+# its parent module.
+#
+# Python versions before 3.0 are ignored, so any module that was already
+# present in 3.0 will have "3.0" as its minimum version. Version ranges
+# for unsupported versions of Python 3 are generally accurate but we do
+# not guarantee their correctness.
+
+__future__: 3.0-
+__main__: 3.0-
+_ast: 3.0-
+_bisect: 3.0-
+_bootlocale: 3.4-3.9
+_codecs: 3.0-
+_collections_abc: 3.3-
+_compat_pickle: 3.1-
+_compression: 3.5-
+_csv: 3.0-
+_ctypes: 3.0-
+_curses: 3.0-
+_decimal: 3.3-
+_dummy_thread: 3.0-3.8
+_dummy_threading: 3.0-3.8
+_heapq: 3.0-
+_imp: 3.0-
+_json: 3.0-
+_locale: 3.0-
+_lsprof: 3.0-
+_markupbase: 3.0-
+_msi: 3.0-
+_operator: 3.4-
+_osx_support: 3.0-
+_posixsubprocess: 3.2-
+_py_abc: 3.7-
+_pydecimal: 3.5-
+_random: 3.0-
+_sitebuiltins: 3.4-
+_socket: 3.0-  # present in 3.0 at runtime, but not in typeshed
+_stat: 3.4-
+_thread: 3.0-
+_threading_local: 3.0-
+_tkinter: 3.0-
+_tracemalloc: 3.4-
+_typeshed: 3.0-  # not present at runtime, only for type checking
+_warnings: 3.0-
+_weakref: 3.0-
+_weakrefset: 3.0-
+_winapi: 3.3-
+abc: 3.0-
+aifc: 3.0-3.12
+antigravity: 3.0-
+argparse: 3.0-
+array: 3.0-
+ast: 3.0-
+asynchat: 3.0-3.11
+asyncio: 3.4-
+asyncio.exceptions: 3.8-
+asyncio.format_helpers: 3.7-
+asyncio.mixins: 3.10-
+asyncio.runners: 3.7-
+asyncio.staggered: 3.8-
+asyncio.taskgroups: 3.11-
+asyncio.threads: 3.9-
+asyncio.timeouts: 3.11-
+asyncio.trsock: 3.8-
+asyncore: 3.0-3.11
+atexit: 3.0-
+audioop: 3.0-3.12
+base64: 3.0-
+bdb: 3.0-
+binascii: 3.0-
+binhex: 3.0-3.10
+bisect: 3.0-
+builtins: 3.0-
+bz2: 3.0-
+cProfile: 3.0-
+calendar: 3.0-
+cgi: 3.0-3.12
+cgitb: 3.0-3.12
+chunk: 3.0-3.12
+cmath: 3.0-
+cmd: 3.0-
+code: 3.0-
+codecs: 3.0-
+codeop: 3.0-
+collections: 3.0-
+collections.abc: 3.3-
+colorsys: 3.0-
+compileall: 3.0-
+concurrent: 3.2-
+configparser: 3.0-
+contextlib: 3.0-
+contextvars: 3.7-
+copy: 3.0-
+copyreg: 3.0-
+crypt: 3.0-3.12
+csv: 3.0-
+ctypes: 3.0-
+curses: 3.0-
+dataclasses: 3.7-
+datetime: 3.0-
+dbm: 3.0-
+decimal: 3.0-
+difflib: 3.0-
+dis: 3.0-
+distutils: 3.0-3.11
+distutils.command.bdist_msi: 3.0-3.10
+distutils.command.bdist_wininst: 3.0-3.9
+doctest: 3.0-
+dummy_threading: 3.0-3.8
+email: 3.0-
+encodings: 3.0-
+ensurepip: 3.0-
+enum: 3.4-
+errno: 3.0-
+faulthandler: 3.3-
+fcntl: 3.0-
+filecmp: 3.0-
+fileinput: 3.0-
+fnmatch: 3.0-
+formatter: 3.0-3.9
+fractions: 3.0-
+ftplib: 3.0-
+functools: 3.0-
+gc: 3.0-
+genericpath: 3.0-
+getopt: 3.0-
+getpass: 3.0-
+gettext: 3.0-
+glob: 3.0-
+graphlib: 3.9-
+grp: 3.0-
+gzip: 3.0-
+hashlib: 3.0-
+heapq: 3.0-
+hmac: 3.0-
+html: 3.0-
+http: 3.0-
+imaplib: 3.0-
+imghdr: 3.0-3.12
+imp: 3.0-3.11
+importlib: 3.0-
+importlib._abc: 3.10-
+importlib.metadata: 3.8-
+importlib.metadata._meta: 3.10-
+importlib.readers: 3.10-
+importlib.resources: 3.7-
+importlib.resources.abc: 3.11-
+importlib.resources.readers: 3.11-
+importlib.resources.simple: 3.11-
+importlib.simple: 3.11-
+inspect: 3.0-
+io: 3.0-
+ipaddress: 3.3-
+itertools: 3.0-
+json: 3.0-
+keyword: 3.0-
+lib2to3: 3.0-3.12
+linecache: 3.0-
+locale: 3.0-
+logging: 3.0-
+lzma: 3.3-
+mailbox: 3.0-
+mailcap: 3.0-3.12
+marshal: 3.0-
+math: 3.0-
+mimetypes: 3.0-
+mmap: 3.0-
+modulefinder: 3.0-
+msilib: 3.0-3.12
+msvcrt: 3.0-
+multiprocessing: 3.0-
+multiprocessing.resource_tracker: 3.8-
+multiprocessing.shared_memory: 3.8-
+netrc: 3.0-
+nis: 3.0-3.12
+nntplib: 3.0-3.12
+nt: 3.0-
+ntpath: 3.0-
+nturl2path: 3.0-
+numbers: 3.0-
+opcode: 3.0-
+operator: 3.0-
+optparse: 3.0-
+os: 3.0-
+ossaudiodev: 3.0-3.12
+parser: 3.0-3.9
+pathlib: 3.4-
+pdb: 3.0-
+pickle: 3.0-
+pickletools: 3.0-
+pipes: 3.0-3.12
+pkgutil: 3.0-
+platform: 3.0-
+plistlib: 3.0-
+poplib: 3.0-
+posix: 3.0-
+posixpath: 3.0-
+pprint: 3.0-
+profile: 3.0-
+pstats: 3.0-
+pty: 3.0-
+pwd: 3.0-
+py_compile: 3.0-
+pyclbr: 3.0-
+pydoc: 3.0-
+pydoc_data: 3.0-
+pyexpat: 3.0-
+queue: 3.0-
+quopri: 3.0-
+random: 3.0-
+re: 3.0-
+readline: 3.0-
+reprlib: 3.0-
+resource: 3.0-
+rlcompleter: 3.0-
+runpy: 3.0-
+sched: 3.0-
+secrets: 3.6-
+select: 3.0-
+selectors: 3.4-
+shelve: 3.0-
+shlex: 3.0-
+shutil: 3.0-
+signal: 3.0-
+site: 3.0-
+smtpd: 3.0-3.11
+smtplib: 3.0-
+sndhdr: 3.0-3.12
+socket: 3.0-
+socketserver: 3.0-
+spwd: 3.0-3.12
+sqlite3: 3.0-
+sre_compile: 3.0-
+sre_constants: 3.0-
+sre_parse: 3.0-
+ssl: 3.0-
+stat: 3.0-
+statistics: 3.4-
+string: 3.0-
+stringprep: 3.0-
+struct: 3.0-
+subprocess: 3.0-
+sunau: 3.0-3.12
+symbol: 3.0-3.9
+symtable: 3.0-
+sys: 3.0-
+sys._monitoring: 3.12-  # Doesn't actually exist. See comments in the stub.
+sysconfig: 3.0-
+syslog: 3.0-
+tabnanny: 3.0-
+tarfile: 3.0-
+telnetlib: 3.0-3.12
+tempfile: 3.0-
+termios: 3.0-
+textwrap: 3.0-
+this: 3.0-
+threading: 3.0-
+time: 3.0-
+timeit: 3.0-
+tkinter: 3.0-
+tkinter.tix: 3.0-3.12
+token: 3.0-
+tokenize: 3.0-
+tomllib: 3.11-
+trace: 3.0-
+traceback: 3.0-
+tracemalloc: 3.4-
+tty: 3.0-
+turtle: 3.0-
+types: 3.0-
+typing: 3.5-
+typing_extensions: 3.0-
+unicodedata: 3.0-
+unittest: 3.0-
+unittest._log: 3.9-
+unittest.async_case: 3.8-
+urllib: 3.0-
+uu: 3.0-3.12
+uuid: 3.0-
+venv: 3.3-
+warnings: 3.0-
+wave: 3.0-
+weakref: 3.0-
+webbrowser: 3.0-
+winreg: 3.0-
+winsound: 3.0-
+wsgiref: 3.0-
+wsgiref.types: 3.11-
+xdrlib: 3.0-3.12
+xml: 3.0-
+xmlrpc: 3.0-
+xxlimited: 3.2-
+zipapp: 3.5-
+zipfile: 3.0-
+zipfile._path: 3.12-
+zipimport: 3.0-
+zlib: 3.0-
+zoneinfo: 3.9-

crates/red_knot_module_resolver/vendor/typeshed/stdlib/__future__.pyi

@@ -0,0 +1,36 @@
+from typing_extensions import TypeAlias
+
+_VersionInfo: TypeAlias = tuple[int, int, int, str, int]
+
+class _Feature:
+    def __init__(self, optionalRelease: _VersionInfo, mandatoryRelease: _VersionInfo | None, compiler_flag: int) -> None: ...
+    def getOptionalRelease(self) -> _VersionInfo: ...
+    def getMandatoryRelease(self) -> _VersionInfo | None: ...
+    compiler_flag: int
+
+absolute_import: _Feature
+division: _Feature
+generators: _Feature
+nested_scopes: _Feature
+print_function: _Feature
+unicode_literals: _Feature
+with_statement: _Feature
+barry_as_FLUFL: _Feature
+generator_stop: _Feature
+annotations: _Feature
+
+all_feature_names: list[str]  # undocumented
+
+__all__ = [
+    "all_feature_names",
+    "absolute_import",
+    "division",
+    "generators",
+    "nested_scopes",
+    "print_function",
+    "unicode_literals",
+    "with_statement",
+    "barry_as_FLUFL",
+    "generator_stop",
+    "annotations",
+]

crates/red_knot_module_resolver/vendor/typeshed/stdlib/__main__.pyi

@@ -0,0 +1,3 @@
+from typing import Any
+
+def __getattr__(name: str) -> Any: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_bisect.pyi

@@ -0,0 +1,84 @@
+import sys
+from _typeshed import SupportsLenAndGetItem, SupportsRichComparisonT
+from collections.abc import Callable, MutableSequence
+from typing import TypeVar, overload
+
+_T = TypeVar("_T")
+
+if sys.version_info >= (3, 10):
+    @overload
+    def bisect_left(
+        a: SupportsLenAndGetItem[SupportsRichComparisonT],
+        x: SupportsRichComparisonT,
+        lo: int = 0,
+        hi: int | None = None,
+        *,
+        key: None = None,
+    ) -> int: ...
+    @overload
+    def bisect_left(
+        a: SupportsLenAndGetItem[_T],
+        x: SupportsRichComparisonT,
+        lo: int = 0,
+        hi: int | None = None,
+        *,
+        key: Callable[[_T], SupportsRichComparisonT],
+    ) -> int: ...
+    @overload
+    def bisect_right(
+        a: SupportsLenAndGetItem[SupportsRichComparisonT],
+        x: SupportsRichComparisonT,
+        lo: int = 0,
+        hi: int | None = None,
+        *,
+        key: None = None,
+    ) -> int: ...
+    @overload
+    def bisect_right(
+        a: SupportsLenAndGetItem[_T],
+        x: SupportsRichComparisonT,
+        lo: int = 0,
+        hi: int | None = None,
+        *,
+        key: Callable[[_T], SupportsRichComparisonT],
+    ) -> int: ...
+    @overload
+    def insort_left(
+        a: MutableSequence[SupportsRichComparisonT],
+        x: SupportsRichComparisonT,
+        lo: int = 0,
+        hi: int | None = None,
+        *,
+        key: None = None,
+    ) -> None: ...
+    @overload
+    def insort_left(
+        a: MutableSequence[_T], x: _T, lo: int = 0, hi: int | None = None, *, key: Callable[[_T], SupportsRichComparisonT]
+    ) -> None: ...
+    @overload
+    def insort_right(
+        a: MutableSequence[SupportsRichComparisonT],
+        x: SupportsRichComparisonT,
+        lo: int = 0,
+        hi: int | None = None,
+        *,
+        key: None = None,
+    ) -> None: ...
+    @overload
+    def insort_right(
+        a: MutableSequence[_T], x: _T, lo: int = 0, hi: int | None = None, *, key: Callable[[_T], SupportsRichComparisonT]
+    ) -> None: ...
+
+else:
+    def bisect_left(
+        a: SupportsLenAndGetItem[SupportsRichComparisonT], x: SupportsRichComparisonT, lo: int = 0, hi: int | None = None
+    ) -> int: ...
+    def bisect_right(
+        a: SupportsLenAndGetItem[SupportsRichComparisonT], x: SupportsRichComparisonT, lo: int = 0, hi: int | None = None
+    ) -> int: ...
+    def insort_left(
+        a: MutableSequence[SupportsRichComparisonT], x: SupportsRichComparisonT, lo: int = 0, hi: int | None = None
+    ) -> None: ...
+    def insort_right(
+        a: MutableSequence[SupportsRichComparisonT], x: SupportsRichComparisonT, lo: int = 0, hi: int | None = None
+    ) -> None: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_bootlocale.pyi

@@ -0,0 +1 @@
+def getpreferredencoding(do_setlocale: bool = True) -> str: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_codecs.pyi

@@ -0,0 +1,133 @@
+import codecs
+import sys
+from _typeshed import ReadableBuffer
+from collections.abc import Callable
+from typing import Literal, overload
+from typing_extensions import TypeAlias
+
+# This type is not exposed; it is defined in unicodeobject.c
+class _EncodingMap:
+    def size(self) -> int: ...
+
+_CharMap: TypeAlias = dict[int, int] | _EncodingMap
+_Handler: TypeAlias = Callable[[UnicodeError], tuple[str | bytes, int]]
+_SearchFunction: TypeAlias = Callable[[str], codecs.CodecInfo | None]
+
+def register(search_function: _SearchFunction, /) -> None: ...
+
+if sys.version_info >= (3, 10):
+    def unregister(search_function: _SearchFunction, /) -> None: ...
+
+def register_error(errors: str, handler: _Handler, /) -> None: ...
+def lookup_error(name: str, /) -> _Handler: ...
+
+# The type ignore on `encode` and `decode` is to avoid issues with overlapping overloads, for more details, see #300
+# https://docs.python.org/3/library/codecs.html#binary-transforms
+_BytesToBytesEncoding: TypeAlias = Literal[
+    "base64",
+    "base_64",
+    "base64_codec",
+    "bz2",
+    "bz2_codec",
+    "hex",
+    "hex_codec",
+    "quopri",
+    "quotedprintable",
+    "quoted_printable",
+    "quopri_codec",
+    "uu",
+    "uu_codec",
+    "zip",
+    "zlib",
+    "zlib_codec",
+]
+# https://docs.python.org/3/library/codecs.html#text-transforms
+_StrToStrEncoding: TypeAlias = Literal["rot13", "rot_13"]
+
+@overload
+def encode(obj: ReadableBuffer, encoding: _BytesToBytesEncoding, errors: str = "strict") -> bytes: ...
+@overload
+def encode(obj: str, encoding: _StrToStrEncoding, errors: str = "strict") -> str: ...  # type: ignore[overload-overlap]
+@overload
+def encode(obj: str, encoding: str = "utf-8", errors: str = "strict") -> bytes: ...
+@overload
+def decode(obj: ReadableBuffer, encoding: _BytesToBytesEncoding, errors: str = "strict") -> bytes: ...  # type: ignore[overload-overlap]
+@overload
+def decode(obj: str, encoding: _StrToStrEncoding, errors: str = "strict") -> str: ...
+
+# these are documented as text encodings but in practice they also accept str as input
+@overload
+def decode(
+    obj: str,
+    encoding: Literal["unicode_escape", "unicode-escape", "raw_unicode_escape", "raw-unicode-escape"],
+    errors: str = "strict",
+) -> str: ...
+
+# hex is officially documented as a bytes to bytes encoding, but it appears to also work with str
+@overload
+def decode(obj: str, encoding: Literal["hex", "hex_codec"], errors: str = "strict") -> bytes: ...
+@overload
+def decode(obj: ReadableBuffer, encoding: str = "utf-8", errors: str = "strict") -> str: ...
+def lookup(encoding: str, /) -> codecs.CodecInfo: ...
+def charmap_build(map: str, /) -> _CharMap: ...
+def ascii_decode(data: ReadableBuffer, errors: str | None = None, /) -> tuple[str, int]: ...
+def ascii_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+def charmap_decode(data: ReadableBuffer, errors: str | None = None, mapping: _CharMap | None = None, /) -> tuple[str, int]: ...
+def charmap_encode(str: str, errors: str | None = None, mapping: _CharMap | None = None, /) -> tuple[bytes, int]: ...
+def escape_decode(data: str | ReadableBuffer, errors: str | None = None, /) -> tuple[str, int]: ...
+def escape_encode(data: bytes, errors: str | None = None, /) -> tuple[bytes, int]: ...
+def latin_1_decode(data: ReadableBuffer, errors: str | None = None, /) -> tuple[str, int]: ...
+def latin_1_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+
+if sys.version_info >= (3, 9):
+    def raw_unicode_escape_decode(
+        data: str | ReadableBuffer, errors: str | None = None, final: bool = True, /
+    ) -> tuple[str, int]: ...
+
+else:
+    def raw_unicode_escape_decode(data: str | ReadableBuffer, errors: str | None = None, /) -> tuple[str, int]: ...
+
+def raw_unicode_escape_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+def readbuffer_encode(data: str | ReadableBuffer, errors: str | None = None, /) -> tuple[bytes, int]: ...
+
+if sys.version_info >= (3, 9):
+    def unicode_escape_decode(
+        data: str | ReadableBuffer, errors: str | None = None, final: bool = True, /
+    ) -> tuple[str, int]: ...
+
+else:
+    def unicode_escape_decode(data: str | ReadableBuffer, errors: str | None = None, /) -> tuple[str, int]: ...
+
+def unicode_escape_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+def utf_16_be_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+def utf_16_be_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+def utf_16_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+def utf_16_encode(str: str, errors: str | None = None, byteorder: int = 0, /) -> tuple[bytes, int]: ...
+def utf_16_ex_decode(
+    data: ReadableBuffer, errors: str | None = None, byteorder: int = 0, final: bool = False, /
+) -> tuple[str, int, int]: ...
+def utf_16_le_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+def utf_16_le_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+def utf_32_be_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+def utf_32_be_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+def utf_32_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+def utf_32_encode(str: str, errors: str | None = None, byteorder: int = 0, /) -> tuple[bytes, int]: ...
+def utf_32_ex_decode(
+    data: ReadableBuffer, errors: str | None = None, byteorder: int = 0, final: bool = False, /
+) -> tuple[str, int, int]: ...
+def utf_32_le_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+def utf_32_le_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+def utf_7_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+def utf_7_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+def utf_8_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+def utf_8_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+
+if sys.platform == "win32":
+    def mbcs_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+    def mbcs_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+    def code_page_decode(
+        codepage: int, data: ReadableBuffer, errors: str | None = None, final: bool = False, /
+    ) -> tuple[str, int]: ...
+    def code_page_encode(code_page: int, str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...
+    def oem_decode(data: ReadableBuffer, errors: str | None = None, final: bool = False, /) -> tuple[str, int]: ...
+    def oem_encode(str: str, errors: str | None = None, /) -> tuple[bytes, int]: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_collections_abc.pyi

@@ -0,0 +1,94 @@
+import sys
+from abc import abstractmethod
+from types import MappingProxyType
+from typing import (  # noqa: Y022,Y038,Y057
+    AbstractSet as Set,
+    AsyncGenerator as AsyncGenerator,
+    AsyncIterable as AsyncIterable,
+    AsyncIterator as AsyncIterator,
+    Awaitable as Awaitable,
+    ByteString as ByteString,
+    Callable as Callable,
+    Collection as Collection,
+    Container as Container,
+    Coroutine as Coroutine,
+    Generator as Generator,
+    Generic,
+    Hashable as Hashable,
+    ItemsView as ItemsView,
+    Iterable as Iterable,
+    Iterator as Iterator,
+    KeysView as KeysView,
+    Mapping as Mapping,
+    MappingView as MappingView,
+    MutableMapping as MutableMapping,
+    MutableSequence as MutableSequence,
+    MutableSet as MutableSet,
+    Protocol,
+    Reversible as Reversible,
+    Sequence as Sequence,
+    Sized as Sized,
+    TypeVar,
+    ValuesView as ValuesView,
+    final,
+    runtime_checkable,
+)
+
+__all__ = [
+    "Awaitable",
+    "Coroutine",
+    "AsyncIterable",
+    "AsyncIterator",
+    "AsyncGenerator",
+    "Hashable",
+    "Iterable",
+    "Iterator",
+    "Generator",
+    "Reversible",
+    "Sized",
+    "Container",
+    "Callable",
+    "Collection",
+    "Set",
+    "MutableSet",
+    "Mapping",
+    "MutableMapping",
+    "MappingView",
+    "KeysView",
+    "ItemsView",
+    "ValuesView",
+    "Sequence",
+    "MutableSequence",
+    "ByteString",
+]
+if sys.version_info >= (3, 12):
+    __all__ += ["Buffer"]
+
+_KT_co = TypeVar("_KT_co", covariant=True)  # Key type covariant containers.
+_VT_co = TypeVar("_VT_co", covariant=True)  # Value type covariant containers.
+
+@final
+class dict_keys(KeysView[_KT_co], Generic[_KT_co, _VT_co]):  # undocumented
+    def __eq__(self, value: object, /) -> bool: ...
+    if sys.version_info >= (3, 10):
+        @property
+        def mapping(self) -> MappingProxyType[_KT_co, _VT_co]: ...
+
+@final
+class dict_values(ValuesView[_VT_co], Generic[_KT_co, _VT_co]):  # undocumented
+    if sys.version_info >= (3, 10):
+        @property
+        def mapping(self) -> MappingProxyType[_KT_co, _VT_co]: ...
+
+@final
+class dict_items(ItemsView[_KT_co, _VT_co]):  # undocumented
+    def __eq__(self, value: object, /) -> bool: ...
+    if sys.version_info >= (3, 10):
+        @property
+        def mapping(self) -> MappingProxyType[_KT_co, _VT_co]: ...
+
+if sys.version_info >= (3, 12):
+    @runtime_checkable
+    class Buffer(Protocol):
+        @abstractmethod
+        def __buffer__(self, flags: int, /) -> memoryview: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_compat_pickle.pyi

@@ -0,0 +1,8 @@
+IMPORT_MAPPING: dict[str, str]
+NAME_MAPPING: dict[tuple[str, str], tuple[str, str]]
+PYTHON2_EXCEPTIONS: tuple[str, ...]
+MULTIPROCESSING_EXCEPTIONS: tuple[str, ...]
+REVERSE_IMPORT_MAPPING: dict[str, str]
+REVERSE_NAME_MAPPING: dict[tuple[str, str], tuple[str, str]]
+PYTHON3_OSERROR_EXCEPTIONS: tuple[str, ...]
+PYTHON3_IMPORTERROR_EXCEPTIONS: tuple[str, ...]

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_compression.pyi

@@ -0,0 +1,25 @@
+from _typeshed import WriteableBuffer
+from collections.abc import Callable
+from io import DEFAULT_BUFFER_SIZE, BufferedIOBase, RawIOBase
+from typing import Any, Protocol
+
+BUFFER_SIZE = DEFAULT_BUFFER_SIZE
+
+class _Reader(Protocol):
+    def read(self, n: int, /) -> bytes: ...
+    def seekable(self) -> bool: ...
+    def seek(self, n: int, /) -> Any: ...
+
+class BaseStream(BufferedIOBase): ...
+
+class DecompressReader(RawIOBase):
+    def __init__(
+        self,
+        fp: _Reader,
+        decomp_factory: Callable[..., object],
+        trailing_error: type[Exception] | tuple[type[Exception], ...] = (),
+        **decomp_args: Any,
+    ) -> None: ...
+    def readinto(self, b: WriteableBuffer) -> int: ...
+    def read(self, size: int = -1) -> bytes: ...
+    def seek(self, offset: int, whence: int = 0) -> int: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_csv.pyi

@@ -0,0 +1,90 @@
+import sys
+from _typeshed import SupportsWrite
+from collections.abc import Iterable, Iterator
+from typing import Any, Final, Literal
+from typing_extensions import TypeAlias
+
+__version__: Final[str]
+
+QUOTE_ALL: Literal[1]
+QUOTE_MINIMAL: Literal[0]
+QUOTE_NONE: Literal[3]
+QUOTE_NONNUMERIC: Literal[2]
+if sys.version_info >= (3, 12):
+    QUOTE_STRINGS: Literal[4]
+    QUOTE_NOTNULL: Literal[5]
+
+# Ideally this would be `QUOTE_ALL | QUOTE_MINIMAL | QUOTE_NONE | QUOTE_NONNUMERIC`
+# However, using literals in situations like these can cause false-positives (see #7258)
+_QuotingType: TypeAlias = int
+
+class Error(Exception): ...
+
+class Dialect:
+    delimiter: str
+    quotechar: str | None
+    escapechar: str | None
+    doublequote: bool
+    skipinitialspace: bool
+    lineterminator: str
+    quoting: _QuotingType
+    strict: bool
+    def __init__(self) -> None: ...
+
+_DialectLike: TypeAlias = str | Dialect | type[Dialect]
+
+class _reader(Iterator[list[str]]):
+    @property
+    def dialect(self) -> Dialect: ...
+    line_num: int
+    def __next__(self) -> list[str]: ...
+
+class _writer:
+    @property
+    def dialect(self) -> Dialect: ...
+    def writerow(self, row: Iterable[Any]) -> Any: ...
+    def writerows(self, rows: Iterable[Iterable[Any]]) -> None: ...
+
+def writer(
+    csvfile: SupportsWrite[str],
+    dialect: _DialectLike = "excel",
+    *,
+    delimiter: str = ",",
+    quotechar: str | None = '"',
+    escapechar: str | None = None,
+    doublequote: bool = True,
+    skipinitialspace: bool = False,
+    lineterminator: str = "\r\n",
+    quoting: _QuotingType = 0,
+    strict: bool = False,
+) -> _writer: ...
+def reader(
+    csvfile: Iterable[str],
+    dialect: _DialectLike = "excel",
+    *,
+    delimiter: str = ",",
+    quotechar: str | None = '"',
+    escapechar: str | None = None,
+    doublequote: bool = True,
+    skipinitialspace: bool = False,
+    lineterminator: str = "\r\n",
+    quoting: _QuotingType = 0,
+    strict: bool = False,
+) -> _reader: ...
+def register_dialect(
+    name: str,
+    dialect: type[Dialect] = ...,
+    *,
+    delimiter: str = ",",
+    quotechar: str | None = '"',
+    escapechar: str | None = None,
+    doublequote: bool = True,
+    skipinitialspace: bool = False,
+    lineterminator: str = "\r\n",
+    quoting: _QuotingType = 0,
+    strict: bool = False,
+) -> None: ...
+def unregister_dialect(name: str) -> None: ...
+def get_dialect(name: str) -> Dialect: ...
+def list_dialects() -> list[str]: ...
+def field_size_limit(new_limit: int = ...) -> int: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_ctypes.pyi

@@ -0,0 +1,211 @@
+import sys
+from _typeshed import ReadableBuffer, WriteableBuffer
+from abc import abstractmethod
+from collections.abc import Callable, Iterable, Iterator, Mapping, Sequence
+from ctypes import CDLL, ArgumentError as ArgumentError
+from typing import Any, ClassVar, Generic, TypeVar, overload
+from typing_extensions import Self, TypeAlias
+
+if sys.version_info >= (3, 9):
+    from types import GenericAlias
+
+_T = TypeVar("_T")
+_CT = TypeVar("_CT", bound=_CData)
+
+FUNCFLAG_CDECL: int
+FUNCFLAG_PYTHONAPI: int
+FUNCFLAG_USE_ERRNO: int
+FUNCFLAG_USE_LASTERROR: int
+RTLD_GLOBAL: int
+RTLD_LOCAL: int
+
+if sys.version_info >= (3, 11):
+    CTYPES_MAX_ARGCOUNT: int
+
+if sys.version_info >= (3, 12):
+    SIZEOF_TIME_T: int
+
+if sys.platform == "win32":
+    # Description, Source, HelpFile, HelpContext, scode
+    _COMError_Details: TypeAlias = tuple[str | None, str | None, str | None, int | None, int | None]
+
+    class COMError(Exception):
+        hresult: int
+        text: str | None
+        details: _COMError_Details
+
+        def __init__(self, hresult: int, text: str | None, details: _COMError_Details) -> None: ...
+
+    def CopyComPointer(src: _PointerLike, dst: _PointerLike | _CArgObject) -> int: ...
+
+    FUNCFLAG_HRESULT: int
+    FUNCFLAG_STDCALL: int
+
+    def FormatError(code: int = ...) -> str: ...
+    def get_last_error() -> int: ...
+    def set_last_error(value: int) -> int: ...
+    def LoadLibrary(name: str, load_flags: int = 0, /) -> int: ...
+    def FreeLibrary(handle: int, /) -> None: ...
+
+class _CDataMeta(type):
+    # By default mypy complains about the following two methods, because strictly speaking cls
+    # might not be a Type[_CT]. However this can never actually happen, because the only class that
+    # uses _CDataMeta as its metaclass is _CData. So it's safe to ignore the errors here.
+    def __mul__(cls: type[_CT], other: int) -> type[Array[_CT]]: ...  # type: ignore[misc]
+    def __rmul__(cls: type[_CT], other: int) -> type[Array[_CT]]: ...  # type: ignore[misc]
+
+class _CData(metaclass=_CDataMeta):
+    _b_base_: int
+    _b_needsfree_: bool
+    _objects: Mapping[Any, int] | None
+    # At runtime the following classmethods are available only on classes, not
+    # on instances. This can't be reflected properly in the type system:
+    #
+    # Structure.from_buffer(...)  # valid at runtime
+    # Structure(...).from_buffer(...)  # invalid at runtime
+    #
+
+    @classmethod
+    def from_buffer(cls, source: WriteableBuffer, offset: int = ...) -> Self: ...
+    @classmethod
+    def from_buffer_copy(cls, source: ReadableBuffer, offset: int = ...) -> Self: ...
+    @classmethod
+    def from_address(cls, address: int) -> Self: ...
+    @classmethod
+    def from_param(cls, obj: Any) -> Self | _CArgObject: ...
+    @classmethod
+    def in_dll(cls, library: CDLL, name: str) -> Self: ...
+    def __buffer__(self, flags: int, /) -> memoryview: ...
+    def __release_buffer__(self, buffer: memoryview, /) -> None: ...
+
+class _SimpleCData(_CData, Generic[_T]):
+    value: _T
+    # The TypeVar can be unsolved here,
+    # but we can't use overloads without creating many, many mypy false-positive errors
+    def __init__(self, value: _T = ...) -> None: ...  # pyright: ignore[reportInvalidTypeVarUse]
+
+class _CanCastTo(_CData): ...
+class _PointerLike(_CanCastTo): ...
+
+class _Pointer(_PointerLike, _CData, Generic[_CT]):
+    _type_: type[_CT]
+    contents: _CT
+    @overload
+    def __init__(self) -> None: ...
+    @overload
+    def __init__(self, arg: _CT) -> None: ...
+    @overload
+    def __getitem__(self, key: int, /) -> Any: ...
+    @overload
+    def __getitem__(self, key: slice, /) -> list[Any]: ...
+    def __setitem__(self, key: int, value: Any, /) -> None: ...
+
+def POINTER(type: type[_CT]) -> type[_Pointer[_CT]]: ...
+def pointer(arg: _CT, /) -> _Pointer[_CT]: ...
+
+class _CArgObject: ...
+
+def byref(obj: _CData, offset: int = ...) -> _CArgObject: ...
+
+_ECT: TypeAlias = Callable[[_CData | None, CFuncPtr, tuple[_CData, ...]], _CData]
+_PF: TypeAlias = tuple[int] | tuple[int, str | None] | tuple[int, str | None, Any]
+
+class CFuncPtr(_PointerLike, _CData):
+    restype: type[_CData] | Callable[[int], Any] | None
+    argtypes: Sequence[type[_CData]]
+    errcheck: _ECT
+    # Abstract attribute that must be defined on subclasses
+    _flags_: ClassVar[int]
+    @overload
+    def __init__(self) -> None: ...
+    @overload
+    def __init__(self, address: int, /) -> None: ...
+    @overload
+    def __init__(self, callable: Callable[..., Any], /) -> None: ...
+    @overload
+    def __init__(self, func_spec: tuple[str | int, CDLL], paramflags: tuple[_PF, ...] | None = ..., /) -> None: ...
+    if sys.platform == "win32":
+        @overload
+        def __init__(
+            self, vtbl_index: int, name: str, paramflags: tuple[_PF, ...] | None = ..., iid: _CData | None = ..., /
+        ) -> None: ...
+
+    def __call__(self, *args: Any, **kwargs: Any) -> Any: ...
+
+_GetT = TypeVar("_GetT")
+_SetT = TypeVar("_SetT")
+
+class _CField(Generic[_CT, _GetT, _SetT]):
+    offset: int
+    size: int
+    @overload
+    def __get__(self, instance: None, owner: type[Any] | None, /) -> Self: ...
+    @overload
+    def __get__(self, instance: Any, owner: type[Any] | None, /) -> _GetT: ...
+    def __set__(self, instance: Any, value: _SetT, /) -> None: ...
+
+class _StructUnionMeta(_CDataMeta):
+    _fields_: Sequence[tuple[str, type[_CData]] | tuple[str, type[_CData], int]]
+    _pack_: int
+    _anonymous_: Sequence[str]
+    def __getattr__(self, name: str) -> _CField[Any, Any, Any]: ...
+
+class _StructUnionBase(_CData, metaclass=_StructUnionMeta):
+    def __init__(self, *args: Any, **kw: Any) -> None: ...
+    def __getattr__(self, name: str) -> Any: ...
+    def __setattr__(self, name: str, value: Any) -> None: ...
+
+class Union(_StructUnionBase): ...
+class Structure(_StructUnionBase): ...
+
+class Array(_CData, Generic[_CT]):
+    @property
+    @abstractmethod
+    def _length_(self) -> int: ...
+    @_length_.setter
+    def _length_(self, value: int) -> None: ...
+    @property
+    @abstractmethod
+    def _type_(self) -> type[_CT]: ...
+    @_type_.setter
+    def _type_(self, value: type[_CT]) -> None: ...
+    # Note: only available if _CT == c_char
+    @property
+    def raw(self) -> bytes: ...
+    @raw.setter
+    def raw(self, value: ReadableBuffer) -> None: ...
+    value: Any  # Note: bytes if _CT == c_char, str if _CT == c_wchar, unavailable otherwise
+    # TODO These methods cannot be annotated correctly at the moment.
+    # All of these "Any"s stand for the array's element type, but it's not possible to use _CT
+    # here, because of a special feature of ctypes.
+    # By default, when accessing an element of an Array[_CT], the returned object has type _CT.
+    # However, when _CT is a "simple type" like c_int, ctypes automatically "unboxes" the object
+    # and converts it to the corresponding Python primitive. For example, when accessing an element
+    # of an Array[c_int], a Python int object is returned, not a c_int.
+    # This behavior does *not* apply to subclasses of "simple types".
+    # If MyInt is a subclass of c_int, then accessing an element of an Array[MyInt] returns
+    # a MyInt, not an int.
+    # This special behavior is not easy to model in a stub, so for now all places where
+    # the array element type would belong are annotated with Any instead.
+    def __init__(self, *args: Any) -> None: ...
+    @overload
+    def __getitem__(self, key: int, /) -> Any: ...
+    @overload
+    def __getitem__(self, key: slice, /) -> list[Any]: ...
+    @overload
+    def __setitem__(self, key: int, value: Any, /) -> None: ...
+    @overload
+    def __setitem__(self, key: slice, value: Iterable[Any], /) -> None: ...
+    def __iter__(self) -> Iterator[Any]: ...
+    # Can't inherit from Sized because the metaclass conflict between
+    # Sized and _CData prevents using _CDataMeta.
+    def __len__(self) -> int: ...
+    if sys.version_info >= (3, 9):
+        def __class_getitem__(cls, item: Any, /) -> GenericAlias: ...
+
+def addressof(obj: _CData) -> int: ...
+def alignment(obj_or_type: _CData | type[_CData]) -> int: ...
+def get_errno() -> int: ...
+def resize(obj: _CData, size: int) -> None: ...
+def set_errno(value: int) -> int: ...
+def sizeof(obj_or_type: _CData | type[_CData]) -> int: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_curses.pyi

@@ -0,0 +1,565 @@
+import sys
+from _typeshed import ReadOnlyBuffer, SupportsRead
+from typing import IO, Any, NamedTuple, final, overload
+from typing_extensions import TypeAlias
+
+# NOTE: This module is ordinarily only available on Unix, but the windows-curses
+# package makes it available on Windows as well with the same contents.
+
+# Handled by PyCurses_ConvertToChtype in _cursesmodule.c.
+_ChType: TypeAlias = str | bytes | int
+
+# ACS codes are only initialized after initscr is called
+ACS_BBSS: int
+ACS_BLOCK: int
+ACS_BOARD: int
+ACS_BSBS: int
+ACS_BSSB: int
+ACS_BSSS: int
+ACS_BTEE: int
+ACS_BULLET: int
+ACS_CKBOARD: int
+ACS_DARROW: int
+ACS_DEGREE: int
+ACS_DIAMOND: int
+ACS_GEQUAL: int
+ACS_HLINE: int
+ACS_LANTERN: int
+ACS_LARROW: int
+ACS_LEQUAL: int
+ACS_LLCORNER: int
+ACS_LRCORNER: int
+ACS_LTEE: int
+ACS_NEQUAL: int
+ACS_PI: int
+ACS_PLMINUS: int
+ACS_PLUS: int
+ACS_RARROW: int
+ACS_RTEE: int
+ACS_S1: int
+ACS_S3: int
+ACS_S7: int
+ACS_S9: int
+ACS_SBBS: int
+ACS_SBSB: int
+ACS_SBSS: int
+ACS_SSBB: int
+ACS_SSBS: int
+ACS_SSSB: int
+ACS_SSSS: int
+ACS_STERLING: int
+ACS_TTEE: int
+ACS_UARROW: int
+ACS_ULCORNER: int
+ACS_URCORNER: int
+ACS_VLINE: int
+ALL_MOUSE_EVENTS: int
+A_ALTCHARSET: int
+A_ATTRIBUTES: int
+A_BLINK: int
+A_BOLD: int
+A_CHARTEXT: int
+A_COLOR: int
+A_DIM: int
+A_HORIZONTAL: int
+A_INVIS: int
+A_ITALIC: int
+A_LEFT: int
+A_LOW: int
+A_NORMAL: int
+A_PROTECT: int
+A_REVERSE: int
+A_RIGHT: int
+A_STANDOUT: int
+A_TOP: int
+A_UNDERLINE: int
+A_VERTICAL: int
+BUTTON1_CLICKED: int
+BUTTON1_DOUBLE_CLICKED: int
+BUTTON1_PRESSED: int
+BUTTON1_RELEASED: int
+BUTTON1_TRIPLE_CLICKED: int
+BUTTON2_CLICKED: int
+BUTTON2_DOUBLE_CLICKED: int
+BUTTON2_PRESSED: int
+BUTTON2_RELEASED: int
+BUTTON2_TRIPLE_CLICKED: int
+BUTTON3_CLICKED: int
+BUTTON3_DOUBLE_CLICKED: int
+BUTTON3_PRESSED: int
+BUTTON3_RELEASED: int
+BUTTON3_TRIPLE_CLICKED: int
+BUTTON4_CLICKED: int
+BUTTON4_DOUBLE_CLICKED: int
+BUTTON4_PRESSED: int
+BUTTON4_RELEASED: int
+BUTTON4_TRIPLE_CLICKED: int
+# Darwin ncurses doesn't provide BUTTON5_* constants
+if sys.version_info >= (3, 10) and sys.platform != "darwin":
+    BUTTON5_PRESSED: int
+    BUTTON5_RELEASED: int
+    BUTTON5_CLICKED: int
+    BUTTON5_DOUBLE_CLICKED: int
+    BUTTON5_TRIPLE_CLICKED: int
+BUTTON_ALT: int
+BUTTON_CTRL: int
+BUTTON_SHIFT: int
+COLOR_BLACK: int
+COLOR_BLUE: int
+COLOR_CYAN: int
+COLOR_GREEN: int
+COLOR_MAGENTA: int
+COLOR_RED: int
+COLOR_WHITE: int
+COLOR_YELLOW: int
+ERR: int
+KEY_A1: int
+KEY_A3: int
+KEY_B2: int
+KEY_BACKSPACE: int
+KEY_BEG: int
+KEY_BREAK: int
+KEY_BTAB: int
+KEY_C1: int
+KEY_C3: int
+KEY_CANCEL: int
+KEY_CATAB: int
+KEY_CLEAR: int
+KEY_CLOSE: int
+KEY_COMMAND: int
+KEY_COPY: int
+KEY_CREATE: int
+KEY_CTAB: int
+KEY_DC: int
+KEY_DL: int
+KEY_DOWN: int
+KEY_EIC: int
+KEY_END: int
+KEY_ENTER: int
+KEY_EOL: int
+KEY_EOS: int
+KEY_EXIT: int
+KEY_F0: int
+KEY_F1: int
+KEY_F10: int
+KEY_F11: int
+KEY_F12: int
+KEY_F13: int
+KEY_F14: int
+KEY_F15: int
+KEY_F16: int
+KEY_F17: int
+KEY_F18: int
+KEY_F19: int
+KEY_F2: int
+KEY_F20: int
+KEY_F21: int
+KEY_F22: int
+KEY_F23: int
+KEY_F24: int
+KEY_F25: int
+KEY_F26: int
+KEY_F27: int
+KEY_F28: int
+KEY_F29: int
+KEY_F3: int
+KEY_F30: int
+KEY_F31: int
+KEY_F32: int
+KEY_F33: int
+KEY_F34: int
+KEY_F35: int
+KEY_F36: int
+KEY_F37: int
+KEY_F38: int
+KEY_F39: int
+KEY_F4: int
+KEY_F40: int
+KEY_F41: int
+KEY_F42: int
+KEY_F43: int
+KEY_F44: int
+KEY_F45: int
+KEY_F46: int
+KEY_F47: int
+KEY_F48: int
+KEY_F49: int
+KEY_F5: int
+KEY_F50: int
+KEY_F51: int
+KEY_F52: int
+KEY_F53: int
+KEY_F54: int
+KEY_F55: int
+KEY_F56: int
+KEY_F57: int
+KEY_F58: int
+KEY_F59: int
+KEY_F6: int
+KEY_F60: int
+KEY_F61: int
+KEY_F62: int
+KEY_F63: int
+KEY_F7: int
+KEY_F8: int
+KEY_F9: int
+KEY_FIND: int
+KEY_HELP: int
+KEY_HOME: int
+KEY_IC: int
+KEY_IL: int
+KEY_LEFT: int
+KEY_LL: int
+KEY_MARK: int
+KEY_MAX: int
+KEY_MESSAGE: int
+KEY_MIN: int
+KEY_MOUSE: int
+KEY_MOVE: int
+KEY_NEXT: int
+KEY_NPAGE: int
+KEY_OPEN: int
+KEY_OPTIONS: int
+KEY_PPAGE: int
+KEY_PREVIOUS: int
+KEY_PRINT: int
+KEY_REDO: int
+KEY_REFERENCE: int
+KEY_REFRESH: int
+KEY_REPLACE: int
+KEY_RESET: int
+KEY_RESIZE: int
+KEY_RESTART: int
+KEY_RESUME: int
+KEY_RIGHT: int
+KEY_SAVE: int
+KEY_SBEG: int
+KEY_SCANCEL: int
+KEY_SCOMMAND: int
+KEY_SCOPY: int
+KEY_SCREATE: int
+KEY_SDC: int
+KEY_SDL: int
+KEY_SELECT: int
+KEY_SEND: int
+KEY_SEOL: int
+KEY_SEXIT: int
+KEY_SF: int
+KEY_SFIND: int
+KEY_SHELP: int
+KEY_SHOME: int
+KEY_SIC: int
+KEY_SLEFT: int
+KEY_SMESSAGE: int
+KEY_SMOVE: int
+KEY_SNEXT: int
+KEY_SOPTIONS: int
+KEY_SPREVIOUS: int
+KEY_SPRINT: int
+KEY_SR: int
+KEY_SREDO: int
+KEY_SREPLACE: int
+KEY_SRESET: int
+KEY_SRIGHT: int
+KEY_SRSUME: int
+KEY_SSAVE: int
+KEY_SSUSPEND: int
+KEY_STAB: int
+KEY_SUNDO: int
+KEY_SUSPEND: int
+KEY_UNDO: int
+KEY_UP: int
+OK: int
+REPORT_MOUSE_POSITION: int
+_C_API: Any
+version: bytes
+
+def baudrate() -> int: ...
+def beep() -> None: ...
+def can_change_color() -> bool: ...
+def cbreak(flag: bool = True, /) -> None: ...
+def color_content(color_number: int, /) -> tuple[int, int, int]: ...
+def color_pair(pair_number: int, /) -> int: ...
+def curs_set(visibility: int, /) -> int: ...
+def def_prog_mode() -> None: ...
+def def_shell_mode() -> None: ...
+def delay_output(ms: int, /) -> None: ...
+def doupdate() -> None: ...
+def echo(flag: bool = True, /) -> None: ...
+def endwin() -> None: ...
+def erasechar() -> bytes: ...
+def filter() -> None: ...
+def flash() -> None: ...
+def flushinp() -> None: ...
+
+if sys.version_info >= (3, 9):
+    def get_escdelay() -> int: ...
+    def get_tabsize() -> int: ...
+
+def getmouse() -> tuple[int, int, int, int, int]: ...
+def getsyx() -> tuple[int, int]: ...
+def getwin(file: SupportsRead[bytes], /) -> _CursesWindow: ...
+def halfdelay(tenths: int, /) -> None: ...
+def has_colors() -> bool: ...
+
+if sys.version_info >= (3, 10):
+    def has_extended_color_support() -> bool: ...
+
+def has_ic() -> bool: ...
+def has_il() -> bool: ...
+def has_key(key: int, /) -> bool: ...
+def init_color(color_number: int, r: int, g: int, b: int, /) -> None: ...
+def init_pair(pair_number: int, fg: int, bg: int, /) -> None: ...
+def initscr() -> _CursesWindow: ...
+def intrflush(flag: bool, /) -> None: ...
+def is_term_resized(nlines: int, ncols: int, /) -> bool: ...
+def isendwin() -> bool: ...
+def keyname(key: int, /) -> bytes: ...
+def killchar() -> bytes: ...
+def longname() -> bytes: ...
+def meta(yes: bool, /) -> None: ...
+def mouseinterval(interval: int, /) -> None: ...
+def mousemask(newmask: int, /) -> tuple[int, int]: ...
+def napms(ms: int, /) -> int: ...
+def newpad(nlines: int, ncols: int, /) -> _CursesWindow: ...
+def newwin(nlines: int, ncols: int, begin_y: int = ..., begin_x: int = ..., /) -> _CursesWindow: ...
+def nl(flag: bool = True, /) -> None: ...
+def nocbreak() -> None: ...
+def noecho() -> None: ...
+def nonl() -> None: ...
+def noqiflush() -> None: ...
+def noraw() -> None: ...
+def pair_content(pair_number: int, /) -> tuple[int, int]: ...
+def pair_number(attr: int, /) -> int: ...
+def putp(string: ReadOnlyBuffer, /) -> None: ...
+def qiflush(flag: bool = True, /) -> None: ...
+def raw(flag: bool = True, /) -> None: ...
+def reset_prog_mode() -> None: ...
+def reset_shell_mode() -> None: ...
+def resetty() -> None: ...
+def resize_term(nlines: int, ncols: int, /) -> None: ...
+def resizeterm(nlines: int, ncols: int, /) -> None: ...
+def savetty() -> None: ...
+
+if sys.version_info >= (3, 9):
+    def set_escdelay(ms: int, /) -> None: ...
+    def set_tabsize(size: int, /) -> None: ...
+
+def setsyx(y: int, x: int, /) -> None: ...
+def setupterm(term: str | None = None, fd: int = -1) -> None: ...
+def start_color() -> None: ...
+def termattrs() -> int: ...
+def termname() -> bytes: ...
+def tigetflag(capname: str, /) -> int: ...
+def tigetnum(capname: str, /) -> int: ...
+def tigetstr(capname: str, /) -> bytes | None: ...
+def tparm(
+    str: ReadOnlyBuffer,
+    i1: int = 0,
+    i2: int = 0,
+    i3: int = 0,
+    i4: int = 0,
+    i5: int = 0,
+    i6: int = 0,
+    i7: int = 0,
+    i8: int = 0,
+    i9: int = 0,
+    /,
+) -> bytes: ...
+def typeahead(fd: int, /) -> None: ...
+def unctrl(ch: _ChType, /) -> bytes: ...
+
+if sys.version_info < (3, 12) or sys.platform != "darwin":
+    # The support for macos was dropped in 3.12
+    def unget_wch(ch: int | str, /) -> None: ...
+
+def ungetch(ch: _ChType, /) -> None: ...
+def ungetmouse(id: int, x: int, y: int, z: int, bstate: int, /) -> None: ...
+def update_lines_cols() -> None: ...
+def use_default_colors() -> None: ...
+def use_env(flag: bool, /) -> None: ...
+
+class error(Exception): ...
+
+@final
+class _CursesWindow:
+    encoding: str
+    @overload
+    def addch(self, ch: _ChType, attr: int = ...) -> None: ...
+    @overload
+    def addch(self, y: int, x: int, ch: _ChType, attr: int = ...) -> None: ...
+    @overload
+    def addnstr(self, str: str, n: int, attr: int = ...) -> None: ...
+    @overload
+    def addnstr(self, y: int, x: int, str: str, n: int, attr: int = ...) -> None: ...
+    @overload
+    def addstr(self, str: str, attr: int = ...) -> None: ...
+    @overload
+    def addstr(self, y: int, x: int, str: str, attr: int = ...) -> None: ...
+    def attroff(self, attr: int, /) -> None: ...
+    def attron(self, attr: int, /) -> None: ...
+    def attrset(self, attr: int, /) -> None: ...
+    def bkgd(self, ch: _ChType, attr: int = ..., /) -> None: ...
+    def bkgdset(self, ch: _ChType, attr: int = ..., /) -> None: ...
+    def border(
+        self,
+        ls: _ChType = ...,
+        rs: _ChType = ...,
+        ts: _ChType = ...,
+        bs: _ChType = ...,
+        tl: _ChType = ...,
+        tr: _ChType = ...,
+        bl: _ChType = ...,
+        br: _ChType = ...,
+    ) -> None: ...
+    @overload
+    def box(self) -> None: ...
+    @overload
+    def box(self, vertch: _ChType = ..., horch: _ChType = ...) -> None: ...
+    @overload
+    def chgat(self, attr: int) -> None: ...
+    @overload
+    def chgat(self, num: int, attr: int) -> None: ...
+    @overload
+    def chgat(self, y: int, x: int, attr: int) -> None: ...
+    @overload
+    def chgat(self, y: int, x: int, num: int, attr: int) -> None: ...
+    def clear(self) -> None: ...
+    def clearok(self, yes: int) -> None: ...
+    def clrtobot(self) -> None: ...
+    def clrtoeol(self) -> None: ...
+    def cursyncup(self) -> None: ...
+    @overload
+    def delch(self) -> None: ...
+    @overload
+    def delch(self, y: int, x: int) -> None: ...
+    def deleteln(self) -> None: ...
+    @overload
+    def derwin(self, begin_y: int, begin_x: int) -> _CursesWindow: ...
+    @overload
+    def derwin(self, nlines: int, ncols: int, begin_y: int, begin_x: int) -> _CursesWindow: ...
+    def echochar(self, ch: _ChType, attr: int = ..., /) -> None: ...
+    def enclose(self, y: int, x: int, /) -> bool: ...
+    def erase(self) -> None: ...
+    def getbegyx(self) -> tuple[int, int]: ...
+    def getbkgd(self) -> tuple[int, int]: ...
+    @overload
+    def getch(self) -> int: ...
+    @overload
+    def getch(self, y: int, x: int) -> int: ...
+    if sys.version_info < (3, 12) or sys.platform != "darwin":
+        # The support for macos was dropped in 3.12
+        @overload
+        def get_wch(self) -> int | str: ...
+        @overload
+        def get_wch(self, y: int, x: int) -> int | str: ...
+
+    @overload
+    def getkey(self) -> str: ...
+    @overload
+    def getkey(self, y: int, x: int) -> str: ...
+    def getmaxyx(self) -> tuple[int, int]: ...
+    def getparyx(self) -> tuple[int, int]: ...
+    @overload
+    def getstr(self) -> bytes: ...
+    @overload
+    def getstr(self, n: int) -> bytes: ...
+    @overload
+    def getstr(self, y: int, x: int) -> bytes: ...
+    @overload
+    def getstr(self, y: int, x: int, n: int) -> bytes: ...
+    def getyx(self) -> tuple[int, int]: ...
+    @overload
+    def hline(self, ch: _ChType, n: int) -> None: ...
+    @overload
+    def hline(self, y: int, x: int, ch: _ChType, n: int) -> None: ...
+    def idcok(self, flag: bool) -> None: ...
+    def idlok(self, yes: bool) -> None: ...
+    def immedok(self, flag: bool) -> None: ...
+    @overload
+    def inch(self) -> int: ...
+    @overload
+    def inch(self, y: int, x: int) -> int: ...
+    @overload
+    def insch(self, ch: _ChType, attr: int = ...) -> None: ...
+    @overload
+    def insch(self, y: int, x: int, ch: _ChType, attr: int = ...) -> None: ...
+    def insdelln(self, nlines: int) -> None: ...
+    def insertln(self) -> None: ...
+    @overload
+    def insnstr(self, str: str, n: int, attr: int = ...) -> None: ...
+    @overload
+    def insnstr(self, y: int, x: int, str: str, n: int, attr: int = ...) -> None: ...
+    @overload
+    def insstr(self, str: str, attr: int = ...) -> None: ...
+    @overload
+    def insstr(self, y: int, x: int, str: str, attr: int = ...) -> None: ...
+    @overload
+    def instr(self, n: int = ...) -> bytes: ...
+    @overload
+    def instr(self, y: int, x: int, n: int = ...) -> bytes: ...
+    def is_linetouched(self, line: int, /) -> bool: ...
+    def is_wintouched(self) -> bool: ...
+    def keypad(self, yes: bool) -> None: ...
+    def leaveok(self, yes: bool) -> None: ...
+    def move(self, new_y: int, new_x: int) -> None: ...
+    def mvderwin(self, y: int, x: int) -> None: ...
+    def mvwin(self, new_y: int, new_x: int) -> None: ...
+    def nodelay(self, yes: bool) -> None: ...
+    def notimeout(self, yes: bool) -> None: ...
+    @overload
+    def noutrefresh(self) -> None: ...
+    @overload
+    def noutrefresh(self, pminrow: int, pmincol: int, sminrow: int, smincol: int, smaxrow: int, smaxcol: int) -> None: ...
+    @overload
+    def overlay(self, destwin: _CursesWindow) -> None: ...
+    @overload
+    def overlay(
+        self, destwin: _CursesWindow, sminrow: int, smincol: int, dminrow: int, dmincol: int, dmaxrow: int, dmaxcol: int
+    ) -> None: ...
+    @overload
+    def overwrite(self, destwin: _CursesWindow) -> None: ...
+    @overload
+    def overwrite(
+        self, destwin: _CursesWindow, sminrow: int, smincol: int, dminrow: int, dmincol: int, dmaxrow: int, dmaxcol: int
+    ) -> None: ...
+    def putwin(self, file: IO[Any], /) -> None: ...
+    def redrawln(self, beg: int, num: int, /) -> None: ...
+    def redrawwin(self) -> None: ...
+    @overload
+    def refresh(self) -> None: ...
+    @overload
+    def refresh(self, pminrow: int, pmincol: int, sminrow: int, smincol: int, smaxrow: int, smaxcol: int) -> None: ...
+    def resize(self, nlines: int, ncols: int) -> None: ...
+    def scroll(self, lines: int = ...) -> None: ...
+    def scrollok(self, flag: bool) -> None: ...
+    def setscrreg(self, top: int, bottom: int, /) -> None: ...
+    def standend(self) -> None: ...
+    def standout(self) -> None: ...
+    @overload
+    def subpad(self, begin_y: int, begin_x: int) -> _CursesWindow: ...
+    @overload
+    def subpad(self, nlines: int, ncols: int, begin_y: int, begin_x: int) -> _CursesWindow: ...
+    @overload
+    def subwin(self, begin_y: int, begin_x: int) -> _CursesWindow: ...
+    @overload
+    def subwin(self, nlines: int, ncols: int, begin_y: int, begin_x: int) -> _CursesWindow: ...
+    def syncdown(self) -> None: ...
+    def syncok(self, flag: bool) -> None: ...
+    def syncup(self) -> None: ...
+    def timeout(self, delay: int) -> None: ...
+    def touchline(self, start: int, count: int, changed: bool = ...) -> None: ...
+    def touchwin(self) -> None: ...
+    def untouchwin(self) -> None: ...
+    @overload
+    def vline(self, ch: _ChType, n: int) -> None: ...
+    @overload
+    def vline(self, y: int, x: int, ch: _ChType, n: int) -> None: ...
+
+class _ncurses_version(NamedTuple):
+    major: int
+    minor: int
+    patch: int
+
+ncurses_version: _ncurses_version
+window = _CursesWindow  # undocumented

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_decimal.pyi

@@ -0,0 +1,281 @@
+import numbers
+import sys
+from collections.abc import Container, Sequence
+from types import TracebackType
+from typing import Any, ClassVar, Final, Literal, NamedTuple, overload
+from typing_extensions import Self, TypeAlias
+
+_Decimal: TypeAlias = Decimal | int
+_DecimalNew: TypeAlias = Decimal | float | str | tuple[int, Sequence[int], int]
+_ComparableNum: TypeAlias = Decimal | float | numbers.Rational
+
+__version__: Final[str]
+__libmpdec_version__: Final[str]
+
+class DecimalTuple(NamedTuple):
+    sign: int
+    digits: tuple[int, ...]
+    exponent: int | Literal["n", "N", "F"]
+
+ROUND_DOWN: str
+ROUND_HALF_UP: str
+ROUND_HALF_EVEN: str
+ROUND_CEILING: str
+ROUND_FLOOR: str
+ROUND_UP: str
+ROUND_HALF_DOWN: str
+ROUND_05UP: str
+HAVE_CONTEXTVAR: bool
+HAVE_THREADS: bool
+MAX_EMAX: int
+MAX_PREC: int
+MIN_EMIN: int
+MIN_ETINY: int
+
+class DecimalException(ArithmeticError): ...
+class Clamped(DecimalException): ...
+class InvalidOperation(DecimalException): ...
+class ConversionSyntax(InvalidOperation): ...
+class DivisionByZero(DecimalException, ZeroDivisionError): ...
+class DivisionImpossible(InvalidOperation): ...
+class DivisionUndefined(InvalidOperation, ZeroDivisionError): ...
+class Inexact(DecimalException): ...
+class InvalidContext(InvalidOperation): ...
+class Rounded(DecimalException): ...
+class Subnormal(DecimalException): ...
+class Overflow(Inexact, Rounded): ...
+class Underflow(Inexact, Rounded, Subnormal): ...
+class FloatOperation(DecimalException, TypeError): ...
+
+def setcontext(context: Context, /) -> None: ...
+def getcontext() -> Context: ...
+
+if sys.version_info >= (3, 11):
+    def localcontext(
+        ctx: Context | None = None,
+        *,
+        prec: int | None = ...,
+        rounding: str | None = ...,
+        Emin: int | None = ...,
+        Emax: int | None = ...,
+        capitals: int | None = ...,
+        clamp: int | None = ...,
+        traps: dict[_TrapType, bool] | None = ...,
+        flags: dict[_TrapType, bool] | None = ...,
+    ) -> _ContextManager: ...
+
+else:
+    def localcontext(ctx: Context | None = None) -> _ContextManager: ...
+
+class Decimal:
+    def __new__(cls, value: _DecimalNew = ..., context: Context | None = ...) -> Self: ...
+    @classmethod
+    def from_float(cls, f: float, /) -> Self: ...
+    def __bool__(self) -> bool: ...
+    def compare(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def __hash__(self) -> int: ...
+    def as_tuple(self) -> DecimalTuple: ...
+    def as_integer_ratio(self) -> tuple[int, int]: ...
+    def to_eng_string(self, context: Context | None = None) -> str: ...
+    def __abs__(self) -> Decimal: ...
+    def __add__(self, value: _Decimal, /) -> Decimal: ...
+    def __divmod__(self, value: _Decimal, /) -> tuple[Decimal, Decimal]: ...
+    def __eq__(self, value: object, /) -> bool: ...
+    def __floordiv__(self, value: _Decimal, /) -> Decimal: ...
+    def __ge__(self, value: _ComparableNum, /) -> bool: ...
+    def __gt__(self, value: _ComparableNum, /) -> bool: ...
+    def __le__(self, value: _ComparableNum, /) -> bool: ...
+    def __lt__(self, value: _ComparableNum, /) -> bool: ...
+    def __mod__(self, value: _Decimal, /) -> Decimal: ...
+    def __mul__(self, value: _Decimal, /) -> Decimal: ...
+    def __neg__(self) -> Decimal: ...
+    def __pos__(self) -> Decimal: ...
+    def __pow__(self, value: _Decimal, mod: _Decimal | None = None, /) -> Decimal: ...
+    def __radd__(self, value: _Decimal, /) -> Decimal: ...
+    def __rdivmod__(self, value: _Decimal, /) -> tuple[Decimal, Decimal]: ...
+    def __rfloordiv__(self, value: _Decimal, /) -> Decimal: ...
+    def __rmod__(self, value: _Decimal, /) -> Decimal: ...
+    def __rmul__(self, value: _Decimal, /) -> Decimal: ...
+    def __rsub__(self, value: _Decimal, /) -> Decimal: ...
+    def __rtruediv__(self, value: _Decimal, /) -> Decimal: ...
+    def __sub__(self, value: _Decimal, /) -> Decimal: ...
+    def __truediv__(self, value: _Decimal, /) -> Decimal: ...
+    def remainder_near(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def __float__(self) -> float: ...
+    def __int__(self) -> int: ...
+    def __trunc__(self) -> int: ...
+    @property
+    def real(self) -> Decimal: ...
+    @property
+    def imag(self) -> Decimal: ...
+    def conjugate(self) -> Decimal: ...
+    def __complex__(self) -> complex: ...
+    @overload
+    def __round__(self) -> int: ...
+    @overload
+    def __round__(self, ndigits: int, /) -> Decimal: ...
+    def __floor__(self) -> int: ...
+    def __ceil__(self) -> int: ...
+    def fma(self, other: _Decimal, third: _Decimal, context: Context | None = None) -> Decimal: ...
+    def __rpow__(self, value: _Decimal, mod: Context | None = None, /) -> Decimal: ...
+    def normalize(self, context: Context | None = None) -> Decimal: ...
+    def quantize(self, exp: _Decimal, rounding: str | None = None, context: Context | None = None) -> Decimal: ...
+    def same_quantum(self, other: _Decimal, context: Context | None = None) -> bool: ...
+    def to_integral_exact(self, rounding: str | None = None, context: Context | None = None) -> Decimal: ...
+    def to_integral_value(self, rounding: str | None = None, context: Context | None = None) -> Decimal: ...
+    def to_integral(self, rounding: str | None = None, context: Context | None = None) -> Decimal: ...
+    def sqrt(self, context: Context | None = None) -> Decimal: ...
+    def max(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def min(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def adjusted(self) -> int: ...
+    def canonical(self) -> Decimal: ...
+    def compare_signal(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def compare_total(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def compare_total_mag(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def copy_abs(self) -> Decimal: ...
+    def copy_negate(self) -> Decimal: ...
+    def copy_sign(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def exp(self, context: Context | None = None) -> Decimal: ...
+    def is_canonical(self) -> bool: ...
+    def is_finite(self) -> bool: ...
+    def is_infinite(self) -> bool: ...
+    def is_nan(self) -> bool: ...
+    def is_normal(self, context: Context | None = None) -> bool: ...
+    def is_qnan(self) -> bool: ...
+    def is_signed(self) -> bool: ...
+    def is_snan(self) -> bool: ...
+    def is_subnormal(self, context: Context | None = None) -> bool: ...
+    def is_zero(self) -> bool: ...
+    def ln(self, context: Context | None = None) -> Decimal: ...
+    def log10(self, context: Context | None = None) -> Decimal: ...
+    def logb(self, context: Context | None = None) -> Decimal: ...
+    def logical_and(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def logical_invert(self, context: Context | None = None) -> Decimal: ...
+    def logical_or(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def logical_xor(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def max_mag(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def min_mag(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def next_minus(self, context: Context | None = None) -> Decimal: ...
+    def next_plus(self, context: Context | None = None) -> Decimal: ...
+    def next_toward(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def number_class(self, context: Context | None = None) -> str: ...
+    def radix(self) -> Decimal: ...
+    def rotate(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def scaleb(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def shift(self, other: _Decimal, context: Context | None = None) -> Decimal: ...
+    def __reduce__(self) -> tuple[type[Self], tuple[str]]: ...
+    def __copy__(self) -> Self: ...
+    def __deepcopy__(self, memo: Any, /) -> Self: ...
+    def __format__(self, specifier: str, context: Context | None = ..., /) -> str: ...
+
+class _ContextManager:
+    new_context: Context
+    saved_context: Context
+    def __init__(self, new_context: Context) -> None: ...
+    def __enter__(self) -> Context: ...
+    def __exit__(self, t: type[BaseException] | None, v: BaseException | None, tb: TracebackType | None) -> None: ...
+
+_TrapType: TypeAlias = type[DecimalException]
+
+class Context:
+    # TODO: Context doesn't allow you to delete *any* attributes from instances of the class at runtime,
+    # even settable attributes like `prec` and `rounding`,
+    # but that's inexpressable in the stub.
+    # Type checkers either ignore it or misinterpret it
+    # if you add a `def __delattr__(self, name: str, /) -> NoReturn` method to the stub
+    prec: int
+    rounding: str
+    Emin: int
+    Emax: int
+    capitals: int
+    clamp: int
+    traps: dict[_TrapType, bool]
+    flags: dict[_TrapType, bool]
+    def __init__(
+        self,
+        prec: int | None = ...,
+        rounding: str | None = ...,
+        Emin: int | None = ...,
+        Emax: int | None = ...,
+        capitals: int | None = ...,
+        clamp: int | None = ...,
+        flags: None | dict[_TrapType, bool] | Container[_TrapType] = ...,
+        traps: None | dict[_TrapType, bool] | Container[_TrapType] = ...,
+        _ignored_flags: list[_TrapType] | None = ...,
+    ) -> None: ...
+    def __reduce__(self) -> tuple[type[Self], tuple[Any, ...]]: ...
+    def clear_flags(self) -> None: ...
+    def clear_traps(self) -> None: ...
+    def copy(self) -> Context: ...
+    def __copy__(self) -> Context: ...
+    # see https://github.com/python/cpython/issues/94107
+    __hash__: ClassVar[None]  # type: ignore[assignment]
+    def Etiny(self) -> int: ...
+    def Etop(self) -> int: ...
+    def create_decimal(self, num: _DecimalNew = "0", /) -> Decimal: ...
+    def create_decimal_from_float(self, f: float, /) -> Decimal: ...
+    def abs(self, x: _Decimal, /) -> Decimal: ...
+    def add(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def canonical(self, x: Decimal, /) -> Decimal: ...
+    def compare(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def compare_signal(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def compare_total(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def compare_total_mag(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def copy_abs(self, x: _Decimal, /) -> Decimal: ...
+    def copy_decimal(self, x: _Decimal, /) -> Decimal: ...
+    def copy_negate(self, x: _Decimal, /) -> Decimal: ...
+    def copy_sign(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def divide(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def divide_int(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def divmod(self, x: _Decimal, y: _Decimal, /) -> tuple[Decimal, Decimal]: ...
+    def exp(self, x: _Decimal, /) -> Decimal: ...
+    def fma(self, x: _Decimal, y: _Decimal, z: _Decimal, /) -> Decimal: ...
+    def is_canonical(self, x: _Decimal, /) -> bool: ...
+    def is_finite(self, x: _Decimal, /) -> bool: ...
+    def is_infinite(self, x: _Decimal, /) -> bool: ...
+    def is_nan(self, x: _Decimal, /) -> bool: ...
+    def is_normal(self, x: _Decimal, /) -> bool: ...
+    def is_qnan(self, x: _Decimal, /) -> bool: ...
+    def is_signed(self, x: _Decimal, /) -> bool: ...
+    def is_snan(self, x: _Decimal, /) -> bool: ...
+    def is_subnormal(self, x: _Decimal, /) -> bool: ...
+    def is_zero(self, x: _Decimal, /) -> bool: ...
+    def ln(self, x: _Decimal, /) -> Decimal: ...
+    def log10(self, x: _Decimal, /) -> Decimal: ...
+    def logb(self, x: _Decimal, /) -> Decimal: ...
+    def logical_and(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def logical_invert(self, x: _Decimal, /) -> Decimal: ...
+    def logical_or(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def logical_xor(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def max(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def max_mag(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def min(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def min_mag(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def minus(self, x: _Decimal, /) -> Decimal: ...
+    def multiply(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def next_minus(self, x: _Decimal, /) -> Decimal: ...
+    def next_plus(self, x: _Decimal, /) -> Decimal: ...
+    def next_toward(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def normalize(self, x: _Decimal, /) -> Decimal: ...
+    def number_class(self, x: _Decimal, /) -> str: ...
+    def plus(self, x: _Decimal, /) -> Decimal: ...
+    def power(self, a: _Decimal, b: _Decimal, modulo: _Decimal | None = None) -> Decimal: ...
+    def quantize(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def radix(self) -> Decimal: ...
+    def remainder(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def remainder_near(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def rotate(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def same_quantum(self, x: _Decimal, y: _Decimal, /) -> bool: ...
+    def scaleb(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def shift(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def sqrt(self, x: _Decimal, /) -> Decimal: ...
+    def subtract(self, x: _Decimal, y: _Decimal, /) -> Decimal: ...
+    def to_eng_string(self, x: _Decimal, /) -> str: ...
+    def to_sci_string(self, x: _Decimal, /) -> str: ...
+    def to_integral_exact(self, x: _Decimal, /) -> Decimal: ...
+    def to_integral_value(self, x: _Decimal, /) -> Decimal: ...
+    def to_integral(self, x: _Decimal, /) -> Decimal: ...
+
+DefaultContext: Context
+BasicContext: Context
+ExtendedContext: Context

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_dummy_thread.pyi

@@ -0,0 +1,33 @@
+from collections.abc import Callable
+from types import TracebackType
+from typing import Any, NoReturn, overload
+from typing_extensions import TypeVarTuple, Unpack
+
+__all__ = ["error", "start_new_thread", "exit", "get_ident", "allocate_lock", "interrupt_main", "LockType", "RLock"]
+
+_Ts = TypeVarTuple("_Ts")
+
+TIMEOUT_MAX: int
+error = RuntimeError
+
+@overload
+def start_new_thread(function: Callable[[Unpack[_Ts]], object], args: tuple[Unpack[_Ts]]) -> None: ...
+@overload
+def start_new_thread(function: Callable[..., object], args: tuple[Any, ...], kwargs: dict[str, Any]) -> None: ...
+def exit() -> NoReturn: ...
+def get_ident() -> int: ...
+def allocate_lock() -> LockType: ...
+def stack_size(size: int | None = None) -> int: ...
+
+class LockType:
+    locked_status: bool
+    def acquire(self, waitflag: bool | None = None, timeout: int = -1) -> bool: ...
+    def __enter__(self, waitflag: bool | None = None, timeout: int = -1) -> bool: ...
+    def __exit__(self, typ: type[BaseException] | None, val: BaseException | None, tb: TracebackType | None) -> None: ...
+    def release(self) -> bool: ...
+    def locked(self) -> bool: ...
+
+class RLock(LockType):
+    def release(self) -> None: ...  # type: ignore[override]
+
+def interrupt_main() -> None: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_dummy_threading.pyi

@@ -0,0 +1,164 @@
+import sys
+from _thread import _excepthook, _ExceptHookArgs
+from _typeshed import ProfileFunction, TraceFunction
+from collections.abc import Callable, Iterable, Mapping
+from types import TracebackType
+from typing import Any, TypeVar
+
+_T = TypeVar("_T")
+
+__all__ = [
+    "get_ident",
+    "active_count",
+    "Condition",
+    "current_thread",
+    "enumerate",
+    "main_thread",
+    "TIMEOUT_MAX",
+    "Event",
+    "Lock",
+    "RLock",
+    "Semaphore",
+    "BoundedSemaphore",
+    "Thread",
+    "Barrier",
+    "BrokenBarrierError",
+    "Timer",
+    "ThreadError",
+    "setprofile",
+    "settrace",
+    "local",
+    "stack_size",
+    "ExceptHookArgs",
+    "excepthook",
+]
+
+def active_count() -> int: ...
+def current_thread() -> Thread: ...
+def currentThread() -> Thread: ...
+def get_ident() -> int: ...
+def enumerate() -> list[Thread]: ...
+def main_thread() -> Thread: ...
+def settrace(func: TraceFunction) -> None: ...
+def setprofile(func: ProfileFunction | None) -> None: ...
+def stack_size(size: int | None = None) -> int: ...
+
+TIMEOUT_MAX: float
+
+class ThreadError(Exception): ...
+
+class local:
+    def __getattribute__(self, name: str) -> Any: ...
+    def __setattr__(self, name: str, value: Any) -> None: ...
+    def __delattr__(self, name: str) -> None: ...
+
+class Thread:
+    name: str
+    daemon: bool
+    @property
+    def ident(self) -> int | None: ...
+    def __init__(
+        self,
+        group: None = None,
+        target: Callable[..., object] | None = None,
+        name: str | None = None,
+        args: Iterable[Any] = (),
+        kwargs: Mapping[str, Any] | None = None,
+        *,
+        daemon: bool | None = None,
+    ) -> None: ...
+    def start(self) -> None: ...
+    def run(self) -> None: ...
+    def join(self, timeout: float | None = None) -> None: ...
+    def getName(self) -> str: ...
+    def setName(self, name: str) -> None: ...
+    @property
+    def native_id(self) -> int | None: ...  # only available on some platforms
+    def is_alive(self) -> bool: ...
+    if sys.version_info < (3, 9):
+        def isAlive(self) -> bool: ...
+
+    def isDaemon(self) -> bool: ...
+    def setDaemon(self, daemonic: bool) -> None: ...
+
+class _DummyThread(Thread): ...
+
+class Lock:
+    def __enter__(self) -> bool: ...
+    def __exit__(
+        self, exc_type: type[BaseException] | None, exc_val: BaseException | None, exc_tb: TracebackType | None
+    ) -> bool | None: ...
+    def acquire(self, blocking: bool = ..., timeout: float = ...) -> bool: ...
+    def release(self) -> None: ...
+    def locked(self) -> bool: ...
+
+class _RLock:
+    def __enter__(self) -> bool: ...
+    def __exit__(
+        self, exc_type: type[BaseException] | None, exc_val: BaseException | None, exc_tb: TracebackType | None
+    ) -> bool | None: ...
+    def acquire(self, blocking: bool = True, timeout: float = -1) -> bool: ...
+    def release(self) -> None: ...
+
+RLock = _RLock
+
+class Condition:
+    def __init__(self, lock: Lock | _RLock | None = None) -> None: ...
+    def __enter__(self) -> bool: ...
+    def __exit__(
+        self, exc_type: type[BaseException] | None, exc_val: BaseException | None, exc_tb: TracebackType | None
+    ) -> bool | None: ...
+    def acquire(self, blocking: bool = ..., timeout: float = ...) -> bool: ...
+    def release(self) -> None: ...
+    def wait(self, timeout: float | None = None) -> bool: ...
+    def wait_for(self, predicate: Callable[[], _T], timeout: float | None = None) -> _T: ...
+    def notify(self, n: int = 1) -> None: ...
+    def notify_all(self) -> None: ...
+    def notifyAll(self) -> None: ...
+
+class Semaphore:
+    def __init__(self, value: int = 1) -> None: ...
+    def __exit__(
+        self, exc_type: type[BaseException] | None, exc_val: BaseException | None, exc_tb: TracebackType | None
+    ) -> bool | None: ...
+    def acquire(self, blocking: bool = True, timeout: float | None = None) -> bool: ...
+    def __enter__(self, blocking: bool = True, timeout: float | None = None) -> bool: ...
+    if sys.version_info >= (3, 9):
+        def release(self, n: int = ...) -> None: ...
+    else:
+        def release(self) -> None: ...
+
+class BoundedSemaphore(Semaphore): ...
+
+class Event:
+    def is_set(self) -> bool: ...
+    def set(self) -> None: ...
+    def clear(self) -> None: ...
+    def wait(self, timeout: float | None = None) -> bool: ...
+
+excepthook = _excepthook
+ExceptHookArgs = _ExceptHookArgs
+
+class Timer(Thread):
+    def __init__(
+        self,
+        interval: float,
+        function: Callable[..., object],
+        args: Iterable[Any] | None = None,
+        kwargs: Mapping[str, Any] | None = None,
+    ) -> None: ...
+    def cancel(self) -> None: ...
+
+class Barrier:
+    @property
+    def parties(self) -> int: ...
+    @property
+    def n_waiting(self) -> int: ...
+    @property
+    def broken(self) -> bool: ...
+    def __init__(self, parties: int, action: Callable[[], None] | None = None, timeout: float | None = None) -> None: ...
+    def wait(self, timeout: float | None = None) -> int: ...
+    def reset(self) -> None: ...
+    def abort(self) -> None: ...
+
+class BrokenBarrierError(RuntimeError): ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_heapq.pyi

@@ -0,0 +1,11 @@
+from typing import Any, Final, TypeVar
+
+_T = TypeVar("_T")
+
+__about__: Final[str]
+
+def heapify(heap: list[Any], /) -> None: ...
+def heappop(heap: list[_T], /) -> _T: ...
+def heappush(heap: list[_T], item: _T, /) -> None: ...
+def heappushpop(heap: list[_T], item: _T, /) -> _T: ...
+def heapreplace(heap: list[_T], item: _T, /) -> _T: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_imp.pyi

@@ -0,0 +1,28 @@
+import sys
+import types
+from _typeshed import ReadableBuffer
+from importlib.machinery import ModuleSpec
+from typing import Any
+
+check_hash_based_pycs: str
+
+def source_hash(key: int, source: ReadableBuffer) -> bytes: ...
+def create_builtin(spec: ModuleSpec, /) -> types.ModuleType: ...
+def create_dynamic(spec: ModuleSpec, file: Any = None, /) -> types.ModuleType: ...
+def acquire_lock() -> None: ...
+def exec_builtin(mod: types.ModuleType, /) -> int: ...
+def exec_dynamic(mod: types.ModuleType, /) -> int: ...
+def extension_suffixes() -> list[str]: ...
+def init_frozen(name: str, /) -> types.ModuleType: ...
+def is_builtin(name: str, /) -> int: ...
+def is_frozen(name: str, /) -> bool: ...
+def is_frozen_package(name: str, /) -> bool: ...
+def lock_held() -> bool: ...
+def release_lock() -> None: ...
+
+if sys.version_info >= (3, 11):
+    def find_frozen(name: str, /, *, withdata: bool = False) -> tuple[memoryview | None, bool, str | None] | None: ...
+    def get_frozen_object(name: str, data: ReadableBuffer | None = None, /) -> types.CodeType: ...
+
+else:
+    def get_frozen_object(name: str, /) -> types.CodeType: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_json.pyi

@@ -0,0 +1,49 @@
+from collections.abc import Callable
+from typing import Any, final
+
+@final
+class make_encoder:
+    @property
+    def sort_keys(self) -> bool: ...
+    @property
+    def skipkeys(self) -> bool: ...
+    @property
+    def key_separator(self) -> str: ...
+    @property
+    def indent(self) -> int | None: ...
+    @property
+    def markers(self) -> dict[int, Any] | None: ...
+    @property
+    def default(self) -> Callable[[Any], Any]: ...
+    @property
+    def encoder(self) -> Callable[[str], str]: ...
+    @property
+    def item_separator(self) -> str: ...
+    def __init__(
+        self,
+        markers: dict[int, Any] | None,
+        default: Callable[[Any], Any],
+        encoder: Callable[[str], str],
+        indent: int | None,
+        key_separator: str,
+        item_separator: str,
+        sort_keys: bool,
+        skipkeys: bool,
+        allow_nan: bool,
+    ) -> None: ...
+    def __call__(self, obj: object, _current_indent_level: int) -> Any: ...
+
+@final
+class make_scanner:
+    object_hook: Any
+    object_pairs_hook: Any
+    parse_int: Any
+    parse_constant: Any
+    parse_float: Any
+    strict: bool
+    # TODO: 'context' needs the attrs above (ducktype), but not __call__.
+    def __init__(self, context: make_scanner) -> None: ...
+    def __call__(self, string: str, index: int) -> tuple[Any, int]: ...
+
+def encode_basestring_ascii(s: str, /) -> str: ...
+def scanstring(string: str, end: int, strict: bool = ...) -> tuple[str, int]: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_locale.pyi

@@ -0,0 +1,100 @@
+import sys
+from _typeshed import StrPath
+from collections.abc import Mapping
+
+LC_CTYPE: int
+LC_COLLATE: int
+LC_TIME: int
+LC_MONETARY: int
+LC_NUMERIC: int
+LC_ALL: int
+CHAR_MAX: int
+
+def setlocale(category: int, locale: str | None = None, /) -> str: ...
+def localeconv() -> Mapping[str, int | str | list[int]]: ...
+
+if sys.version_info >= (3, 11):
+    def getencoding() -> str: ...
+
+def strcoll(os1: str, os2: str, /) -> int: ...
+def strxfrm(string: str, /) -> str: ...
+
+# native gettext functions
+# https://docs.python.org/3/library/locale.html#access-to-message-catalogs
+# https://github.com/python/cpython/blob/f4c03484da59049eb62a9bf7777b963e2267d187/Modules/_localemodule.c#L626
+if sys.platform != "win32":
+    LC_MESSAGES: int
+
+    ABDAY_1: int
+    ABDAY_2: int
+    ABDAY_3: int
+    ABDAY_4: int
+    ABDAY_5: int
+    ABDAY_6: int
+    ABDAY_7: int
+
+    ABMON_1: int
+    ABMON_2: int
+    ABMON_3: int
+    ABMON_4: int
+    ABMON_5: int
+    ABMON_6: int
+    ABMON_7: int
+    ABMON_8: int
+    ABMON_9: int
+    ABMON_10: int
+    ABMON_11: int
+    ABMON_12: int
+
+    DAY_1: int
+    DAY_2: int
+    DAY_3: int
+    DAY_4: int
+    DAY_5: int
+    DAY_6: int
+    DAY_7: int
+
+    ERA: int
+    ERA_D_T_FMT: int
+    ERA_D_FMT: int
+    ERA_T_FMT: int
+
+    MON_1: int
+    MON_2: int
+    MON_3: int
+    MON_4: int
+    MON_5: int
+    MON_6: int
+    MON_7: int
+    MON_8: int
+    MON_9: int
+    MON_10: int
+    MON_11: int
+    MON_12: int
+
+    CODESET: int
+    D_T_FMT: int
+    D_FMT: int
+    T_FMT: int
+    T_FMT_AMPM: int
+    AM_STR: int
+    PM_STR: int
+
+    RADIXCHAR: int
+    THOUSEP: int
+    YESEXPR: int
+    NOEXPR: int
+    CRNCYSTR: int
+    ALT_DIGITS: int
+
+    def nl_langinfo(key: int, /) -> str: ...
+
+    # This is dependent on `libintl.h` which is a part of `gettext`
+    # system dependency. These functions might be missing.
+    # But, we always say that they are present.
+    def gettext(msg: str, /) -> str: ...
+    def dgettext(domain: str | None, msg: str, /) -> str: ...
+    def dcgettext(domain: str | None, msg: str, category: int, /) -> str: ...
+    def textdomain(domain: str | None, /) -> str: ...
+    def bindtextdomain(domain: str, dir: StrPath | None, /) -> str: ...
+    def bind_textdomain_codeset(domain: str, codeset: str | None, /) -> str | None: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_lsprof.pyi

@@ -0,0 +1,35 @@
+import sys
+from _typeshed import structseq
+from collections.abc import Callable
+from types import CodeType
+from typing import Any, Final, final
+
+class Profiler:
+    def __init__(
+        self, timer: Callable[[], float] | None = None, timeunit: float = 0.0, subcalls: bool = True, builtins: bool = True
+    ) -> None: ...
+    def getstats(self) -> list[profiler_entry]: ...
+    def enable(self, subcalls: bool = True, builtins: bool = True) -> None: ...
+    def disable(self) -> None: ...
+    def clear(self) -> None: ...
+
+@final
+class profiler_entry(structseq[Any], tuple[CodeType | str, int, int, float, float, list[profiler_subentry]]):
+    if sys.version_info >= (3, 10):
+        __match_args__: Final = ("code", "callcount", "reccallcount", "totaltime", "inlinetime", "calls")
+    code: CodeType | str
+    callcount: int
+    reccallcount: int
+    totaltime: float
+    inlinetime: float
+    calls: list[profiler_subentry]
+
+@final
+class profiler_subentry(structseq[Any], tuple[CodeType | str, int, int, float, float]):
+    if sys.version_info >= (3, 10):
+        __match_args__: Final = ("code", "callcount", "reccallcount", "totaltime", "inlinetime")
+    code: CodeType | str
+    callcount: int
+    reccallcount: int
+    totaltime: float
+    inlinetime: float

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_markupbase.pyi

@@ -0,0 +1,16 @@
+import sys
+from typing import Any
+
+class ParserBase:
+    def reset(self) -> None: ...
+    def getpos(self) -> tuple[int, int]: ...
+    def unknown_decl(self, data: str) -> None: ...
+    def parse_comment(self, i: int, report: int = 1) -> int: ...  # undocumented
+    def parse_declaration(self, i: int) -> int: ...  # undocumented
+    def parse_marked_section(self, i: int, report: int = 1) -> int: ...  # undocumented
+    def updatepos(self, i: int, j: int) -> int: ...  # undocumented
+    if sys.version_info < (3, 10):
+        # Removed from ParserBase: https://bugs.python.org/issue31844
+        def error(self, message: str) -> Any: ...  # undocumented
+    lineno: int  # undocumented
+    offset: int  # undocumented

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_msi.pyi

@@ -0,0 +1,92 @@
+import sys
+
+if sys.platform == "win32":
+    class MSIError(Exception): ...
+    # Actual typename View, not exposed by the implementation
+    class _View:
+        def Execute(self, params: _Record | None = ...) -> None: ...
+        def GetColumnInfo(self, kind: int) -> _Record: ...
+        def Fetch(self) -> _Record: ...
+        def Modify(self, mode: int, record: _Record) -> None: ...
+        def Close(self) -> None: ...
+        # Don't exist at runtime
+        __new__: None  # type: ignore[assignment]
+        __init__: None  # type: ignore[assignment]
+
+    # Actual typename SummaryInformation, not exposed by the implementation
+    class _SummaryInformation:
+        def GetProperty(self, field: int) -> int | bytes | None: ...
+        def GetPropertyCount(self) -> int: ...
+        def SetProperty(self, field: int, value: int | str) -> None: ...
+        def Persist(self) -> None: ...
+        # Don't exist at runtime
+        __new__: None  # type: ignore[assignment]
+        __init__: None  # type: ignore[assignment]
+
+    # Actual typename Database, not exposed by the implementation
+    class _Database:
+        def OpenView(self, sql: str) -> _View: ...
+        def Commit(self) -> None: ...
+        def GetSummaryInformation(self, updateCount: int) -> _SummaryInformation: ...
+        def Close(self) -> None: ...
+        # Don't exist at runtime
+        __new__: None  # type: ignore[assignment]
+        __init__: None  # type: ignore[assignment]
+
+    # Actual typename Record, not exposed by the implementation
+    class _Record:
+        def GetFieldCount(self) -> int: ...
+        def GetInteger(self, field: int) -> int: ...
+        def GetString(self, field: int) -> str: ...
+        def SetString(self, field: int, str: str) -> None: ...
+        def SetStream(self, field: int, stream: str) -> None: ...
+        def SetInteger(self, field: int, int: int) -> None: ...
+        def ClearData(self) -> None: ...
+        # Don't exist at runtime
+        __new__: None  # type: ignore[assignment]
+        __init__: None  # type: ignore[assignment]
+
+    def UuidCreate() -> str: ...
+    def FCICreate(cabname: str, files: list[str], /) -> None: ...
+    def OpenDatabase(path: str, persist: int, /) -> _Database: ...
+    def CreateRecord(count: int, /) -> _Record: ...
+
+    MSICOLINFO_NAMES: int
+    MSICOLINFO_TYPES: int
+    MSIDBOPEN_CREATE: int
+    MSIDBOPEN_CREATEDIRECT: int
+    MSIDBOPEN_DIRECT: int
+    MSIDBOPEN_PATCHFILE: int
+    MSIDBOPEN_READONLY: int
+    MSIDBOPEN_TRANSACT: int
+    MSIMODIFY_ASSIGN: int
+    MSIMODIFY_DELETE: int
+    MSIMODIFY_INSERT: int
+    MSIMODIFY_INSERT_TEMPORARY: int
+    MSIMODIFY_MERGE: int
+    MSIMODIFY_REFRESH: int
+    MSIMODIFY_REPLACE: int
+    MSIMODIFY_SEEK: int
+    MSIMODIFY_UPDATE: int
+    MSIMODIFY_VALIDATE: int
+    MSIMODIFY_VALIDATE_DELETE: int
+    MSIMODIFY_VALIDATE_FIELD: int
+    MSIMODIFY_VALIDATE_NEW: int
+
+    PID_APPNAME: int
+    PID_AUTHOR: int
+    PID_CHARCOUNT: int
+    PID_CODEPAGE: int
+    PID_COMMENTS: int
+    PID_CREATE_DTM: int
+    PID_KEYWORDS: int
+    PID_LASTAUTHOR: int
+    PID_LASTPRINTED: int
+    PID_LASTSAVE_DTM: int
+    PID_PAGECOUNT: int
+    PID_REVNUMBER: int
+    PID_SECURITY: int
+    PID_SUBJECT: int
+    PID_TEMPLATE: int
+    PID_TITLE: int
+    PID_WORDCOUNT: int

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_operator.pyi

@@ -0,0 +1,147 @@
+import sys
+from _typeshed import SupportsGetItem
+from collections.abc import Callable, Container, Iterable, MutableMapping, MutableSequence, Sequence
+from typing import Any, AnyStr, Generic, Protocol, SupportsAbs, SupportsIndex, TypeVar, final, overload
+from typing_extensions import ParamSpec, TypeAlias, TypeVarTuple, Unpack
+
+_R = TypeVar("_R")
+_T = TypeVar("_T")
+_T_co = TypeVar("_T_co", covariant=True)
+_T1 = TypeVar("_T1")
+_T2 = TypeVar("_T2")
+_K = TypeVar("_K")
+_V = TypeVar("_V")
+_P = ParamSpec("_P")
+_Ts = TypeVarTuple("_Ts")
+
+# The following protocols return "Any" instead of bool, since the comparison
+# operators can be overloaded to return an arbitrary object. For example,
+# the numpy.array comparison dunders return another numpy.array.
+
+class _SupportsDunderLT(Protocol):
+    def __lt__(self, other: Any, /) -> Any: ...
+
+class _SupportsDunderGT(Protocol):
+    def __gt__(self, other: Any, /) -> Any: ...
+
+class _SupportsDunderLE(Protocol):
+    def __le__(self, other: Any, /) -> Any: ...
+
+class _SupportsDunderGE(Protocol):
+    def __ge__(self, other: Any, /) -> Any: ...
+
+_SupportsComparison: TypeAlias = _SupportsDunderLE | _SupportsDunderGE | _SupportsDunderGT | _SupportsDunderLT
+
+class _SupportsInversion(Protocol[_T_co]):
+    def __invert__(self) -> _T_co: ...
+
+class _SupportsNeg(Protocol[_T_co]):
+    def __neg__(self) -> _T_co: ...
+
+class _SupportsPos(Protocol[_T_co]):
+    def __pos__(self) -> _T_co: ...
+
+# All four comparison functions must have the same signature, or we get false-positive errors
+def lt(a: _SupportsComparison, b: _SupportsComparison, /) -> Any: ...
+def le(a: _SupportsComparison, b: _SupportsComparison, /) -> Any: ...
+def eq(a: object, b: object, /) -> Any: ...
+def ne(a: object, b: object, /) -> Any: ...
+def ge(a: _SupportsComparison, b: _SupportsComparison, /) -> Any: ...
+def gt(a: _SupportsComparison, b: _SupportsComparison, /) -> Any: ...
+def not_(a: object, /) -> bool: ...
+def truth(a: object, /) -> bool: ...
+def is_(a: object, b: object, /) -> bool: ...
+def is_not(a: object, b: object, /) -> bool: ...
+def abs(a: SupportsAbs[_T], /) -> _T: ...
+def add(a: Any, b: Any, /) -> Any: ...
+def and_(a: Any, b: Any, /) -> Any: ...
+def floordiv(a: Any, b: Any, /) -> Any: ...
+def index(a: SupportsIndex, /) -> int: ...
+def inv(a: _SupportsInversion[_T_co], /) -> _T_co: ...
+def invert(a: _SupportsInversion[_T_co], /) -> _T_co: ...
+def lshift(a: Any, b: Any, /) -> Any: ...
+def mod(a: Any, b: Any, /) -> Any: ...
+def mul(a: Any, b: Any, /) -> Any: ...
+def matmul(a: Any, b: Any, /) -> Any: ...
+def neg(a: _SupportsNeg[_T_co], /) -> _T_co: ...
+def or_(a: Any, b: Any, /) -> Any: ...
+def pos(a: _SupportsPos[_T_co], /) -> _T_co: ...
+def pow(a: Any, b: Any, /) -> Any: ...
+def rshift(a: Any, b: Any, /) -> Any: ...
+def sub(a: Any, b: Any, /) -> Any: ...
+def truediv(a: Any, b: Any, /) -> Any: ...
+def xor(a: Any, b: Any, /) -> Any: ...
+def concat(a: Sequence[_T], b: Sequence[_T], /) -> Sequence[_T]: ...
+def contains(a: Container[object], b: object, /) -> bool: ...
+def countOf(a: Iterable[object], b: object, /) -> int: ...
+@overload
+def delitem(a: MutableSequence[Any], b: SupportsIndex, /) -> None: ...
+@overload
+def delitem(a: MutableSequence[Any], b: slice, /) -> None: ...
+@overload
+def delitem(a: MutableMapping[_K, Any], b: _K, /) -> None: ...
+@overload
+def getitem(a: Sequence[_T], b: slice, /) -> Sequence[_T]: ...
+@overload
+def getitem(a: SupportsGetItem[_K, _V], b: _K, /) -> _V: ...
+def indexOf(a: Iterable[_T], b: _T, /) -> int: ...
+@overload
+def setitem(a: MutableSequence[_T], b: SupportsIndex, c: _T, /) -> None: ...
+@overload
+def setitem(a: MutableSequence[_T], b: slice, c: Sequence[_T], /) -> None: ...
+@overload
+def setitem(a: MutableMapping[_K, _V], b: _K, c: _V, /) -> None: ...
+def length_hint(obj: object, default: int = 0, /) -> int: ...
+@final
+class attrgetter(Generic[_T_co]):
+    @overload
+    def __new__(cls, attr: str, /) -> attrgetter[Any]: ...
+    @overload
+    def __new__(cls, attr: str, attr2: str, /) -> attrgetter[tuple[Any, Any]]: ...
+    @overload
+    def __new__(cls, attr: str, attr2: str, attr3: str, /) -> attrgetter[tuple[Any, Any, Any]]: ...
+    @overload
+    def __new__(cls, attr: str, attr2: str, attr3: str, attr4: str, /) -> attrgetter[tuple[Any, Any, Any, Any]]: ...
+    @overload
+    def __new__(cls, attr: str, /, *attrs: str) -> attrgetter[tuple[Any, ...]]: ...
+    def __call__(self, obj: Any, /) -> _T_co: ...
+
+@final
+class itemgetter(Generic[_T_co]):
+    @overload
+    def __new__(cls, item: _T, /) -> itemgetter[_T]: ...
+    @overload
+    def __new__(cls, item1: _T1, item2: _T2, /, *items: Unpack[_Ts]) -> itemgetter[tuple[_T1, _T2, Unpack[_Ts]]]: ...
+    # __key: _KT_contra in SupportsGetItem seems to be causing variance issues, ie:
+    # TypeVar "_KT_contra@SupportsGetItem" is contravariant
+    #   "tuple[int, int]" is incompatible with protocol "SupportsIndex"
+    # preventing [_T_co, ...] instead of [Any, ...]
+    #
+    # A suspected mypy issue prevents using [..., _T] instead of [..., Any] here.
+    # https://github.com/python/mypy/issues/14032
+    def __call__(self, obj: SupportsGetItem[Any, Any]) -> Any: ...
+
+@final
+class methodcaller:
+    def __init__(self, name: str, /, *args: Any, **kwargs: Any) -> None: ...
+    def __call__(self, obj: Any) -> Any: ...
+
+def iadd(a: Any, b: Any, /) -> Any: ...
+def iand(a: Any, b: Any, /) -> Any: ...
+def iconcat(a: Any, b: Any, /) -> Any: ...
+def ifloordiv(a: Any, b: Any, /) -> Any: ...
+def ilshift(a: Any, b: Any, /) -> Any: ...
+def imod(a: Any, b: Any, /) -> Any: ...
+def imul(a: Any, b: Any, /) -> Any: ...
+def imatmul(a: Any, b: Any, /) -> Any: ...
+def ior(a: Any, b: Any, /) -> Any: ...
+def ipow(a: Any, b: Any, /) -> Any: ...
+def irshift(a: Any, b: Any, /) -> Any: ...
+def isub(a: Any, b: Any, /) -> Any: ...
+def itruediv(a: Any, b: Any, /) -> Any: ...
+def ixor(a: Any, b: Any, /) -> Any: ...
+
+if sys.version_info >= (3, 11):
+    def call(obj: Callable[_P, _R], /, *args: _P.args, **kwargs: _P.kwargs) -> _R: ...
+
+def _compare_digest(a: AnyStr, b: AnyStr, /) -> bool: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_osx_support.pyi

@@ -0,0 +1,34 @@
+from collections.abc import Iterable, Sequence
+from typing import TypeVar
+
+_T = TypeVar("_T")
+_K = TypeVar("_K")
+_V = TypeVar("_V")
+
+__all__ = ["compiler_fixup", "customize_config_vars", "customize_compiler", "get_platform_osx"]
+
+_UNIVERSAL_CONFIG_VARS: tuple[str, ...]  # undocumented
+_COMPILER_CONFIG_VARS: tuple[str, ...]  # undocumented
+_INITPRE: str  # undocumented
+
+def _find_executable(executable: str, path: str | None = None) -> str | None: ...  # undocumented
+def _read_output(commandstring: str, capture_stderr: bool = False) -> str | None: ...  # undocumented
+def _find_build_tool(toolname: str) -> str: ...  # undocumented
+
+_SYSTEM_VERSION: str | None  # undocumented
+
+def _get_system_version() -> str: ...  # undocumented
+def _remove_original_values(_config_vars: dict[str, str]) -> None: ...  # undocumented
+def _save_modified_value(_config_vars: dict[str, str], cv: str, newvalue: str) -> None: ...  # undocumented
+def _supports_universal_builds() -> bool: ...  # undocumented
+def _find_appropriate_compiler(_config_vars: dict[str, str]) -> dict[str, str]: ...  # undocumented
+def _remove_universal_flags(_config_vars: dict[str, str]) -> dict[str, str]: ...  # undocumented
+def _remove_unsupported_archs(_config_vars: dict[str, str]) -> dict[str, str]: ...  # undocumented
+def _override_all_archs(_config_vars: dict[str, str]) -> dict[str, str]: ...  # undocumented
+def _check_for_unavailable_sdk(_config_vars: dict[str, str]) -> dict[str, str]: ...  # undocumented
+def compiler_fixup(compiler_so: Iterable[str], cc_args: Sequence[str]) -> list[str]: ...
+def customize_config_vars(_config_vars: dict[str, str]) -> dict[str, str]: ...
+def customize_compiler(_config_vars: dict[str, str]) -> dict[str, str]: ...
+def get_platform_osx(
+    _config_vars: dict[str, str], osname: _T, release: _K, machine: _V
+) -> tuple[str | _T, str | _K, str | _V]: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_posixsubprocess.pyi

@@ -0,0 +1,33 @@
+import sys
+from _typeshed import StrOrBytesPath
+from collections.abc import Callable, Sequence
+from typing import SupportsIndex
+
+if sys.platform != "win32":
+    def cloexec_pipe() -> tuple[int, int]: ...
+    def fork_exec(
+        args: Sequence[StrOrBytesPath] | None,
+        executable_list: Sequence[bytes],
+        close_fds: bool,
+        pass_fds: tuple[int, ...],
+        cwd: str,
+        env: Sequence[bytes] | None,
+        p2cread: int,
+        p2cwrite: int,
+        c2pread: int,
+        c2pwrite: int,
+        errread: int,
+        errwrite: int,
+        errpipe_read: int,
+        errpipe_write: int,
+        restore_signals: int,
+        call_setsid: int,
+        pgid_to_set: int,
+        gid: SupportsIndex | None,
+        extra_groups: list[int] | None,
+        uid: SupportsIndex | None,
+        child_umask: int,
+        preexec_fn: Callable[[], None],
+        allow_vfork: bool,
+        /,
+    ) -> int: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_py_abc.pyi

@@ -0,0 +1,14 @@
+import _typeshed
+from typing import Any, NewType, TypeVar
+
+_T = TypeVar("_T")
+
+_CacheToken = NewType("_CacheToken", int)
+
+def get_cache_token() -> _CacheToken: ...
+
+class ABCMeta(type):
+    def __new__(
+        mcls: type[_typeshed.Self], name: str, bases: tuple[type[Any], ...], namespace: dict[str, Any], /
+    ) -> _typeshed.Self: ...
+    def register(cls, subclass: type[_T]) -> type[_T]: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_pydecimal.pyi

@@ -0,0 +1,43 @@
+# This is a slight lie, the implementations aren't exactly identical
+# However, in all likelihood, the differences are inconsequential
+from _decimal import *
+
+__all__ = [
+    "Decimal",
+    "Context",
+    "DecimalTuple",
+    "DefaultContext",
+    "BasicContext",
+    "ExtendedContext",
+    "DecimalException",
+    "Clamped",
+    "InvalidOperation",
+    "DivisionByZero",
+    "Inexact",
+    "Rounded",
+    "Subnormal",
+    "Overflow",
+    "Underflow",
+    "FloatOperation",
+    "DivisionImpossible",
+    "InvalidContext",
+    "ConversionSyntax",
+    "DivisionUndefined",
+    "ROUND_DOWN",
+    "ROUND_HALF_UP",
+    "ROUND_HALF_EVEN",
+    "ROUND_CEILING",
+    "ROUND_FLOOR",
+    "ROUND_UP",
+    "ROUND_HALF_DOWN",
+    "ROUND_05UP",
+    "setcontext",
+    "getcontext",
+    "localcontext",
+    "MAX_PREC",
+    "MAX_EMAX",
+    "MIN_EMIN",
+    "MIN_ETINY",
+    "HAVE_THREADS",
+    "HAVE_CONTEXTVAR",
+]

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_random.pyi

@@ -0,0 +1,12 @@
+from typing_extensions import TypeAlias
+
+# Actually Tuple[(int,) * 625]
+_State: TypeAlias = tuple[int, ...]
+
+class Random:
+    def __init__(self, seed: object = ...) -> None: ...
+    def seed(self, n: object = None, /) -> None: ...
+    def getstate(self) -> _State: ...
+    def setstate(self, state: _State, /) -> None: ...
+    def random(self) -> float: ...
+    def getrandbits(self, k: int, /) -> int: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_sitebuiltins.pyi

@@ -0,0 +1,16 @@
+from collections.abc import Iterable
+from typing import ClassVar, Literal, NoReturn
+
+class Quitter:
+    name: str
+    eof: str
+    def __init__(self, name: str, eof: str) -> None: ...
+    def __call__(self, code: int | None = None) -> NoReturn: ...
+
+class _Printer:
+    MAXLINES: ClassVar[Literal[23]]
+    def __init__(self, name: str, data: str, files: Iterable[str] = (), dirs: Iterable[str] = ()) -> None: ...
+    def __call__(self) -> None: ...
+
+class _Helper:
+    def __call__(self, request: object) -> None: ...

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_socket.pyi

@@ -0,0 +1,803 @@
+import sys
+from _typeshed import ReadableBuffer, WriteableBuffer
+from collections.abc import Iterable
+from typing import Any, SupportsIndex, overload
+from typing_extensions import TypeAlias
+
+_CMSG: TypeAlias = tuple[int, int, bytes]
+_CMSGArg: TypeAlias = tuple[int, int, ReadableBuffer]
+
+# Addresses can be either tuples of varying lengths (AF_INET, AF_INET6,
+# AF_NETLINK, AF_TIPC) or strings/buffers (AF_UNIX).
+# See getsockaddrarg() in socketmodule.c.
+_Address: TypeAlias = tuple[Any, ...] | str | ReadableBuffer
+_RetAddress: TypeAlias = Any
+
+# ===== Constants =====
+# This matches the order in the CPython documentation
+# https://docs.python.org/3/library/socket.html#constants
+
+if sys.platform != "win32":
+    AF_UNIX: int
+
+AF_INET: int
+AF_INET6: int
+
+AF_UNSPEC: int
+
+SOCK_STREAM: int
+SOCK_DGRAM: int
+SOCK_RAW: int
+SOCK_RDM: int
+SOCK_SEQPACKET: int
+
+if sys.platform == "linux":
+    # Availability: Linux >= 2.6.27
+    SOCK_CLOEXEC: int
+    SOCK_NONBLOCK: int
+
+# --------------------
+# Many constants of these forms, documented in the Unix documentation on
+# sockets and/or the IP protocol, are also defined in the socket module.
+# SO_*
+# socket.SOMAXCONN
+# MSG_*
+# SOL_*
+# SCM_*
+# IPPROTO_*
+# IPPORT_*
+# INADDR_*
+# IP_*
+# IPV6_*
+# EAI_*
+# AI_*
+# NI_*
+# TCP_*
+# --------------------
+
+SO_ACCEPTCONN: int
+SO_BROADCAST: int
+SO_DEBUG: int
+SO_DONTROUTE: int
+SO_ERROR: int
+SO_KEEPALIVE: int
+SO_LINGER: int
+SO_OOBINLINE: int
+SO_RCVBUF: int
+SO_RCVLOWAT: int
+SO_RCVTIMEO: int
+SO_REUSEADDR: int
+SO_SNDBUF: int
+SO_SNDLOWAT: int
+SO_SNDTIMEO: int
+SO_TYPE: int
+SO_USELOOPBACK: int
+if sys.platform == "win32":
+    SO_EXCLUSIVEADDRUSE: int
+if sys.platform != "win32":
+    SO_REUSEPORT: int
+if sys.platform != "win32" and sys.platform != "darwin":
+    SO_BINDTODEVICE: int
+    SO_DOMAIN: int
+    SO_MARK: int
+    SO_PASSCRED: int
+    SO_PASSSEC: int
+    SO_PEERCRED: int
+    SO_PEERSEC: int
+    SO_PRIORITY: int
+    SO_PROTOCOL: int
+    SO_SETFIB: int
+
+SOMAXCONN: int
+
+MSG_CTRUNC: int
+MSG_DONTROUTE: int
+MSG_OOB: int
+MSG_PEEK: int
+MSG_TRUNC: int
+MSG_WAITALL: int
+if sys.platform != "win32":
+    MSG_DONTWAIT: int
+    MSG_EOF: int
+    MSG_EOR: int
+    MSG_NOSIGNAL: int  # Sometimes this exists on darwin, sometimes not
+if sys.platform != "darwin":
+    MSG_BCAST: int
+    MSG_ERRQUEUE: int
+    MSG_MCAST: int
+if sys.platform != "win32" and sys.platform != "darwin":
+    MSG_BTAG: int
+    MSG_CMSG_CLOEXEC: int
+    MSG_CONFIRM: int
+    MSG_ETAG: int
+    MSG_FASTOPEN: int
+    MSG_MORE: int
+    MSG_NOTIFICATION: int
+
+SOL_IP: int
+SOL_SOCKET: int
+SOL_TCP: int
+SOL_UDP: int
+if sys.platform != "win32" and sys.platform != "darwin":
+    SOL_ATALK: int
+    SOL_AX25: int
+    SOL_HCI: int
+    SOL_IPX: int
+    SOL_NETROM: int
+    SOL_ROSE: int
+
+if sys.platform != "win32":
+    SCM_CREDS: int
+    SCM_RIGHTS: int
+if sys.platform != "win32" and sys.platform != "darwin":
+    SCM_CREDENTIALS: int
+
+IPPROTO_ICMP: int
+IPPROTO_IP: int
+IPPROTO_RAW: int
+IPPROTO_TCP: int
+IPPROTO_UDP: int
+IPPROTO_AH: int
+IPPROTO_DSTOPTS: int
+IPPROTO_EGP: int
+IPPROTO_ESP: int
+IPPROTO_FRAGMENT: int
+IPPROTO_GGP: int
+IPPROTO_HOPOPTS: int
+IPPROTO_ICMPV6: int
+IPPROTO_IDP: int
+IPPROTO_IGMP: int
+IPPROTO_IPV4: int
+IPPROTO_IPV6: int
+IPPROTO_MAX: int
+IPPROTO_ND: int
+IPPROTO_NONE: int
+IPPROTO_PIM: int
+IPPROTO_PUP: int
+IPPROTO_ROUTING: int
+IPPROTO_SCTP: int
+if sys.platform != "darwin":
+    IPPROTO_CBT: int
+    IPPROTO_ICLFXBM: int
+    IPPROTO_IGP: int
+    IPPROTO_L2TP: int
+    IPPROTO_PGM: int
+    IPPROTO_RDP: int
+    IPPROTO_ST: int
+if sys.platform != "win32":
+    IPPROTO_EON: int
+    IPPROTO_GRE: int
+    IPPROTO_HELLO: int
+    IPPROTO_IPCOMP: int
+    IPPROTO_IPIP: int
+    IPPROTO_RSVP: int
+    IPPROTO_TP: int
+    IPPROTO_XTP: int
+if sys.platform != "win32" and sys.platform != "darwin":
+    IPPROTO_BIP: int
+    IPPROTO_MOBILE: int
+    IPPROTO_VRRP: int
+if sys.version_info >= (3, 9) and sys.platform == "linux":
+    # Availability: Linux >= 2.6.20, FreeBSD >= 10.1
+    IPPROTO_UDPLITE: int
+if sys.version_info >= (3, 10) and sys.platform == "linux":
+    IPPROTO_MPTCP: int
+
+IPPORT_RESERVED: int
+IPPORT_USERRESERVED: int
+
+INADDR_ALLHOSTS_GROUP: int
+INADDR_ANY: int
+INADDR_BROADCAST: int
+INADDR_LOOPBACK: int
+INADDR_MAX_LOCAL_GROUP: int
+INADDR_NONE: int
+INADDR_UNSPEC_GROUP: int
+
+IP_ADD_MEMBERSHIP: int
+IP_DROP_MEMBERSHIP: int
+IP_HDRINCL: int
+IP_MULTICAST_IF: int
+IP_MULTICAST_LOOP: int
+IP_MULTICAST_TTL: int
+IP_OPTIONS: int
+IP_RECVDSTADDR: int
+if sys.version_info >= (3, 10):
+    IP_RECVTOS: int
+elif sys.platform != "win32" and sys.platform != "darwin":
+    IP_RECVTOS: int
+IP_TOS: int
+IP_TTL: int
+if sys.platform != "win32":
+    IP_DEFAULT_MULTICAST_LOOP: int
+    IP_DEFAULT_MULTICAST_TTL: int
+    IP_MAX_MEMBERSHIPS: int
+    IP_RECVOPTS: int
+    IP_RECVRETOPTS: int
+    IP_RETOPTS: int
+if sys.platform != "win32" and sys.platform != "darwin":
+    IP_TRANSPARENT: int
+    IP_BIND_ADDRESS_NO_PORT: int
+if sys.version_info >= (3, 12):
+    IP_ADD_SOURCE_MEMBERSHIP: int
+    IP_BLOCK_SOURCE: int
+    IP_DROP_SOURCE_MEMBERSHIP: int
+    IP_PKTINFO: int
+    IP_UNBLOCK_SOURCE: int
+
+IPV6_CHECKSUM: int
+IPV6_JOIN_GROUP: int
+IPV6_LEAVE_GROUP: int
+IPV6_MULTICAST_HOPS: int
+IPV6_MULTICAST_IF: int
+IPV6_MULTICAST_LOOP: int
+IPV6_RECVTCLASS: int
+IPV6_TCLASS: int
+IPV6_UNICAST_HOPS: int
+IPV6_V6ONLY: int
+if sys.version_info >= (3, 9) or sys.platform != "darwin":
+    IPV6_DONTFRAG: int
+    IPV6_HOPLIMIT: int
+    IPV6_HOPOPTS: int
+    IPV6_PKTINFO: int
+    IPV6_RECVRTHDR: int
+    IPV6_RTHDR: int
+if sys.platform != "win32":
+    IPV6_RTHDR_TYPE_0: int
+    if sys.version_info >= (3, 9) or sys.platform != "darwin":
+        IPV6_DSTOPTS: int
+        IPV6_NEXTHOP: int
+        IPV6_PATHMTU: int
+        IPV6_RECVDSTOPTS: int
+        IPV6_RECVHOPLIMIT: int
+        IPV6_RECVHOPOPTS: int
+        IPV6_RECVPATHMTU: int
+        IPV6_RECVPKTINFO: int
+        IPV6_RTHDRDSTOPTS: int
+        IPV6_USE_MIN_MTU: int
+
+EAI_AGAIN: int
+EAI_BADFLAGS: int
+EAI_FAIL: int
+EAI_FAMILY: int
+EAI_MEMORY: int
+EAI_NODATA: int
+EAI_NONAME: int
+EAI_SERVICE: int
+EAI_SOCKTYPE: int
+if sys.platform != "win32":
+    EAI_ADDRFAMILY: int
+    EAI_BADHINTS: int
+    EAI_MAX: int
+    EAI_OVERFLOW: int
+    EAI_PROTOCOL: int
+    EAI_SYSTEM: int
+
+AI_ADDRCONFIG: int
+AI_ALL: int
+AI_CANONNAME: int
+AI_NUMERICHOST: int
+AI_NUMERICSERV: int
+AI_PASSIVE: int
+AI_V4MAPPED: int
+if sys.platform != "win32":
+    AI_DEFAULT: int
+    AI_MASK: int
+    AI_V4MAPPED_CFG: int
+
+NI_DGRAM: int
+NI_MAXHOST: int
+NI_MAXSERV: int
+NI_NAMEREQD: int
+NI_NOFQDN: int
+NI_NUMERICHOST: int
+NI_NUMERICSERV: int
+
+TCP_FASTOPEN: int
+TCP_KEEPCNT: int
+TCP_KEEPINTVL: int
+TCP_MAXSEG: int
+TCP_NODELAY: int
+if sys.platform != "win32":
+    TCP_NOTSENT_LOWAT: int
+if sys.platform != "darwin":
+    TCP_KEEPIDLE: int
+if sys.version_info >= (3, 10) and sys.platform == "darwin":
+    TCP_KEEPALIVE: int
+if sys.version_info >= (3, 11) and sys.platform == "darwin":
+    TCP_CONNECTION_INFO: int
+
+if sys.platform != "win32" and sys.platform != "darwin":
+    TCP_CONGESTION: int
+    TCP_CORK: int
+    TCP_DEFER_ACCEPT: int
+    TCP_INFO: int
+    TCP_LINGER2: int
+    TCP_QUICKACK: int
+    TCP_SYNCNT: int
+    TCP_USER_TIMEOUT: int
+    TCP_WINDOW_CLAMP: int
+
+# --------------------
+# Specifically documented constants
+# --------------------
+
+if sys.platform == "linux":
+    # Availability: Linux >= 2.6.25, NetBSD >= 8
+    AF_CAN: int
+    PF_CAN: int
+    SOL_CAN_BASE: int
+    SOL_CAN_RAW: int
+    CAN_EFF_FLAG: int
+    CAN_EFF_MASK: int
+    CAN_ERR_FLAG: int
+    CAN_ERR_MASK: int
+    CAN_RAW: int
+    CAN_RAW_ERR_FILTER: int
+    CAN_RAW_FILTER: int
+    CAN_RAW_LOOPBACK: int
+    CAN_RAW_RECV_OWN_MSGS: int
+    CAN_RTR_FLAG: int
+    CAN_SFF_MASK: int
+
+if sys.platform == "linux":
+    # Availability: Linux >= 2.6.25
+    CAN_BCM: int
+    CAN_BCM_TX_SETUP: int
+    CAN_BCM_TX_DELETE: int
+    CAN_BCM_TX_READ: int
+    CAN_BCM_TX_SEND: int
+    CAN_BCM_RX_SETUP: int
+    CAN_BCM_RX_DELETE: int
+    CAN_BCM_RX_READ: int
+    CAN_BCM_TX_STATUS: int
+    CAN_BCM_TX_EXPIRED: int
+    CAN_BCM_RX_STATUS: int
+    CAN_BCM_RX_TIMEOUT: int
+    CAN_BCM_RX_CHANGED: int
+    CAN_BCM_SETTIMER: int
+    CAN_BCM_STARTTIMER: int
+    CAN_BCM_TX_COUNTEVT: int
+    CAN_BCM_TX_ANNOUNCE: int
+    CAN_BCM_TX_CP_CAN_ID: int
+    CAN_BCM_RX_FILTER_ID: int
+    CAN_BCM_RX_CHECK_DLC: int
+    CAN_BCM_RX_NO_AUTOTIMER: int
+    CAN_BCM_RX_ANNOUNCE_RESUME: int
+    CAN_BCM_TX_RESET_MULTI_IDX: int
+    CAN_BCM_RX_RTR_FRAME: int
+    CAN_BCM_CAN_FD_FRAME: int
+
+if sys.platform == "linux":
+    # Availability: Linux >= 3.6
+    CAN_RAW_FD_FRAMES: int
+
+if sys.platform == "linux" and sys.version_info >= (3, 9):
+    # Availability: Linux >= 4.1
+    CAN_RAW_JOIN_FILTERS: int
+
+if sys.platform == "linux":
+    # Availability: Linux >= 2.6.25
+    CAN_ISOTP: int
+
+if sys.platform == "linux" and sys.version_info >= (3, 9):
+    # Availability: Linux >= 5.4
+    CAN_J1939: int
+
+    J1939_MAX_UNICAST_ADDR: int
+    J1939_IDLE_ADDR: int
+    J1939_NO_ADDR: int
+    J1939_NO_NAME: int
+    J1939_PGN_REQUEST: int
+    J1939_PGN_ADDRESS_CLAIMED: int
+    J1939_PGN_ADDRESS_COMMANDED: int
+    J1939_PGN_PDU1_MAX: int
+    J1939_PGN_MAX: int
+    J1939_NO_PGN: int
+
+    SO_J1939_FILTER: int
+    SO_J1939_PROMISC: int
+    SO_J1939_SEND_PRIO: int
+    SO_J1939_ERRQUEUE: int
+
+    SCM_J1939_DEST_ADDR: int
+    SCM_J1939_DEST_NAME: int
+    SCM_J1939_PRIO: int
+    SCM_J1939_ERRQUEUE: int
+
+    J1939_NLA_PAD: int
+    J1939_NLA_BYTES_ACKED: int
+    J1939_EE_INFO_NONE: int
+    J1939_EE_INFO_TX_ABORT: int
+    J1939_FILTER_MAX: int
+
+if sys.version_info >= (3, 12) and sys.platform != "linux" and sys.platform != "win32" and sys.platform != "darwin":
+    # Availability: FreeBSD >= 14.0
+    AF_DIVERT: int
+    PF_DIVERT: int
+
+if sys.platform == "linux":
+    # Availability: Linux >= 2.2
+    AF_PACKET: int
+    PF_PACKET: int
+    PACKET_BROADCAST: int
+    PACKET_FASTROUTE: int
+    PACKET_HOST: int
+    PACKET_LOOPBACK: int
+    PACKET_MULTICAST: int
+    PACKET_OTHERHOST: int
+    PACKET_OUTGOING: int
+
+if sys.version_info >= (3, 12) and sys.platform == "linux":
+    ETH_P_ALL: int
+
+if sys.platform == "linux":
+    # Availability: Linux >= 2.6.30
+    AF_RDS: int
+    PF_RDS: int
+    SOL_RDS: int
+    RDS_CANCEL_SENT_TO: int
+    RDS_CMSG_RDMA_ARGS: int
+    RDS_CMSG_RDMA_DEST: int
+    RDS_CMSG_RDMA_MAP: int
+    RDS_CMSG_RDMA_STATUS: int
+    RDS_CMSG_RDMA_UPDATE: int
+    RDS_CONG_MONITOR: int
+    RDS_FREE_MR: int
+    RDS_GET_MR: int
+    RDS_GET_MR_FOR_DEST: int
+    RDS_RDMA_DONTWAIT: int
+    RDS_RDMA_FENCE: int
+    RDS_RDMA_INVALIDATE: int
+    RDS_RDMA_NOTIFY_ME: int
+    RDS_RDMA_READWRITE: int
+    RDS_RDMA_SILENT: int
+    RDS_RDMA_USE_ONCE: int
+    RDS_RECVERR: int
+
+if sys.platform == "win32":
+    SIO_RCVALL: int
+    SIO_KEEPALIVE_VALS: int
+    SIO_LOOPBACK_FAST_PATH: int
+    RCVALL_MAX: int
+    RCVALL_OFF: int
+    RCVALL_ON: int
+    RCVALL_SOCKETLEVELONLY: int
+
+if sys.platform == "linux":
+    AF_TIPC: int
+    SOL_TIPC: int
+    TIPC_ADDR_ID: int
+    TIPC_ADDR_NAME: int
+    TIPC_ADDR_NAMESEQ: int
+    TIPC_CFG_SRV: int
+    TIPC_CLUSTER_SCOPE: int
+    TIPC_CONN_TIMEOUT: int
+    TIPC_CRITICAL_IMPORTANCE: int
+    TIPC_DEST_DROPPABLE: int
+    TIPC_HIGH_IMPORTANCE: int
+    TIPC_IMPORTANCE: int
+    TIPC_LOW_IMPORTANCE: int
+    TIPC_MEDIUM_IMPORTANCE: int
+    TIPC_NODE_SCOPE: int
+    TIPC_PUBLISHED: int
+    TIPC_SRC_DROPPABLE: int
+    TIPC_SUBSCR_TIMEOUT: int
+    TIPC_SUB_CANCEL: int
+    TIPC_SUB_PORTS: int
+    TIPC_SUB_SERVICE: int
+    TIPC_TOP_SRV: int
+    TIPC_WAIT_FOREVER: int
+    TIPC_WITHDRAWN: int
+    TIPC_ZONE_SCOPE: int
+
+if sys.platform == "linux":
+    # Availability: Linux >= 2.6.38
+    AF_ALG: int
+    SOL_ALG: int
+    ALG_OP_DECRYPT: int
+    ALG_OP_ENCRYPT: int
+    ALG_OP_SIGN: int
+    ALG_OP_VERIFY: int
+    ALG_SET_AEAD_ASSOCLEN: int
+    ALG_SET_AEAD_AUTHSIZE: int
+    ALG_SET_IV: int
+    ALG_SET_KEY: int
+    ALG_SET_OP: int
+    ALG_SET_PUBKEY: int
+
+if sys.platform == "linux":
+    # Availability: Linux >= 4.8 (or maybe 3.9, CPython docs are confusing)
+    AF_VSOCK: int
+    IOCTL_VM_SOCKETS_GET_LOCAL_CID: int
+    VMADDR_CID_ANY: int
+    VMADDR_CID_HOST: int
+    VMADDR_PORT_ANY: int
+    SO_VM_SOCKETS_BUFFER_MAX_SIZE: int
+    SO_VM_SOCKETS_BUFFER_SIZE: int
+    SO_VM_SOCKETS_BUFFER_MIN_SIZE: int
+    VM_SOCKETS_INVALID_VERSION: int  # undocumented
+
+if sys.platform != "win32" or sys.version_info >= (3, 9):
+    # Documented as only available on BSD, macOS, but empirically sometimes
+    # available on Windows
+    AF_LINK: int
+
+has_ipv6: bool
+
+if sys.platform != "darwin":
+    if sys.platform != "win32" or sys.version_info >= (3, 9):
+        BDADDR_ANY: str
+        BDADDR_LOCAL: str
+
+if sys.platform != "win32" and sys.platform != "darwin":
+    HCI_FILTER: int  # not in NetBSD or DragonFlyBSD
+    HCI_TIME_STAMP: int  # not in FreeBSD, NetBSD, or DragonFlyBSD
+    HCI_DATA_DIR: int  # not in FreeBSD, NetBSD, or DragonFlyBSD
+
+if sys.platform == "linux":
+    AF_QIPCRTR: int  # Availability: Linux >= 4.7
+
+if sys.version_info >= (3, 11) and sys.platform != "linux" and sys.platform != "win32" and sys.platform != "darwin":
+    # FreeBSD
+    SCM_CREDS2: int
+    LOCAL_CREDS: int
+    LOCAL_CREDS_PERSISTENT: int
+
+if sys.version_info >= (3, 11) and sys.platform == "linux":
+    SO_INCOMING_CPU: int  # Availability: Linux >= 3.9
+
+if sys.version_info >= (3, 12) and sys.platform == "win32":
+    # Availability: Windows
+    AF_HYPERV: int
+    HV_PROTOCOL_RAW: int
+    HVSOCKET_CONNECT_TIMEOUT: int
+    HVSOCKET_CONNECT_TIMEOUT_MAX: int
+    HVSOCKET_CONNECTED_SUSPEND: int
+    HVSOCKET_ADDRESS_FLAG_PASSTHRU: int
+    HV_GUID_ZERO: str
+    HV_GUID_WILDCARD: str
+    HV_GUID_BROADCAST: str
+    HV_GUID_CHILDREN: str
+    HV_GUID_LOOPBACK: str
+    HV_GUID_PARENT: str
+
+if sys.version_info >= (3, 12):
+    if sys.platform != "win32":
+        # Availability: Linux, FreeBSD, macOS
+        ETHERTYPE_ARP: int
+        ETHERTYPE_IP: int
+        ETHERTYPE_IPV6: int
+        ETHERTYPE_VLAN: int
+
+# --------------------
+# Semi-documented constants
+# These are alluded to under the "Socket families" section in the docs
+# https://docs.python.org/3/library/socket.html#socket-families
+# --------------------
+
+if sys.platform == "linux":
+    # Netlink is defined by Linux
+    AF_NETLINK: int
+    NETLINK_ARPD: int
+    NETLINK_CRYPTO: int
+    NETLINK_DNRTMSG: int
+    NETLINK_FIREWALL: int
+    NETLINK_IP6_FW: int
+    NETLINK_NFLOG: int
+    NETLINK_ROUTE6: int
+    NETLINK_ROUTE: int
+    NETLINK_SKIP: int
+    NETLINK_TAPBASE: int
+    NETLINK_TCPDIAG: int
+    NETLINK_USERSOCK: int
+    NETLINK_W1: int
+    NETLINK_XFRM: int
+
+if sys.platform == "darwin":
+    PF_SYSTEM: int
+    SYSPROTO_CONTROL: int
+
+if sys.platform != "darwin":
+    if sys.version_info >= (3, 9) or sys.platform != "win32":
+        AF_BLUETOOTH: int
+
+if sys.platform != "win32" and sys.platform != "darwin":
+    # Linux and some BSD support is explicit in the docs
+    # Windows and macOS do not support in practice
+    BTPROTO_HCI: int
+    BTPROTO_L2CAP: int
+    BTPROTO_SCO: int  # not in FreeBSD
+if sys.platform != "darwin":
+    if sys.version_info >= (3, 9) or sys.platform != "win32":
+        BTPROTO_RFCOMM: int
+
+if sys.version_info >= (3, 9) and sys.platform == "linux":
+    UDPLITE_RECV_CSCOV: int
+    UDPLITE_SEND_CSCOV: int
+
+# --------------------
+# Documented under socket.shutdown
+# --------------------
+SHUT_RD: int
+SHUT_RDWR: int
+SHUT_WR: int
+
+# --------------------
+# Undocumented constants
+# --------------------
+
+# Undocumented address families
+AF_APPLETALK: int
+AF_DECnet: int
+AF_IPX: int
+AF_SNA: int
+
+if sys.platform != "win32":
+    AF_ROUTE: int
+    AF_SYSTEM: int
+
+if sys.platform != "darwin":
+    AF_IRDA: int
+
+if sys.platform != "win32" and sys.platform != "darwin":
+    AF_AAL5: int
+    AF_ASH: int
+    AF_ATMPVC: int
+    AF_ATMSVC: int
+    AF_AX25: int
+    AF_BRIDGE: int
+    AF_ECONET: int
+    AF_KEY: int
+    AF_LLC: int
+    AF_NETBEUI: int
+    AF_NETROM: int
+    AF_PPPOX: int
+    AF_ROSE: int
+    AF_SECURITY: int
+    AF_WANPIPE: int
+    AF_X25: int
+
+# Miscellaneous undocumented
+
+if sys.platform != "win32":
+    LOCAL_PEERCRED: int
+
+if sys.platform != "win32" and sys.platform != "darwin":
+    IPX_TYPE: int
+
+# ===== Exceptions =====
+
+error = OSError
+
+class herror(error): ...
+class gaierror(error): ...
+
+if sys.version_info >= (3, 10):
+    timeout = TimeoutError
+else:
+    class timeout(error): ...
+
+# ===== Classes =====
+
+class socket:
+    @property
+    def family(self) -> int: ...
+    @property
+    def type(self) -> int: ...
+    @property
+    def proto(self) -> int: ...
+    @property
+    def timeout(self) -> float | None: ...
+    if sys.platform == "win32":
+        def __init__(
+            self, family: int = ..., type: int = ..., proto: int = ..., fileno: SupportsIndex | bytes | None = ...
+        ) -> None: ...
+    else:
+        def __init__(self, family: int = ..., type: int = ..., proto: int = ..., fileno: SupportsIndex | None = ...) -> None: ...
+
+    def bind(self, address: _Address, /) -> None: ...
+    def close(self) -> None: ...
+    def connect(self, address: _Address, /) -> None: ...
+    def connect_ex(self, address: _Address, /) -> int: ...
+    def detach(self) -> int: ...
+    def fileno(self) -> int: ...
+    def getpeername(self) -> _RetAddress: ...
+    def getsockname(self) -> _RetAddress: ...
+    @overload
+    def getsockopt(self, level: int, optname: int, /) -> int: ...
+    @overload
+    def getsockopt(self, level: int, optname: int, buflen: int, /) -> bytes: ...
+    def getblocking(self) -> bool: ...
+    def gettimeout(self) -> float | None: ...
+    if sys.platform == "win32":
+        def ioctl(self, control: int, option: int | tuple[int, int, int] | bool, /) -> None: ...
+
+    def listen(self, backlog: int = ..., /) -> None: ...
+    def recv(self, bufsize: int, flags: int = ..., /) -> bytes: ...
+    def recvfrom(self, bufsize: int, flags: int = ..., /) -> tuple[bytes, _RetAddress]: ...
+    if sys.platform != "win32":
+        def recvmsg(self, bufsize: int, ancbufsize: int = ..., flags: int = ..., /) -> tuple[bytes, list[_CMSG], int, Any]: ...
+        def recvmsg_into(
+            self, buffers: Iterable[WriteableBuffer], ancbufsize: int = ..., flags: int = ..., /
+        ) -> tuple[int, list[_CMSG], int, Any]: ...
+
+    def recvfrom_into(self, buffer: WriteableBuffer, nbytes: int = ..., flags: int = ...) -> tuple[int, _RetAddress]: ...
+    def recv_into(self, buffer: WriteableBuffer, nbytes: int = ..., flags: int = ...) -> int: ...
+    def send(self, data: ReadableBuffer, flags: int = ..., /) -> int: ...
+    def sendall(self, data: ReadableBuffer, flags: int = ..., /) -> None: ...
+    @overload
+    def sendto(self, data: ReadableBuffer, address: _Address, /) -> int: ...
+    @overload
+    def sendto(self, data: ReadableBuffer, flags: int, address: _Address, /) -> int: ...
+    if sys.platform != "win32":
+        def sendmsg(
+            self,
+            buffers: Iterable[ReadableBuffer],
+            ancdata: Iterable[_CMSGArg] = ...,
+            flags: int = ...,
+            address: _Address | None = ...,
+            /,
+        ) -> int: ...
+    if sys.platform == "linux":
+        def sendmsg_afalg(
+            self, msg: Iterable[ReadableBuffer] = ..., *, op: int, iv: Any = ..., assoclen: int = ..., flags: int = ...
+        ) -> int: ...
+
+    def setblocking(self, flag: bool, /) -> None: ...
+    def settimeout(self, value: float | None, /) -> None: ...
+    @overload
+    def setsockopt(self, level: int, optname: int, value: int | ReadableBuffer, /) -> None: ...
+    @overload
+    def setsockopt(self, level: int, optname: int, value: None, optlen: int, /) -> None: ...
+    if sys.platform == "win32":
+        def share(self, process_id: int, /) -> bytes: ...
+
+    def shutdown(self, how: int, /) -> None: ...
+
+SocketType = socket
+
+# ===== Functions =====
+
+def close(fd: SupportsIndex, /) -> None: ...
+def dup(fd: SupportsIndex, /) -> int: ...
+
+# the 5th tuple item is an address
+def getaddrinfo(
+    host: bytes | str | None,
+    port: bytes | str | int | None,
+    family: int = ...,
+    type: int = ...,
+    proto: int = ...,
+    flags: int = ...,
+) -> list[tuple[int, int, int, str, tuple[str, int] | tuple[str, int, int, int]]]: ...
+def gethostbyname(hostname: str, /) -> str: ...
+def gethostbyname_ex(hostname: str, /) -> tuple[str, list[str], list[str]]: ...
+def gethostname() -> str: ...
+def gethostbyaddr(ip_address: str, /) -> tuple[str, list[str], list[str]]: ...
+def getnameinfo(sockaddr: tuple[str, int] | tuple[str, int, int, int], flags: int, /) -> tuple[str, str]: ...
+def getprotobyname(protocolname: str, /) -> int: ...
+def getservbyname(servicename: str, protocolname: str = ..., /) -> int: ...
+def getservbyport(port: int, protocolname: str = ..., /) -> str: ...
+def ntohl(x: int, /) -> int: ...  # param & ret val are 32-bit ints
+def ntohs(x: int, /) -> int: ...  # param & ret val are 16-bit ints
+def htonl(x: int, /) -> int: ...  # param & ret val are 32-bit ints
+def htons(x: int, /) -> int: ...  # param & ret val are 16-bit ints
+def inet_aton(ip_addr: str, /) -> bytes: ...  # ret val 4 bytes in length
+def inet_ntoa(packed_ip: ReadableBuffer, /) -> str: ...
+def inet_pton(address_family: int, ip_string: str, /) -> bytes: ...
+def inet_ntop(address_family: int, packed_ip: ReadableBuffer, /) -> str: ...
+def getdefaulttimeout() -> float | None: ...
+def setdefaulttimeout(timeout: float | None, /) -> None: ...
+
+if sys.platform != "win32":
+    def sethostname(name: str, /) -> None: ...
+    def CMSG_LEN(length: int, /) -> int: ...
+    def CMSG_SPACE(length: int, /) -> int: ...
+    def socketpair(family: int = ..., type: int = ..., proto: int = ..., /) -> tuple[socket, socket]: ...
+
+def if_nameindex() -> list[tuple[int, str]]: ...
+def if_nametoindex(oname: str, /) -> int: ...
+def if_indextoname(index: int, /) -> str: ...
+
+CAPI: object

crates/red_knot_module_resolver/vendor/typeshed/stdlib/_stat.pyi

@@ -0,0 +1,117 @@
+import sys
+from typing import Literal
+
+SF_APPEND: Literal[0x00040000]
+SF_ARCHIVED: Literal[0x00010000]
+SF_IMMUTABLE: Literal[0x00020000]
+SF_NOUNLINK: Literal[0x00100000]
+SF_SNAPSHOT: Literal[0x00200000]
+
+ST_MODE: Literal[0]
+ST_INO: Literal[1]
+ST_DEV: Literal[2]
+ST_NLINK: Literal[3]
+ST_UID: Literal[4]
+ST_GID: Literal[5]
+ST_SIZE: Literal[6]
+ST_ATIME: Literal[7]
+ST_MTIME: Literal[8]
+ST_CTIME: Literal[9]
+
+S_IFIFO: Literal[0o010000]
+S_IFLNK: Literal[0o120000]
+S_IFREG: Literal[0o100000]
+S_IFSOCK: Literal[0o140000]
+S_IFBLK: Literal[0o060000]
+S_IFCHR: Literal[0o020000]
+S_IFDIR: Literal[0o040000]
+
+# These are 0 on systems that don't support the specific kind of file.
+# Example: Linux doesn't support door files, so S_IFDOOR is 0 on linux.
+S_IFDOOR: int
+S_IFPORT: int
+S_IFWHT: int
+
+S_ISUID: Literal[0o4000]
+S_ISGID: Literal[0o2000]
+S_ISVTX: Literal[0o1000]
+
+S_IRWXU: Literal[0o0700]
+S_IRUSR: Literal[0o0400]
+S_IWUSR: Literal[0o0200]
+S_IXUSR: Literal[0o0100]
+
+S_IRWXG: Literal[0o0070]
+S_IRGRP: Literal[0o0040]
+S_IWGRP: Literal[0o0020]
+S_IXGRP: Literal[0o0010]
+
+S_IRWXO: Literal[0o0007]
+S_IROTH: Literal[0o0004]
+S_IWOTH: Literal[0o0002]
+S_IXOTH: Literal[0o0001]
+
+S_ENFMT: Literal[0o2000]
+S_IREAD: Literal[0o0400]
+S_IWRITE: Literal[0o0200]
+S_IEXEC: Literal[0o0100]
+
+UF_APPEND: Literal[0x00000004]
+UF_COMPRESSED: Literal[0x00000020]  # OS X 10.6+ only
+UF_HIDDEN: Literal[0x00008000]  # OX X 10.5+ only
+UF_IMMUTABLE: Literal[0x00000002]
+UF_NODUMP: Literal[0x00000001]
+UF_NOUNLINK: Literal[0x00000010]
+UF_OPAQUE: Literal[0x00000008]
+
+def S_IMODE(mode: int, /) -> int: ...
+def S_IFMT(mode: int, /) -> int: ...
+def S_ISBLK(mode: int, /) -> bool: ...
+def S_ISCHR(mode: int, /) -> bool: ...
+def S_ISDIR(mode: int, /) -> bool: ...
+def S_ISDOOR(mode: int, /) -> bool: ...
+def S_ISFIFO(mode: int, /) -> bool: ...
+def S_ISLNK(mode: int, /) -> bool: ...
+def S_ISPORT(mode: int, /) -> bool: ...
+def S_ISREG(mode: int, /) -> bool: ...
+def S_ISSOCK(mode: int, /) -> bool: ...
+def S_ISWHT(mode: int, /) -> bool: ...
+def filemode(mode: int, /) -> str: ...
+
+if sys.platform == "win32":
+    IO_REPARSE_TAG_SYMLINK: int
+    IO_REPARSE_TAG_MOUNT_POINT: int
+    IO_REPARSE_TAG_APPEXECLINK: int
+
+if sys.platform == "win32":
+    FILE_ATTRIBUTE_ARCHIVE: Literal[32]
+    FILE_ATTRIBUTE_COMPRESSED: Literal[2048]
+    FILE_ATTRIBUTE_DEVICE: Literal[64]
+    FILE_ATTRIBUTE_DIRECTORY: Literal[16]
+    FILE_ATTRIBUTE_ENCRYPTED: Literal[16384]
+    FILE_ATTRIBUTE_HIDDEN: Literal[2]
+    FILE_ATTRIBUTE_INTEGRITY_STREAM: Literal[32768]
+    FILE_ATTRIBUTE_NORMAL: Literal[128]
+    FILE_ATTRIBUTE_NOT_CONTENT_INDEXED: Literal[8192]
+    FILE_ATTRIBUTE_NO_SCRUB_DATA: Literal[131072]
+    FILE_ATTRIBUTE_OFFLINE: Literal[4096]
+    FILE_ATTRIBUTE_READONLY: Literal[1]
+    FILE_ATTRIBUTE_REPARSE_POINT: Literal[1024]
+    FILE_ATTRIBUTE_SPARSE_FILE: Literal[512]
+    FILE_ATTRIBUTE_SYSTEM: Literal[4]
+    FILE_ATTRIBUTE_TEMPORARY: Literal[256]
+    FILE_ATTRIBUTE_VIRTUAL: Literal[65536]
+
+if sys.version_info >= (3, 13):
+    SF_SETTABLE: Literal[0x3FFF0000]
+    # https://github.com/python/cpython/issues/114081#issuecomment-2119017790
+    # SF_RESTRICTED: Literal[0x00080000]
+    SF_FIRMLINK: Literal[0x00800000]
+    SF_DATALESS: Literal[0x40000000]
+
+    SF_SUPPORTED: Literal[0x9F0000]
+    SF_SYNTHETIC: Literal[0xC0000000]
+
+    UF_TRACKED: Literal[0x00000040]
+    UF_DATAVAULT: Literal[0x00000080]
+    UF_SETTABLE: Literal[0x0000FFFF]