Add formatter to runes

## Summary

We weren't recursing into the comprehension body.

.config/nextest.toml

@@ -0,0 +1,8 @@
+[profile.ci]
+# Print out output for failing tests as soon as they fail, and also at the end
+# of the run (for easy scrollability).
+failure-output = "immediate-final"
+# Do not cancel the test run on the first failure.
+fail-fast = false
+
+status-level = "skip"

.gitattributes

@@ -2,9 +2,13 @@
 
 crates/ruff_linter/resources/test/fixtures/isort/line_ending_crlf.py text eol=crlf
 crates/ruff_linter/resources/test/fixtures/pycodestyle/W605_1.py text eol=crlf
+crates/ruff_linter/resources/test/fixtures/pycodestyle/W391_2.py text eol=crlf
+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/inline linguist-generated=true
+
 ruff.schema.json linguist-generated=true text=auto eol=lf
 *.md.snap linguist-language=Markdown

.github/CODEOWNERS

@@ -5,5 +5,13 @@
 # - The '*' pattern is global owners.
 # - Order is important. The last matching pattern has the most precedence.
 
-# Jupyter
-/crates/ruff_linter/src/jupyter/ @dhruvmanila
+/crates/ruff_notebook/ @dhruvmanila
+/crates/ruff_formatter/ @MichaReiser
+/crates/ruff_python_formatter/ @MichaReiser
+/crates/ruff_python_parser/ @MichaReiser @dhruvmanila
+
+# flake8-pyi
+/crates/ruff_linter/src/rules/flake8_pyi/ @AlexWaygood
+
+# Script for fuzzing the parser
+/scripts/fuzz-parser/ @AlexWaygood

.github/ISSUE_TEMPLATE.md

@@ -3,6 +3,8 @@ Thank you for taking the time to report an issue! We're glad to have you involve
 
 If you're filing a bug report, please consider including the following information:
 
+* List of keywords you searched for before creating this issue. Write them down here so that others can find this issue more easily and help provide feedback.
+  e.g. "RUF001", "unused variable", "Jupyter notebook"
 * A minimal code snippet that reproduces the bug.
 * The command you invoked (e.g., `ruff /path/to/file.py --fix`), ideally including the `--isolated` flag.
 * The current Ruff settings (any relevant sections from your `pyproject.toml`).

.github/dependabot.yml

@@ -1,21 +0,0 @@
-version: 2
-updates:
-  - package-ecosystem: "github-actions"
-    directory: "/"
-    schedule:
-      interval: "weekly"
-    labels: ["internal"]
-    groups:
-      actions:
-        patterns:
-          - "*"
-    ignore:
-      # The latest versions of these are not compatible with our release workflow
-      - dependency-name: "actions/upload-artifact"
-      - dependency-name: "actions/download-artifact"
-
-  - package-ecosystem: "cargo"
-    directory: "/"
-    schedule:
-      interval: "weekly"
-    labels: ["internal"]

.github/renovate.json5

@@ -0,0 +1,68 @@
+{
+  $schema: "https://docs.renovatebot.com/renovate-schema.json",
+  dependencyDashboard: true,
+  suppressNotifications: ["prEditedNotification"],
+  extends: ["config:recommended"],
+  labels: ["internal"],
+  schedule: ["before 4am on Monday"],
+  semanticCommits: "disabled",
+  separateMajorMinor: false,
+  prHourlyLimit: 10,
+  enabledManagers: ["github-actions", "pre-commit", "cargo", "pep621", "npm"],
+  cargo: {
+    // See https://docs.renovatebot.com/configuration-options/#rangestrategy
+    rangeStrategy: "update-lockfile",
+  },
+  pep621: {
+    fileMatch: ["^(python|scripts)/.*pyproject\\.toml$"],
+  },
+  npm: {
+    fileMatch: ["^playground/.*package\\.json$"],
+  },
+  "pre-commit": {
+    enabled: true,
+  },
+  packageRules: [
+    {
+      // Group upload/download artifact updates, the versions are dependent
+      groupName: "Artifact GitHub Actions dependencies",
+      matchManagers: ["github-actions"],
+      matchPackagePatterns: ["actions/.*-artifact"],
+      description: "Weekly update of artifact-related GitHub Actions dependencies",
+    },
+    {
+      groupName: "pre-commit dependencies",
+      matchManagers: ["pre-commit"],
+      description: "Weekly update of pre-commit dependencies",
+    },
+    {
+      groupName: "NPM Development dependencies",
+      matchManagers: ["npm"],
+      matchDepTypes: ["devDependencies"],
+      description: "Weekly update of NPM development dependencies",
+    },
+    {
+      groupName: "Monaco",
+      matchManagers: ["npm"],
+      matchPackagePatterns: ["monaco"],
+      description: "Weekly update of the Monaco editor",
+    },
+    {
+      groupName: "strum",
+      matchManagers: ["cargo"],
+      matchPackagePatterns: ["strum"],
+      description: "Weekly update of strum dependencies",
+    },
+    {
+      groupName: "ESLint",
+      matchManagers: ["npm"],
+      matchPackageNames: ["eslint"],
+      allowedVersions: "<9",
+      description: "Constraint ESLint to version 8 until TypeScript-eslint supports ESLint 9", // https://github.com/typescript-eslint/typescript-eslint/issues/8211
+    },
+  ],
+  vulnerabilityAlerts: {
+    commitMessageSuffix: "",
+    labels: ["internal", "security"],
+  },
+}

.github/workflows/ci.yaml

@@ -23,6 +23,8 @@ jobs:
     name: "Determine changes"
     runs-on: ubuntu-latest
     outputs:
+      # Flag that is raised when any code that affects parser is changed
+      parser: ${{ steps.changed.outputs.parser_any_changed }}
       # Flag that is raised when any code that affects linter is changed
       linter: ${{ steps.changed.outputs.linter_any_changed }}
       # Flag that is raised when any code that affects formatter is changed
@@ -35,10 +37,21 @@ jobs:
         with:
           fetch-depth: 0
 
-      - uses: tj-actions/changed-files@v42
+      - uses: tj-actions/changed-files@v44
         id: changed
         with:
           files_yaml: |
+            parser:
+              - Cargo.toml
+              - Cargo.lock
+              - crates/ruff_python_trivia/**
+              - crates/ruff_source_file/**
+              - crates/ruff_text_size/**
+              - crates/ruff_python_ast/**
+              - crates/ruff_python_parser/**
+              - scripts/fuzz-parser/**
+              - .github/workflows/ci.yaml
+
             linter:
               - Cargo.toml
               - Cargo.lock
@@ -111,22 +124,29 @@ jobs:
       - uses: actions/checkout@v4
       - name: "Install Rust toolchain"
         run: rustup show
+      - 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"
-        run: cargo insta test --all --exclude ruff_dev --all-features --unreferenced reject
-      - name: "Run dev tests"
-        # e.g. generating the schema — these should not run with all features enabled
-        run: cargo insta test -p ruff_dev --unreferenced reject
+        shell: bash
+        env:
+          NEXTEST_PROFILE: "ci"
+        run: cargo insta test --all-features --unreferenced reject --test-runner nextest
+
       # Check for broken links in the documentation.
       - run: cargo doc --all --no-deps
         env:
           # Setting RUSTDOCFLAGS because `cargo doc --check` isn't yet implemented (https://github.com/rust-lang/cargo/issues/10025).
           RUSTDOCFLAGS: "-D warnings"
-      - uses: actions/upload-artifact@v3
+      - uses: actions/upload-artifact@v4
         with:
           name: ruff
           path: target/debug/ruff
@@ -141,15 +161,16 @@ jobs:
       - uses: actions/checkout@v4
       - name: "Install Rust toolchain"
         run: rustup show
-      - name: "Install cargo insta"
+      - name: "Install cargo nextest"
         uses: taiki-e/install-action@v2
         with:
-          tool: cargo-insta
+          tool: cargo-nextest
       - uses: Swatinem/rust-cache@v2
       - name: "Run tests"
         shell: bash
-        # We can't reject unreferenced snapshots on windows because flake8_executable can't run on windows
-        run: cargo insta test --all --exclude ruff_dev --all-features
+        run: |
+          cargo nextest run --all-features --profile ci
+          cargo test --all-features --doc
 
   cargo-test-wasm:
     name: "cargo test (wasm)"
@@ -173,6 +194,22 @@ 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-fuzz:
     name: "cargo fuzz"
     runs-on: ubuntu-latest
@@ -192,6 +229,38 @@ jobs:
           tool: cargo-fuzz@0.11.2
       - run: cargo fuzz build -s none
 
+  fuzz-parser:
+    name: "Fuzz the parser"
+    runs-on: ubuntu-latest
+    needs:
+      - cargo-test-linux
+      - determine_changes
+    if: ${{ needs.determine_changes.outputs.parser == 'true' }}
+    timeout-minutes: 20
+    env:
+      FORCE_COLOR: 1
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ env.PYTHON_VERSION }}
+      - 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
+      - uses: actions/download-artifact@v4
+        name: Download Ruff binary to test
+        id: download-cached-binary
+        with:
+          name: ruff
+          path: ruff-to-test
+      - name: Fuzz
+        run: |
+          # Make executable, since artifact download doesn't preserve this
+          chmod +x ${{ steps.download-cached-binary.outputs.download-path }}/ruff
+
+          python scripts/fuzz-parser/fuzz.py 0-500 --test-executable ${{ steps.download-cached-binary.outputs.download-path }}/ruff
+
   scripts:
     name: "test scripts"
     runs-on: ubuntu-latest
@@ -220,9 +289,7 @@ jobs:
       - determine_changes
     # Only runs on pull requests, since that is the only we way we can find the base version for comparison.
     # Ecosystem check needs linter and/or formatter changes.
-    if: github.event_name == 'pull_request' && ${{
-      needs.determine_changes.outputs.code == 'true'
-      }}
+    if: ${{ github.event_name == 'pull_request' && needs.determine_changes.outputs.code == 'true' }}
     timeout-minutes: 20
     steps:
       - uses: actions/checkout@v4
@@ -230,7 +297,7 @@ jobs:
         with:
           python-version: ${{ env.PYTHON_VERSION }}
 
-      - uses: actions/download-artifact@v3
+      - uses: actions/download-artifact@v4
         name: Download comparison Ruff binary
         id: ruff-target
         with:
@@ -242,6 +309,7 @@ jobs:
         with:
           name: ruff
           branch: ${{ github.event.pull_request.base.ref }}
+          workflow: "ci.yaml"
           check_artifacts: true
 
       - name: Install ruff-ecosystem
@@ -316,13 +384,13 @@ jobs:
         run: |
           echo ${{ github.event.number }} > pr-number
 
-      - uses: actions/upload-artifact@v3
+      - uses: actions/upload-artifact@v4
         name: Upload PR Number
         with:
           name: pr-number
           path: pr-number
 
-      - uses: actions/upload-artifact@v3
+      - uses: actions/upload-artifact@v4
         name: Upload Results
         with:
           name: ecosystem-result
@@ -410,7 +478,7 @@ jobs:
       - uses: actions/setup-python@v5
       - name: "Add SSH key"
         if: ${{ env.MKDOCS_INSIDERS_SSH_KEY_EXISTS == 'true' }}
-        uses: webfactory/ssh-agent@v0.8.0
+        uses: webfactory/ssh-agent@v0.9.0
         with:
           ssh-private-key: ${{ secrets.MKDOCS_INSIDERS_SSH_KEY }}
       - name: "Install Rust toolchain"
@@ -463,7 +531,7 @@ jobs:
       - determine_changes
     if: ${{ needs.determine_changes.outputs.code == 'true' || github.ref == 'refs/heads/main' }}
     steps:
-      - uses: extractions/setup-just@v1
+      - uses: extractions/setup-just@v2
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
@@ -476,7 +544,7 @@ jobs:
         with:
           python-version: ${{ env.PYTHON_VERSION }}
 
-      - uses: actions/download-artifact@v3
+      - uses: actions/download-artifact@v4
         name: Download development ruff binary
         id: ruff-target
         with:
@@ -500,7 +568,7 @@ jobs:
   benchmarks:
     runs-on: ubuntu-latest
     needs: determine_changes
-    if: ${{ needs.determine_changes.outputs.code == 'true' || github.ref == 'refs/heads/main' }}
+    if: ${{ github.repository == 'astral-sh/ruff' && (needs.determine_changes.outputs.code == 'true' || github.ref == 'refs/heads/main') }}
     timeout-minutes: 20
     steps:
       - name: "Checkout Branch"

.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/docs.yaml

@@ -23,7 +23,7 @@ jobs:
       - uses: actions/setup-python@v5
       - name: "Add SSH key"
         if: ${{ env.MKDOCS_INSIDERS_SSH_KEY_EXISTS == 'true' }}
-        uses: webfactory/ssh-agent@v0.8.0
+        uses: webfactory/ssh-agent@v0.9.0
         with:
           ssh-private-key: ${{ secrets.MKDOCS_INSIDERS_SSH_KEY }}
       - name: "Install Rust toolchain"

.github/workflows/release.yaml

@@ -52,13 +52,13 @@ jobs:
           ruff --help
           python -m ruff --help
       - name: "Upload sdist"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: wheels
+          name: wheels-sdist
           path: dist
 
   macos-x86_64:
-    runs-on: macos-latest
+    runs-on: macos-12
     steps:
       - uses: actions/checkout@v4
         with:
@@ -80,9 +80,9 @@ jobs:
           ruff --help
           python -m ruff --help
       - name: "Upload wheels"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: wheels
+          name: wheels-macos-x86_64
           path: dist
       - name: "Archive binary"
         run: |
@@ -90,15 +90,15 @@ jobs:
           tar czvf $ARCHIVE_FILE -C target/x86_64-apple-darwin/release ruff
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: binaries
+          name: binaries-macos-x86_64
           path: |
             *.tar.gz
             *.sha256
 
-  macos-universal:
-    runs-on: macos-latest
+  macos-aarch64:
+    runs-on: macos-14
     steps:
       - uses: actions/checkout@v4
         with:
@@ -106,22 +106,23 @@ jobs:
       - uses: actions/setup-python@v5
         with:
           python-version: ${{ env.PYTHON_VERSION }}
-          architecture: x64
+          architecture: arm64
       - name: "Prep README.md"
         run: python scripts/transform_readme.py --target pypi
-      - name: "Build wheels - universal2"
+      - name: "Build wheels - aarch64"
         uses: PyO3/maturin-action@v1
         with:
-          args: --release --locked --target universal2-apple-darwin --out dist
-      - name: "Test wheel - universal2"
+          target: aarch64
+          args: --release --locked --out dist
+      - name: "Test wheel - aarch64"
         run: |
-          pip install dist/${{ env.PACKAGE_NAME }}-*universal2.whl --force-reinstall
+          pip install dist/${{ env.PACKAGE_NAME }}-*.whl --force-reinstall
           ruff --help
           python -m ruff --help
       - name: "Upload wheels"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: wheels
+          name: wheels-aarch64-apple-darwin
           path: dist
       - name: "Archive binary"
         run: |
@@ -129,9 +130,9 @@ jobs:
           tar czvf $ARCHIVE_FILE -C target/aarch64-apple-darwin/release ruff
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: binaries
+          name: binaries-aarch64-apple-darwin
           path: |
             *.tar.gz
             *.sha256
@@ -170,9 +171,9 @@ jobs:
           ruff --help
           python -m ruff --help
       - name: "Upload wheels"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: wheels
+          name: wheels-${{ matrix.platform.target }}
           path: dist
       - name: "Archive binary"
         shell: bash
@@ -181,9 +182,9 @@ jobs:
           7z a $ARCHIVE_FILE ./target/${{ matrix.platform.target }}/release/ruff.exe
           sha256sum $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: binaries
+          name: binaries-${{ matrix.platform.target }}
           path: |
             *.zip
             *.sha256
@@ -218,9 +219,9 @@ jobs:
           ruff --help
           python -m ruff --help
       - name: "Upload wheels"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: wheels
+          name: wheels-${{ matrix.target }}
           path: dist
       - name: "Archive binary"
         run: |
@@ -228,9 +229,9 @@ jobs:
           tar czvf $ARCHIVE_FILE -C target/${{ matrix.target }}/release ruff
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: binaries
+          name: binaries-${{ matrix.target }}
           path: |
             *.tar.gz
             *.sha256
@@ -251,8 +252,12 @@ jobs:
             arch: s390x
           - target: powerpc64le-unknown-linux-gnu
             arch: ppc64le
+            # see https://github.com/astral-sh/ruff/issues/10073
+            maturin_docker_options: -e JEMALLOC_SYS_WITH_LG_PAGE=16
           - target: powerpc64-unknown-linux-gnu
             arch: ppc64
+            # see https://github.com/astral-sh/ruff/issues/10073
+            maturin_docker_options: -e JEMALLOC_SYS_WITH_LG_PAGE=16
 
     steps:
       - uses: actions/checkout@v4
@@ -285,9 +290,9 @@ jobs:
             pip3 install ${{ env.PACKAGE_NAME }} --no-index --find-links dist/ --force-reinstall
             ruff --help
       - name: "Upload wheels"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: wheels
+          name: wheels-${{ matrix.platform.target }}
           path: dist
       - name: "Archive binary"
         run: |
@@ -295,9 +300,9 @@ jobs:
           tar czvf $ARCHIVE_FILE -C target/${{ matrix.platform.target }}/release ruff
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: binaries
+          name: binaries-${{ matrix.platform.target }}
           path: |
             *.tar.gz
             *.sha256
@@ -337,9 +342,9 @@ jobs:
             .venv/bin/pip3 install ${{ env.PACKAGE_NAME }} --no-index --find-links dist/ --force-reinstall
             .venv/bin/ruff check --help
       - name: "Upload wheels"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: wheels
+          name: wheels-${{ matrix.target }}
           path: dist
       - name: "Archive binary"
         run: |
@@ -347,9 +352,9 @@ jobs:
           tar czvf $ARCHIVE_FILE -C target/${{ matrix.target }}/release ruff
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: binaries
+          name: binaries-${{ matrix.target }}
           path: |
             *.tar.gz
             *.sha256
@@ -394,9 +399,9 @@ jobs:
             .venv/bin/pip3 install ${{ env.PACKAGE_NAME }} --no-index --find-links dist/ --force-reinstall
             .venv/bin/ruff check --help
       - name: "Upload wheels"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: wheels
+          name: wheels-${{ matrix.platform.target }}
           path: dist
       - name: "Archive binary"
         run: |
@@ -404,9 +409,9 @@ jobs:
           tar czvf $ARCHIVE_FILE -C target/${{ matrix.platform.target }}/release ruff
           shasum -a 256 $ARCHIVE_FILE > $ARCHIVE_FILE.sha256
       - name: "Upload binary"
-        uses: actions/upload-artifact@v3
+        uses: actions/upload-artifact@v4
         with:
-          name: binaries
+          name: binaries-${{ matrix.platform.target }}
           path: |
             *.tar.gz
             *.sha256
@@ -447,7 +452,7 @@ jobs:
     name: Upload to PyPI
     runs-on: ubuntu-latest
     needs:
-      - macos-universal
+      - macos-aarch64
       - macos-x86_64
       - windows
       - linux
@@ -463,10 +468,11 @@ jobs:
       # For pypi trusted publishing
       id-token: write
     steps:
-      - uses: actions/download-artifact@v3
+      - uses: actions/download-artifact@v4
         with:
-          name: wheels
+          pattern: wheels-*
           path: wheels
+          merge-multiple: true
       - name: Publish to PyPi
         uses: pypa/gh-action-pypi-publish@release/v1
         with:
@@ -506,12 +512,13 @@ jobs:
       # For GitHub release publishing
       contents: write
     steps:
-      - uses: actions/download-artifact@v3
+      - uses: actions/download-artifact@v4
         with:
-          name: binaries
+          pattern: binaries-*
           path: binaries
+          merge-multiple: true
       - name: "Publish to GitHub"
-        uses: softprops/action-gh-release@v1
+        uses: softprops/action-gh-release@v2
         with:
           draft: true
           files: binaries/*

.gitignore

@@ -92,6 +92,7 @@ coverage.xml
 .hypothesis/
 .pytest_cache/
 cover/
+repos/
 
 # Translations
 *.mo

.markdownlint.yaml

@@ -17,4 +17,4 @@ MD013: false
 # MD024/no-duplicate-heading
 MD024:
   # Allow when nested under different parents e.g. CHANGELOG.md
-  allow_different_nesting: true
+  siblings_only: true

.pre-commit-config.yaml

@@ -13,7 +13,7 @@ exclude: |
 
 repos:
   - repo: https://github.com/abravalheri/validate-pyproject
-    rev: v0.15
+    rev: v0.16
     hooks:
       - id: validate-pyproject
 
@@ -31,7 +31,7 @@ repos:
           )$
 
   - repo: https://github.com/igorshubovych/markdownlint-cli
-    rev: v0.37.0
+    rev: v0.39.0
     hooks:
       - id: markdownlint-fix
         exclude: |
@@ -41,7 +41,7 @@ repos:
           )$
 
   - repo: https://github.com/crate-ci/typos
-    rev: v1.16.22
+    rev: v1.20.10
     hooks:
       - id: typos
 
@@ -55,7 +55,7 @@ repos:
         pass_filenames: false # This makes it a lot faster
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.1.4
+    rev: v0.4.2
     hooks:
       - id: ruff-format
       - id: ruff
@@ -70,7 +70,7 @@ repos:
 
   # Prettier
   - repo: https://github.com/pre-commit/mirrors-prettier
-    rev: v3.0.3
+    rev: v3.1.0
     hooks:
       - id: prettier
         types: [yaml]

BREAKING_CHANGES.md

@@ -1,5 +1,56 @@
 # Breaking Changes
 
+## 0.3.0
+
+### Ruff 2024.2 style
+
+The formatter now formats code according to the Ruff 2024.2 style guide. Read the [changelog](./CHANGELOG.md#030) for a detailed list of stabilized style changes.
+
+### `isort`: Use one blank line after imports in typing stub files ([#9971](https://github.com/astral-sh/ruff/pull/9971))
+
+Previously, Ruff used one or two blank lines (or the number configured by `isort.lines-after-imports`) after imports in typing stub files (`.pyi` files).
+The [typing style guide for stubs](https://typing.readthedocs.io/en/latest/source/stubs.html#style-guide) recommends using at most 1 blank line for grouping.
+As of this release, `isort` now always uses one blank line after imports in stub files, the same as the formatter.
+
+### `build` is no longer excluded by default ([#10093](https://github.com/astral-sh/ruff/pull/10093))
+
+Ruff maintains a list of directories and files that are excluded by default. This list now consists of the following patterns:
+
+- `.bzr`
+- `.direnv`
+- `.eggs`
+- `.git`
+- `.git-rewrite`
+- `.hg`
+- `.ipynb_checkpoints`
+- `.mypy_cache`
+- `.nox`
+- `.pants.d`
+- `.pyenv`
+- `.pytest_cache`
+- `.pytype`
+- `.ruff_cache`
+- `.svn`
+- `.tox`
+- `.venv`
+- `.vscode`
+- `__pypackages__`
+- `_build`
+- `buck-out`
+- `dist`
+- `node_modules`
+- `site-packages`
+- `venv`
+
+Previously, the `build` directory was included in this list. However, the `build` directory tends to be a not-unpopular directory
+name, and excluding it by default caused confusion. Ruff now no longer excludes `build` except if it is excluded by a `.gitignore` file
+or because it is listed in `extend-exclude`.
+
+### `--format` is no longer a valid `rule` or `linter` command option
+
+Previously, `ruff rule` and `ruff linter` accepted the `--format <FORMAT>` option as an alias for `--output-format`. Ruff no longer
+supports this alias. Please use `ruff rule --output-format <FORMAT>` and `ruff linter --output-format <FORMAT>` instead.
+
 ## 0.1.9
 
 ### `site-packages` is now excluded by default ([#5513](https://github.com/astral-sh/ruff/pull/5513))

CHANGELOG.md

@@ -1,5 +1,496 @@
 # Changelog
 
+## 0.4.2
+
+### Rule changes
+
+- \[`flake8-pyi`\] Allow for overloaded `__exit__` and `__aexit__` definitions (`PYI036`) ([#11057](https://github.com/astral-sh/ruff/pull/11057))
+- \[`pyupgrade`\] Catch usages of `"%s" % var` and provide an unsafe fix (`UP031`) ([#11019](https://github.com/astral-sh/ruff/pull/11019))
+- \[`refurb`\] Implement new rule that suggests min/max over `sorted()` (`FURB192`) ([#10868](https://github.com/astral-sh/ruff/pull/10868))
+
+### Server
+
+- Fix an issue with missing diagnostics for Neovim and Helix ([#11092](https://github.com/astral-sh/ruff/pull/11092))
+- Implement hover documentation for `noqa` codes ([#11096](https://github.com/astral-sh/ruff/pull/11096))
+- Introduce common Ruff configuration options with new server settings ([#11062](https://github.com/astral-sh/ruff/pull/11062))
+
+### Bug fixes
+
+- Use `macos-12` for building release wheels to enable macOS 11 compatibility ([#11146](https://github.com/astral-sh/ruff/pull/11146))
+- \[`flake8-blind-expect`\] Allow raise from in `BLE001` ([#11131](https://github.com/astral-sh/ruff/pull/11131))
+- \[`flake8-pyi`\] Allow simple assignments to `None` in enum class scopes (`PYI026`) ([#11128](https://github.com/astral-sh/ruff/pull/11128))
+- \[`flake8-simplify`\] Avoid raising `SIM911` for non-`zip` attribute calls ([#11126](https://github.com/astral-sh/ruff/pull/11126))
+- \[`refurb`\] Avoid `operator.itemgetter` suggestion for single-item tuple ([#11095](https://github.com/astral-sh/ruff/pull/11095))
+- \[`ruff`\] Respect per-file-ignores for `RUF100` with no other diagnostics ([#11058](https://github.com/astral-sh/ruff/pull/11058))
+- \[`ruff`\] Fix async comprehension false positive (`RUF029`) ([#11070](https://github.com/astral-sh/ruff/pull/11070))
+
+### Documentation
+
+- \[`flake8-bugbear`\] Document explicitly disabling strict zip (`B905`) ([#11040](https://github.com/astral-sh/ruff/pull/11040))
+- \[`flake8-type-checking`\] Mention `lint.typing-modules` in `TCH001`, `TCH002`, and `TCH003` ([#11144](https://github.com/astral-sh/ruff/pull/11144))
+- \[`isort`\] Improve documentation around custom `isort` sections ([#11050](https://github.com/astral-sh/ruff/pull/11050))
+- \[`pylint`\] Fix documentation oversight for `invalid-X-returns` ([#11094](https://github.com/astral-sh/ruff/pull/11094))
+
+### Performance
+
+- Use `matchit` to resolve per-file settings ([#11111](https://github.com/astral-sh/ruff/pull/11111))
+
+## 0.4.1
+
+### Preview features
+
+- \[`pylint`\] Implement `invalid-hash-returned` (`PLE0309`) ([#10961](https://github.com/astral-sh/ruff/pull/10961))
+- \[`pylint`\] Implement `invalid-index-returned` (`PLE0305`) ([#10962](https://github.com/astral-sh/ruff/pull/10962))
+
+### Bug fixes
+
+- \[`pylint`\] Allow `NoReturn`-like functions for `__str__`, `__len__`, etc. (`PLE0307`) ([#11017](https://github.com/astral-sh/ruff/pull/11017))
+- Parser: Use empty range when there's "gap" in token source ([#11032](https://github.com/astral-sh/ruff/pull/11032))
+- \[`ruff`\] Ignore stub functions in `unused-async` (`RUF029`) ([#11026](https://github.com/astral-sh/ruff/pull/11026))
+- Parser: Expect indented case block instead of match stmt ([#11033](https://github.com/astral-sh/ruff/pull/11033))
+
+## 0.4.0
+
+### A new, hand-written parser
+
+Ruff's new parser is **>2x faster**, which translates to a **20-40% speedup** for all linting and formatting invocations.
+There's a lot to say about this exciting change, so check out the [blog post](https://astral.sh/blog/ruff-v0.4.0) for more details!
+
+See [#10036](https://github.com/astral-sh/ruff/pull/10036) for implementation details.
+
+### A new language server in Rust
+
+With this release, we also want to highlight our new language server. `ruff server` is a Rust-powered language
+server that comes built-in with Ruff. It can be used with any editor that supports the [Language Server Protocol](https://microsoft.github.io/language-server-protocol/) (LSP).
+It uses a multi-threaded, lock-free architecture inspired by `rust-analyzer` and it will open the door for a lot
+of exciting features. It’s also faster than our previous [Python-based language server](https://github.com/astral-sh/ruff-lsp)
+-- but you probably guessed that already.
+
+`ruff server` is only in alpha, but it has a lot of features that you can try out today:
+
+- Lints Python files automatically and shows quick-fixes when available
+- Formats Python files, with support for range formatting
+- Comes with commands for quickly performing actions: `ruff.applyAutofix`, `ruff.applyFormat`, and `ruff.applyOrganizeImports`
+- Supports `source.fixAll` and `source.organizeImports` source actions
+- Automatically reloads your project configuration when you change it
+
+To setup `ruff server` with your editor, refer to the [README.md](https://github.com/astral-sh/ruff/blob/main/crates/ruff_server/README.md).
+
+### Preview features
+
+- \[`pycodestyle`\] Do not trigger `E3` rules on `def`s following a function/method with a dummy body ([#10704](https://github.com/astral-sh/ruff/pull/10704))
+- \[`pylint`\] Implement `invalid-bytes-returned` (`E0308`) ([#10959](https://github.com/astral-sh/ruff/pull/10959))
+- \[`pylint`\] Implement `invalid-length-returned` (`E0303`) ([#10963](https://github.com/astral-sh/ruff/pull/10963))
+- \[`pylint`\] Implement `self-cls-assignment` (`W0642`) ([#9267](https://github.com/astral-sh/ruff/pull/9267))
+- \[`pylint`\] Omit stubs from `invalid-bool` and `invalid-str-return-type` ([#11008](https://github.com/astral-sh/ruff/pull/11008))
+- \[`ruff`\] New rule `unused-async` (`RUF029`) to detect unneeded `async` keywords on functions ([#9966](https://github.com/astral-sh/ruff/pull/9966))
+
+### Rule changes
+
+- \[`flake8-bandit`\] Allow `urllib.request.urlopen` calls with static `Request` argument (`S310`) ([#10964](https://github.com/astral-sh/ruff/pull/10964))
+- \[`flake8-bugbear`\] Treat `raise NotImplemented`-only bodies as stub functions (`B006`) ([#10990](https://github.com/astral-sh/ruff/pull/10990))
+- \[`flake8-slots`\] Respect same-file `Enum` subclasses (`SLOT000`) ([#11006](https://github.com/astral-sh/ruff/pull/11006))
+- \[`pylint`\] Support inverted comparisons (`PLR1730`) ([#10920](https://github.com/astral-sh/ruff/pull/10920))
+
+### Linter
+
+- Improve handling of builtin symbols in linter rules ([#10919](https://github.com/astral-sh/ruff/pull/10919))
+- Improve display of rules in `--show-settings` ([#11003](https://github.com/astral-sh/ruff/pull/11003))
+- Improve inference capabilities of the `BuiltinTypeChecker` ([#10976](https://github.com/astral-sh/ruff/pull/10976))
+- Resolve classes and functions relative to script name ([#10965](https://github.com/astral-sh/ruff/pull/10965))
+- Improve performance of `RuleTable::any_enabled` ([#10971](https://github.com/astral-sh/ruff/pull/10971))
+
+### Server
+
+*This section is devoted to updates for our new language server, written in Rust.*
+
+- Enable ruff-specific source actions ([#10916](https://github.com/astral-sh/ruff/pull/10916))
+- Refreshes diagnostics for open files when file configuration is changed ([#10988](https://github.com/astral-sh/ruff/pull/10988))
+- Important errors are now shown as popups ([#10951](https://github.com/astral-sh/ruff/pull/10951))
+- Introduce settings for directly configuring the linter and formatter ([#10984](https://github.com/astral-sh/ruff/pull/10984))
+- Resolve configuration for each document individually ([#10950](https://github.com/astral-sh/ruff/pull/10950))
+- Write a setup guide for Neovim ([#10987](https://github.com/astral-sh/ruff/pull/10987))
+
+### Configuration
+
+- Add `RUFF_OUTPUT_FILE` environment variable support ([#10992](https://github.com/astral-sh/ruff/pull/10992))
+
+### Bug fixes
+
+- Avoid `non-augmented-assignment` for reversed, non-commutative operators (`PLR6104`) ([#10909](https://github.com/astral-sh/ruff/pull/10909))
+- Limit commutative non-augmented-assignments to primitive data types (`PLR6104`) ([#10912](https://github.com/astral-sh/ruff/pull/10912))
+- Respect `per-file-ignores` for `RUF100` on blanket `# noqa` ([#10908](https://github.com/astral-sh/ruff/pull/10908))
+- Consider `if` expression for parenthesized with items parsing ([#11010](https://github.com/astral-sh/ruff/pull/11010))
+- Consider binary expr for parenthesized with items parsing ([#11012](https://github.com/astral-sh/ruff/pull/11012))
+- Reset `FOR_TARGET` context for all kinds of parentheses ([#11009](https://github.com/astral-sh/ruff/pull/11009))
+
+## 0.3.7
+
+### Preview features
+
+- \[`flake8-bugbear`\] Implement `loop-iterator-mutation` (`B909`) ([#9578](https://github.com/astral-sh/ruff/pull/9578))
+- \[`pylint`\] Implement rule to prefer augmented assignment (`PLR6104`) ([#9932](https://github.com/astral-sh/ruff/pull/9932))
+
+### Bug fixes
+
+- Avoid TOCTOU errors in cache initialization ([#10884](https://github.com/astral-sh/ruff/pull/10884))
+- \[`pylint`\] Recode `nan-comparison` rule to `W0177` ([#10894](https://github.com/astral-sh/ruff/pull/10894))
+- \[`pylint`\] Reverse min-max logic in `if-stmt-min-max` ([#10890](https://github.com/astral-sh/ruff/pull/10890))
+
+## 0.3.6
+
+### Preview features
+
+- \[`pylint`\] Implement `bad-staticmethod-argument` (`PLW0211`) ([#10781](https://github.com/astral-sh/ruff/pull/10781))
+- \[`pylint`\] Implement `if-stmt-min-max` (`PLR1730`, `PLR1731`) ([#10002](https://github.com/astral-sh/ruff/pull/10002))
+- \[`pyupgrade`\] Replace `str,Enum` multiple inheritance with `StrEnum` `UP042` ([#10713](https://github.com/astral-sh/ruff/pull/10713))
+- \[`refurb`\] Implement `if-expr-instead-of-or-operator` (`FURB110`) ([#10687](https://github.com/astral-sh/ruff/pull/10687))
+- \[`refurb`\] Implement `int-on-sliced-str` (`FURB166`) ([#10650](https://github.com/astral-sh/ruff/pull/10650))
+- \[`refurb`\] Implement `write-whole-file` (`FURB103`) ([#10802](https://github.com/astral-sh/ruff/pull/10802))
+- \[`refurb`\] Support `itemgetter` in `reimplemented-operator` (`FURB118`) ([#10526](https://github.com/astral-sh/ruff/pull/10526))
+- \[`flake8_comprehensions`\] Add `sum`/`min`/`max` to unnecessary comprehension check (`C419`) ([#10759](https://github.com/astral-sh/ruff/pull/10759))
+
+### Rule changes
+
+- \[`pydocstyle`\] Require capitalizing docstrings where the first sentence is a single word (`D403`) ([#10776](https://github.com/astral-sh/ruff/pull/10776))
+- \[`pycodestyle`\] Ignore annotated lambdas in class scopes (`E731`) ([#10720](https://github.com/astral-sh/ruff/pull/10720))
+- \[`flake8-pyi`\] Various improvements to PYI034 ([#10807](https://github.com/astral-sh/ruff/pull/10807))
+- \[`flake8-slots`\] Flag subclasses of call-based `typing.NamedTuple`s as well as subclasses of `collections.namedtuple()` (`SLOT002`) ([#10808](https://github.com/astral-sh/ruff/pull/10808))
+- \[`pyflakes`\] Allow forward references in class bases in stub files (`F821`) ([#10779](https://github.com/astral-sh/ruff/pull/10779))
+- \[`pygrep-hooks`\] Improve `blanket-noqa` error message (`PGH004`) ([#10851](https://github.com/astral-sh/ruff/pull/10851))
+
+### CLI
+
+- Support `FORCE_COLOR` env var ([#10839](https://github.com/astral-sh/ruff/pull/10839))
+
+### Configuration
+
+- Support negated patterns in `[extend-]per-file-ignores` ([#10852](https://github.com/astral-sh/ruff/pull/10852))
+
+### Bug fixes
+
+- \[`flake8-import-conventions`\] Accept non-aliased (but correct) import in `unconventional-import-alias` (`ICN001`) ([#10729](https://github.com/astral-sh/ruff/pull/10729))
+- \[`flake8-quotes`\] Add semantic model flag when inside f-string replacement field ([#10766](https://github.com/astral-sh/ruff/pull/10766))
+- \[`pep8-naming`\] Recursively resolve `TypeDicts` for N815 violations ([#10719](https://github.com/astral-sh/ruff/pull/10719))
+- \[`flake8-quotes`\] Respect `Q00*` ignores in `flake8-quotes` rules ([#10728](https://github.com/astral-sh/ruff/pull/10728))
+- \[`flake8-simplify`\] Show negated condition in `needless-bool` diagnostics (`SIM103`) ([#10854](https://github.com/astral-sh/ruff/pull/10854))
+- \[`ruff`\] Use within-scope shadowed bindings in `asyncio-dangling-task` (`RUF006`) ([#10793](https://github.com/astral-sh/ruff/pull/10793))
+- \[`flake8-pytest-style`\] Fix single-tuple conversion in `pytest-parametrize-values-wrong-type` (`PT007`) ([#10862](https://github.com/astral-sh/ruff/pull/10862))
+- \[`flake8-return`\] Ignore assignments to annotated variables in `unnecessary-assign` (`RET504`) ([#10741](https://github.com/astral-sh/ruff/pull/10741))
+- \[`refurb`\] Do not allow any keyword arguments for `read-whole-file` in `rb` mode (`FURB101`) ([#10803](https://github.com/astral-sh/ruff/pull/10803))
+- \[`pylint`\] Don't recommend decorating staticmethods with `@singledispatch` (`PLE1519`, `PLE1520`) ([#10637](https://github.com/astral-sh/ruff/pull/10637))
+- \[`pydocstyle`\] Use section name range for all section-related docstring diagnostics ([#10740](https://github.com/astral-sh/ruff/pull/10740))
+- Respect `# noqa` directives on `__all__` openers ([#10798](https://github.com/astral-sh/ruff/pull/10798))
+
+## 0.3.5
+
+### Preview features
+
+- \[`pylint`\] Implement `modified-iterating-set` (`E4703`) ([#10473](https://github.com/astral-sh/ruff/pull/10473))
+- \[`refurb`\] Implement `for-loop-set-mutations` (`FURB142`) ([#10583](https://github.com/astral-sh/ruff/pull/10583))
+- \[`refurb`\] Implement `unnecessary-from-float` (`FURB164`) ([#10647](https://github.com/astral-sh/ruff/pull/10647))
+- \[`refurb`\] Implement `verbose-decimal-constructor` (`FURB157`) ([#10533](https://github.com/astral-sh/ruff/pull/10533))
+
+### Rule changes
+
+- \[`flake8-comprehensions`\] Handled special case for `C401` which also matches `C416` ([#10596](https://github.com/astral-sh/ruff/pull/10596))
+- \[`flake8-pyi`\] Mark `unaliased-collections-abc-set-import` fix as "safe" for more cases in stub files (`PYI025`) ([#10547](https://github.com/astral-sh/ruff/pull/10547))
+- \[`numpy`\] Add `row_stack` to NumPy 2.0 migration rule ([#10646](https://github.com/astral-sh/ruff/pull/10646))
+- \[`pycodestyle`\] Allow cell magics before an import (`E402`) ([#10545](https://github.com/astral-sh/ruff/pull/10545))
+- \[`pycodestyle`\] Avoid blank line rules for the first logical line in cell ([#10291](https://github.com/astral-sh/ruff/pull/10291))
+
+### Configuration
+
+- Respected nested namespace packages ([#10541](https://github.com/astral-sh/ruff/pull/10541))
+- \[`flake8-boolean-trap`\] Add setting for user defined allowed boolean trap ([#10531](https://github.com/astral-sh/ruff/pull/10531))
+
+### Bug fixes
+
+- Correctly handle references in `__all__` definitions when renaming symbols in autofixes ([#10527](https://github.com/astral-sh/ruff/pull/10527))
+- Track ranges of names inside `__all__` definitions ([#10525](https://github.com/astral-sh/ruff/pull/10525))
+- \[`flake8-bugbear`\] Avoid false positive for usage after `continue` (`B031`) ([#10539](https://github.com/astral-sh/ruff/pull/10539))
+- \[`flake8-copyright`\] Accept commas in default copyright pattern ([#9498](https://github.com/astral-sh/ruff/pull/9498))
+- \[`flake8-datetimez`\] Allow f-strings with `%z` for `DTZ007` ([#10651](https://github.com/astral-sh/ruff/pull/10651))
+- \[`flake8-pytest-style`\] Fix `PT014` autofix for last item in list ([#10532](https://github.com/astral-sh/ruff/pull/10532))
+- \[`flake8-quotes`\] Ignore `Q000`, `Q001` when string is inside forward ref ([#10585](https://github.com/astral-sh/ruff/pull/10585))
+- \[`isort`\] Always place non-relative imports after relative imports ([#10669](https://github.com/astral-sh/ruff/pull/10669))
+- \[`isort`\] Respect Unicode characters in import sorting ([#10529](https://github.com/astral-sh/ruff/pull/10529))
+- \[`pyflakes`\] Fix F821 false negatives when `from __future__ import annotations` is active (attempt 2) ([#10524](https://github.com/astral-sh/ruff/pull/10524))
+- \[`pyflakes`\] Make `unnecessary-lambda` an always-unsafe fix ([#10668](https://github.com/astral-sh/ruff/pull/10668))
+- \[`pylint`\] Fixed false-positive on the rule `PLW1641` (`eq-without-hash`) ([#10566](https://github.com/astral-sh/ruff/pull/10566))
+- \[`ruff`\] Fix panic in unused `# noqa` removal with multi-byte space (`RUF100`) ([#10682](https://github.com/astral-sh/ruff/pull/10682))
+
+### Documentation
+
+- Add PR title format to `CONTRIBUTING.md` ([#10665](https://github.com/astral-sh/ruff/pull/10665))
+- Fix list markup to include blank lines required ([#10591](https://github.com/astral-sh/ruff/pull/10591))
+- Put `flake8-logging` next to the other flake8 plugins in registry ([#10587](https://github.com/astral-sh/ruff/pull/10587))
+- \[`flake8-bandit`\] Update warning message for rule `S305` to address insecure block cipher mode use ([#10602](https://github.com/astral-sh/ruff/pull/10602))
+- \[`flake8-bugbear`\] Document use of anonymous assignment in `useless-expression` ([#10551](https://github.com/astral-sh/ruff/pull/10551))
+- \[`flake8-datetimez`\] Clarify error messages and docs for `DTZ` rules ([#10621](https://github.com/astral-sh/ruff/pull/10621))
+- \[`pycodestyle`\] Use same before vs. after numbers for `space-around-operator` ([#10640](https://github.com/astral-sh/ruff/pull/10640))
+- \[`ruff`\] Change `quadratic-list-summation` docs to use `iadd` consistently ([#10666](https://github.com/astral-sh/ruff/pull/10666))
+
+## 0.3.4
+
+### Preview features
+
+- \[`flake8-simplify`\] Detect implicit `else` cases in `needless-bool` (`SIM103`) ([#10414](https://github.com/astral-sh/ruff/pull/10414))
+- \[`pylint`\] Implement `nan-comparison` (`PLW0117`) ([#10401](https://github.com/astral-sh/ruff/pull/10401))
+- \[`pylint`\] Implement `nonlocal-and-global` (`E115`) ([#10407](https://github.com/astral-sh/ruff/pull/10407))
+- \[`pylint`\] Implement `singledispatchmethod-function` (`PLE5120`) ([#10428](https://github.com/astral-sh/ruff/pull/10428))
+- \[`refurb`\] Implement `list-reverse-copy` (`FURB187`) ([#10212](https://github.com/astral-sh/ruff/pull/10212))
+
+### Rule changes
+
+- \[`flake8-pytest-style`\] Add automatic fix for `pytest-parametrize-values-wrong-type` (`PT007`) ([#10461](https://github.com/astral-sh/ruff/pull/10461))
+- \[`pycodestyle`\] Allow SPDX license headers to exceed the line length (`E501`) ([#10481](https://github.com/astral-sh/ruff/pull/10481))
+
+### Formatter
+
+- Fix unstable formatting for trailing subscript end-of-line comment ([#10492](https://github.com/astral-sh/ruff/pull/10492))
+
+### Bug fixes
+
+- Avoid code comment detection in PEP 723 script tags ([#10464](https://github.com/astral-sh/ruff/pull/10464))
+- Avoid incorrect tuple transformation in single-element case (`C409`) ([#10491](https://github.com/astral-sh/ruff/pull/10491))
+- Bug fix: Prevent fully defined links [`name`](link) from being reformatted ([#10442](https://github.com/astral-sh/ruff/pull/10442))
+- Consider raw source code for `W605` ([#10480](https://github.com/astral-sh/ruff/pull/10480))
+- Docs: Link inline settings when not part of options section ([#10499](https://github.com/astral-sh/ruff/pull/10499))
+- Don't treat annotations as redefinitions in `.pyi` files ([#10512](https://github.com/astral-sh/ruff/pull/10512))
+- Fix `E231` bug: Inconsistent catch compared to pycodestyle, such as when dict nested in list ([#10469](https://github.com/astral-sh/ruff/pull/10469))
+- Fix pylint upstream categories not showing in docs ([#10441](https://github.com/astral-sh/ruff/pull/10441))
+- Add missing `Options` references to blank line docs ([#10498](https://github.com/astral-sh/ruff/pull/10498))
+- 'Revert "F821: Fix false negatives in .py files when `from __future__ import annotations` is active (#10362)"' ([#10513](https://github.com/astral-sh/ruff/pull/10513))
+- Apply NFKC normalization to unicode identifiers in the lexer ([#10412](https://github.com/astral-sh/ruff/pull/10412))
+- Avoid failures due to non-deterministic binding ordering ([#10478](https://github.com/astral-sh/ruff/pull/10478))
+- \[`flake8-bugbear`\] Allow tuples of exceptions (`B030`) ([#10437](https://github.com/astral-sh/ruff/pull/10437))
+- \[`flake8-quotes`\] Avoid syntax errors due to invalid quotes (`Q000, Q002`) ([#10199](https://github.com/astral-sh/ruff/pull/10199))
+
+## 0.3.3
+
+### Preview features
+
+- \[`flake8-bandit`\]: Implement `S610` rule ([#10316](https://github.com/astral-sh/ruff/pull/10316))
+- \[`pycodestyle`\] Implement `blank-line-at-end-of-file` (`W391`) ([#10243](https://github.com/astral-sh/ruff/pull/10243))
+- \[`pycodestyle`\] Implement `redundant-backslash` (`E502`) ([#10292](https://github.com/astral-sh/ruff/pull/10292))
+- \[`pylint`\] - implement `redeclared-assigned-name` (`W0128`) ([#9268](https://github.com/astral-sh/ruff/pull/9268))
+
+### Rule changes
+
+- \[`flake8_comprehensions`\] Handled special case for `C400` which also matches `C416` ([#10419](https://github.com/astral-sh/ruff/pull/10419))
+- \[`flake8-bandit`\] Implement upstream updates for `S311`, `S324` and `S605` ([#10313](https://github.com/astral-sh/ruff/pull/10313))
+- \[`pyflakes`\] Remove `F401` fix for `__init__` imports by default and allow opt-in to unsafe fix ([#10365](https://github.com/astral-sh/ruff/pull/10365))
+- \[`pylint`\] Implement `invalid-bool-return-type` (`E304`) ([#10377](https://github.com/astral-sh/ruff/pull/10377))
+- \[`pylint`\] Include builtin warnings in useless-exception-statement (`PLW0133`) ([#10394](https://github.com/astral-sh/ruff/pull/10394))
+
+### CLI
+
+- Add message on success to `ruff check` ([#8631](https://github.com/astral-sh/ruff/pull/8631))
+
+### Bug fixes
+
+- \[`PIE970`\] Allow trailing ellipsis in `typing.TYPE_CHECKING` ([#10413](https://github.com/astral-sh/ruff/pull/10413))
+- Avoid `TRIO115` if the argument is a variable ([#10376](https://github.com/astral-sh/ruff/pull/10376))
+- \[`F811`\] Avoid removing shadowed imports that point to different symbols ([#10387](https://github.com/astral-sh/ruff/pull/10387))
+- Fix `F821` and `F822` false positives in `.pyi` files ([#10341](https://github.com/astral-sh/ruff/pull/10341))
+- Fix `F821` false negatives in `.py` files when `from __future__ import annotations` is active ([#10362](https://github.com/astral-sh/ruff/pull/10362))
+- Fix case where `Indexer` fails to identify continuation preceded by newline #10351 ([#10354](https://github.com/astral-sh/ruff/pull/10354))
+- Sort hash maps in `Settings` display ([#10370](https://github.com/astral-sh/ruff/pull/10370))
+- Track conditional deletions in the semantic model ([#10415](https://github.com/astral-sh/ruff/pull/10415))
+- \[`C413`\] Wrap expressions in parentheses when negating ([#10346](https://github.com/astral-sh/ruff/pull/10346))
+- \[`pycodestyle`\] Do not ignore lines before the first logical line in blank lines rules. ([#10382](https://github.com/astral-sh/ruff/pull/10382))
+- \[`pycodestyle`\] Do not trigger `E225` and `E275` when the next token is a ')' ([#10315](https://github.com/astral-sh/ruff/pull/10315))
+- \[`pylint`\] Avoid false-positive slot non-assignment for `__dict__` (`PLE0237`) ([#10348](https://github.com/astral-sh/ruff/pull/10348))
+- Gate f-string struct size test for Rustc \< 1.76 ([#10371](https://github.com/astral-sh/ruff/pull/10371))
+
+### Documentation
+
+- Use `ruff.toml` format in README ([#10393](https://github.com/astral-sh/ruff/pull/10393))
+- \[`RUF008`\] Make it clearer that a mutable default in a dataclass is only valid if it is typed as a ClassVar ([#10395](https://github.com/astral-sh/ruff/pull/10395))
+- \[`pylint`\] Extend docs and test in `invalid-str-return-type` (`E307`) ([#10400](https://github.com/astral-sh/ruff/pull/10400))
+- Remove `.` from `check` and `format` commands ([#10217](https://github.com/astral-sh/ruff/pull/10217))
+
+## 0.3.2
+
+### Preview features
+
+- Improve single-`with` item formatting for Python 3.8 or older ([#10276](https://github.com/astral-sh/ruff/pull/10276))
+
+### Rule changes
+
+- \[`pyupgrade`\] Allow fixes for f-string rule regardless of line length (`UP032`) ([#10263](https://github.com/astral-sh/ruff/pull/10263))
+- \[`pycodestyle`\] Include actual conditions in E712 diagnostics ([#10254](https://github.com/astral-sh/ruff/pull/10254))
+
+### Bug fixes
+
+- Fix trailing kwargs end of line comment after slash ([#10297](https://github.com/astral-sh/ruff/pull/10297))
+- Fix unstable `with` items formatting ([#10274](https://github.com/astral-sh/ruff/pull/10274))
+- Avoid repeating function calls in f-string conversions ([#10265](https://github.com/astral-sh/ruff/pull/10265))
+- Fix E203 false positive for slices in format strings ([#10280](https://github.com/astral-sh/ruff/pull/10280))
+- Fix incorrect `Parameter` range for `*args` and `**kwargs` ([#10283](https://github.com/astral-sh/ruff/pull/10283))
+- Treat `typing.Annotated` subscripts as type definitions ([#10285](https://github.com/astral-sh/ruff/pull/10285))
+
+## 0.3.1
+
+### Preview features
+
+- \[`pycodestyle`\] Fix E301 not triggering on decorated methods. ([#10117](https://github.com/astral-sh/ruff/pull/10117))
+- \[`pycodestyle`\] Respect `isort` settings in blank line rules (`E3*`) ([#10096](https://github.com/astral-sh/ruff/pull/10096))
+- \[`pycodestyle`\] Make blank lines in typing stub files optional (`E3*`) ([#10098](https://github.com/astral-sh/ruff/pull/10098))
+- \[`pylint`\] Implement `singledispatch-method` (`E1519`) ([#10140](https://github.com/astral-sh/ruff/pull/10140))
+- \[`pylint`\] Implement `useless-exception-statement` (`W0133`) ([#10176](https://github.com/astral-sh/ruff/pull/10176))
+
+### Rule changes
+
+- \[`flake8-debugger`\] Check for use of `debugpy` and `ptvsd` debug modules (#10177) ([#10194](https://github.com/astral-sh/ruff/pull/10194))
+- \[`pyupgrade`\] Generate diagnostic for all valid f-string conversions regardless of line length (`UP032`) ([#10238](https://github.com/astral-sh/ruff/pull/10238))
+- \[`pep8_naming`\] Add fixes for `N804` and `N805` ([#10215](https://github.com/astral-sh/ruff/pull/10215))
+
+### CLI
+
+- Colorize the output of `ruff format --diff` ([#10110](https://github.com/astral-sh/ruff/pull/10110))
+- Make `--config` and `--isolated` global flags ([#10150](https://github.com/astral-sh/ruff/pull/10150))
+- Correctly expand tildes and environment variables in paths passed to `--config` ([#10219](https://github.com/astral-sh/ruff/pull/10219))
+
+### Configuration
+
+- Accept a PEP 440 version specifier for `required-version` ([#10216](https://github.com/astral-sh/ruff/pull/10216))
+- Implement isort's `default-section` setting ([#10149](https://github.com/astral-sh/ruff/pull/10149))
+
+### Bug fixes
+
+- Remove trailing space from `CapWords` message ([#10220](https://github.com/astral-sh/ruff/pull/10220))
+- Respect external codes in file-level exemptions ([#10203](https://github.com/astral-sh/ruff/pull/10203))
+- \[`flake8-raise`\] Avoid false-positives for parens-on-raise with `future.exception()` (`RSE102`) ([#10206](https://github.com/astral-sh/ruff/pull/10206))
+- \[`pylint`\] Add fix for unary expressions in `PLC2801` ([#9587](https://github.com/astral-sh/ruff/pull/9587))
+- \[`ruff`\] Fix RUF028 not allowing `# fmt: skip` on match cases ([#10178](https://github.com/astral-sh/ruff/pull/10178))
+
+## 0.3.0
+
+This release introduces the new Ruff formatter 2024.2 style and adds a new lint rule to
+detect invalid formatter suppression comments.
+
+### Preview features
+
+- \[`flake8-bandit`\] Remove suspicious-lxml-import (`S410`) ([#10154](https://github.com/astral-sh/ruff/pull/10154))
+- \[`pycodestyle`\] Allow `os.environ` modifications between imports (`E402`) ([#10066](https://github.com/astral-sh/ruff/pull/10066))
+- \[`pycodestyle`\] Don't warn about a single whitespace character before a comma in a tuple (`E203`) ([#10094](https://github.com/astral-sh/ruff/pull/10094))
+
+### Rule changes
+
+- \[`eradicate`\] Detect commented out `case` statements (`ERA001`) ([#10055](https://github.com/astral-sh/ruff/pull/10055))
+- \[`eradicate`\] Detect single-line code for `try:`, `except:`, etc. (`ERA001`) ([#10057](https://github.com/astral-sh/ruff/pull/10057))
+- \[`flake8-boolean-trap`\] Allow boolean positionals in `__post_init__` ([#10027](https://github.com/astral-sh/ruff/pull/10027))
+- \[`flake8-copyright`\] Allow © in copyright notices ([#10065](https://github.com/astral-sh/ruff/pull/10065))
+- \[`isort`\]: Use one blank line after imports in typing stub files ([#9971](https://github.com/astral-sh/ruff/pull/9971))
+- \[`pylint`\] New Rule `dict-iter-missing-items` (`PLE1141`) ([#9845](https://github.com/astral-sh/ruff/pull/9845))
+- \[`pylint`\] Ignore `sys.version` and `sys.platform` (`PLR1714`) ([#10054](https://github.com/astral-sh/ruff/pull/10054))
+- \[`pyupgrade`\] Detect literals with unary operators (`UP018`) ([#10060](https://github.com/astral-sh/ruff/pull/10060))
+- \[`ruff`\] Expand rule for `list(iterable).pop(0)` idiom (`RUF015`) ([#10148](https://github.com/astral-sh/ruff/pull/10148))
+
+### Formatter
+
+This release introduces the Ruff 2024.2 style, stabilizing the following changes:
+
+- Prefer splitting the assignment's value over the target or type annotation ([#8943](https://github.com/astral-sh/ruff/pull/8943))
+- Remove blank lines before class docstrings ([#9154](https://github.com/astral-sh/ruff/pull/9154))
+- Wrap multiple context managers in `with` parentheses when targeting Python 3.9 or newer ([#9222](https://github.com/astral-sh/ruff/pull/9222))
+- Add a blank line after nested classes with a dummy body (`...`) in typing stub files ([#9155](https://github.com/astral-sh/ruff/pull/9155))
+- Reduce vertical spacing for classes and functions with a dummy (`...`) body ([#7440](https://github.com/astral-sh/ruff/issues/7440), [#9240](https://github.com/astral-sh/ruff/pull/9240))
+- Add a blank line after the module docstring ([#8283](https://github.com/astral-sh/ruff/pull/8283))
+- Parenthesize long type hints in assignments ([#9210](https://github.com/astral-sh/ruff/pull/9210))
+- Preserve indent for single multiline-string call-expressions ([#9673](https://github.com/astral-sh/ruff/pull/9637))
+- Normalize hex escape and unicode escape sequences ([#9280](https://github.com/astral-sh/ruff/pull/9280))
+- Format module docstrings ([#9725](https://github.com/astral-sh/ruff/pull/9725))
+
+### CLI
+
+- Explicitly disallow `extend` as part of a `--config` flag ([#10135](https://github.com/astral-sh/ruff/pull/10135))
+- Remove `build` from the default exclusion list ([#10093](https://github.com/astral-sh/ruff/pull/10093))
+- Deprecate `ruff <path>`, `ruff --explain`, `ruff --clean`, and `ruff --generate-shell-completion` in favor of `ruff check <path>`, `ruff rule`, `ruff clean`, and `ruff generate-shell-completion` ([#10169](https://github.com/astral-sh/ruff/pull/10169))
+- Remove the deprecated CLI option `--format` from `ruff rule` and `ruff linter` ([#10170](https://github.com/astral-sh/ruff/pull/10170))
+
+### Bug fixes
+
+- \[`flake8-bugbear`\] Avoid adding default initializers to stubs (`B006`) ([#10152](https://github.com/astral-sh/ruff/pull/10152))
+- \[`flake8-type-checking`\] Respect runtime-required decorators for function signatures ([#10091](https://github.com/astral-sh/ruff/pull/10091))
+- \[`pycodestyle`\] Mark fixes overlapping with a multiline string as unsafe (`W293`) ([#10049](https://github.com/astral-sh/ruff/pull/10049))
+- \[`pydocstyle`\] Trim whitespace when removing blank lines after section (`D413`) ([#10162](https://github.com/astral-sh/ruff/pull/10162))
+- \[`pylint`\] Delete entire statement, including semicolons (`PLR0203`) ([#10074](https://github.com/astral-sh/ruff/pull/10074))
+- \[`ruff`\] Avoid f-string false positives in `gettext` calls (`RUF027`) ([#10118](https://github.com/astral-sh/ruff/pull/10118))
+- Fix `ruff` crashing on PowerPC systems because of too small page size ([#10080](https://github.com/astral-sh/ruff/pull/10080))
+
+### Performance
+
+- Add cold attribute to less likely printer queue branches in the formatter ([#10121](https://github.com/astral-sh/ruff/pull/10121))
+- Skip unnecessary string normalization in the formatter ([#10116](https://github.com/astral-sh/ruff/pull/10116))
+
+### Documentation
+
+- Remove "Beta" Label from formatter documentation ([#10144](https://github.com/astral-sh/ruff/pull/10144))
+- `line-length` option: fix link to `pycodestyle.max-line-length` ([#10136](https://github.com/astral-sh/ruff/pull/10136))
+
+## 0.2.2
+
+Highlights include:
+
+- Initial support formatting f-strings (in `--preview`).
+- Support for overriding arbitrary configuration options via the CLI through an expanded `--config` argument (e.g., `--config "lint.isort.combine-as-imports=false"`).
+- Significant performance improvements in Ruff's lexer, parser, and lint rules.
+
+### Preview features
+
+- Implement minimal f-string formatting ([#9642](https://github.com/astral-sh/ruff/pull/9642))
+- \[`pycodestyle`\] Add blank line(s) rules (`E301`, `E302`, `E303`, `E304`, `E305`, `E306`) ([#9266](https://github.com/astral-sh/ruff/pull/9266))
+- \[`refurb`\] Implement `readlines_in_for` (`FURB129`) ([#9880](https://github.com/astral-sh/ruff/pull/9880))
+
+### Rule changes
+
+- \[`ruff`\] Ensure closing parentheses for multiline sequences are always on their own line (`RUF022`, `RUF023`) ([#9793](https://github.com/astral-sh/ruff/pull/9793))
+- \[`numpy`\] Add missing deprecation violations (`NPY002`) ([#9862](https://github.com/astral-sh/ruff/pull/9862))
+- \[`flake8-bandit`\] Detect `mark_safe` usages in decorators ([#9887](https://github.com/astral-sh/ruff/pull/9887))
+- \[`ruff`\] Expand `asyncio-dangling-task` (`RUF006`) to include `new_event_loop` ([#9976](https://github.com/astral-sh/ruff/pull/9976))
+- \[`flake8-pyi`\] Ignore 'unused' private type dicts in class scopes ([#9952](https://github.com/astral-sh/ruff/pull/9952))
+
+### Formatter
+
+- Docstring formatting: Preserve tab indentation when using `indent-style=tabs` ([#9915](https://github.com/astral-sh/ruff/pull/9915))
+- Disable top-level docstring formatting for notebooks ([#9957](https://github.com/astral-sh/ruff/pull/9957))
+- Stabilize quote-style's `preserve` mode ([#9922](https://github.com/astral-sh/ruff/pull/9922))
+
+### CLI
+
+- Allow arbitrary configuration options to be overridden via the CLI ([#9599](https://github.com/astral-sh/ruff/pull/9599))
+
+### Bug fixes
+
+- Make `show-settings` filters directory-agnostic ([#9866](https://github.com/astral-sh/ruff/pull/9866))
+- Respect duplicates when rewriting type aliases ([#9905](https://github.com/astral-sh/ruff/pull/9905))
+- Respect tuple assignments in typing analyzer ([#9969](https://github.com/astral-sh/ruff/pull/9969))
+- Use atomic write when persisting cache ([#9981](https://github.com/astral-sh/ruff/pull/9981))
+- Use non-parenthesized range for `DebugText` ([#9953](https://github.com/astral-sh/ruff/pull/9953))
+- \[`flake8-simplify`\] Avoid false positive with `async` for loops (`SIM113`) ([#9996](https://github.com/astral-sh/ruff/pull/9996))
+- \[`flake8-trio`\] Respect `async with` in `timeout-without-await` ([#9859](https://github.com/astral-sh/ruff/pull/9859))
+- \[`perflint`\] Catch a wider range of mutations in `PERF101` ([#9955](https://github.com/astral-sh/ruff/pull/9955))
+- \[`pycodestyle`\] Fix `E30X` panics on blank lines with trailing white spaces ([#9907](https://github.com/astral-sh/ruff/pull/9907))
+- \[`pydocstyle`\] Allow using `parameters` as a subsection header (`D405`) ([#9894](https://github.com/astral-sh/ruff/pull/9894))
+- \[`pydocstyle`\] Fix blank-line docstring rules for module-level docstrings ([#9878](https://github.com/astral-sh/ruff/pull/9878))
+- \[`pylint`\] Accept 0.0 and 1.0 as common magic values (`PLR2004`) ([#9964](https://github.com/astral-sh/ruff/pull/9964))
+- \[`pylint`\] Avoid suggesting set rewrites for non-hashable types ([#9956](https://github.com/astral-sh/ruff/pull/9956))
+- \[`ruff`\] Avoid false negatives with string literals inside of method calls (`RUF027`) ([#9865](https://github.com/astral-sh/ruff/pull/9865))
+- \[`ruff`\] Fix panic on with f-string detection (`RUF027`) ([#9990](https://github.com/astral-sh/ruff/pull/9990))
+- \[`ruff`\] Ignore builtins when detecting missing f-strings ([#9849](https://github.com/astral-sh/ruff/pull/9849))
+
+### Performance
+
+- Use `memchr` for string lexing ([#9888](https://github.com/astral-sh/ruff/pull/9888))
+- Use `memchr` for tab-indentation detection ([#9853](https://github.com/astral-sh/ruff/pull/9853))
+- Reduce `Result<Tok, LexicalError>` size by using `Box<str>` instead of `String` ([#9885](https://github.com/astral-sh/ruff/pull/9885))
+- Reduce size of `Expr` from 80 to 64 bytes ([#9900](https://github.com/astral-sh/ruff/pull/9900))
+- Improve trailing comma rule performance ([#9867](https://github.com/astral-sh/ruff/pull/9867))
+- Remove unnecessary string cloning from the parser ([#9884](https://github.com/astral-sh/ruff/pull/9884))
+
 ## 0.2.1
 
 This release includes support for range formatting (i.e., the ability to format specific lines
@@ -593,7 +1084,7 @@ docstrings via the `docstring-code-format` setting.
 - \[`pylint`\] Default `max-positional-args` to `max-args` ([#8998](https://github.com/astral-sh/ruff/pull/8998))
 - \[`pylint`\] Add `allow-dunder-method-names` setting for `bad-dunder-method-name` (`PLW3201`) ([#8812](https://github.com/astral-sh/ruff/pull/8812))
 - \[`isort`\] Add support for `from-first` setting ([#8663](https://github.com/astral-sh/ruff/pull/8663))
-- \[`isort`\]  Add support for `length-sort` settings ([#8841](https://github.com/astral-sh/ruff/pull/8841))
+- \[`isort`\] Add support for `length-sort` settings ([#8841](https://github.com/astral-sh/ruff/pull/8841))
 
 ### Bug fixes
 
@@ -722,7 +1213,7 @@ docstrings via the `docstring-code-format` setting.
 - \[`flake8-trio`\] Implement `TRIO115` ([#8486](https://github.com/astral-sh/ruff/pull/8486))
 - \[`refurb`\] Implement `type-none-comparison` (`FURB169`) ([#8487](https://github.com/astral-sh/ruff/pull/8487))
 - Flag all comparisons against builtin types in `E721` ([#8491](https://github.com/astral-sh/ruff/pull/8491))
-- Make `SIM118` fix as safe when the expression is a known dictionary  ([#8525](https://github.com/astral-sh/ruff/pull/8525))
+- Make `SIM118` fix as safe when the expression is a known dictionary ([#8525](https://github.com/astral-sh/ruff/pull/8525))
 
 ### Formatter
 
@@ -890,7 +1381,7 @@ Try it today with `ruff format`! [Check out the blog post](https://astral.sh/blo
 - Add `backports.strenum` to `deprecated-imports` ([#8113](https://github.com/astral-sh/ruff/pull/8113))
 - Update `SIM112` to ignore `https_proxy`, `http_proxy`, and `no_proxy` ([#8140](https://github.com/astral-sh/ruff/pull/8140))
 - Update fix for `literal-membership` (`PLR6201`) to be unsafe ([#8097](https://github.com/astral-sh/ruff/pull/8097))
-- Update fix for  `mutable-argument-defaults` (`B006`) to be unsafe ([#8108](https://github.com/astral-sh/ruff/pull/8108))
+- Update fix for `mutable-argument-defaults` (`B006`) to be unsafe ([#8108](https://github.com/astral-sh/ruff/pull/8108))
 
 ### Formatter
 
@@ -1018,7 +1509,7 @@ Read Ruff's new [versioning policy](https://docs.astral.sh/ruff/versioning/).
 - \[`refurb`\] Add `single-item-membership-test` (`FURB171`) ([#7815](https://github.com/astral-sh/ruff/pull/7815))
 - \[`pylint`\] Add `and-or-ternary` (`R1706`) ([#7811](https://github.com/astral-sh/ruff/pull/7811))
 
-_New rules are added in [preview](https://docs.astral.sh/ruff/preview/)._
+*New rules are added in [preview](https://docs.astral.sh/ruff/preview/).*
 
 ### Configuration
 

CONTRIBUTING.md

@@ -26,30 +26,25 @@ Welcome! We're happy to have you here. Thank you in advance for your contributio
 - [`cargo dev`](#cargo-dev)
 - [Subsystems](#subsystems)
     - [Compilation Pipeline](#compilation-pipeline)
+    - [Import Categorization](#import-categorization)
+        - [Project root](#project-root)
+        - [Package root](#package-root)
+        - [Import categorization](#import-categorization-1)
 
 ## The Basics
 
-Ruff welcomes contributions in the form of Pull Requests.
+Ruff welcomes contributions in the form of pull requests.
 
 For small changes (e.g., bug fixes), feel free to submit a PR.
 
 For larger changes (e.g., new lint rules, new functionality, new configuration options), consider
 creating an [**issue**](https://github.com/astral-sh/ruff/issues) outlining your proposed change.
-You can also join us on [**Discord**](https://discord.gg/c9MhzV8aU5) to discuss your idea with the
+You can also join us on [Discord](https://discord.com/invite/astral-sh) to discuss your idea with the
 community. We've labeled [beginner-friendly tasks](https://github.com/astral-sh/ruff/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22)
 in the issue tracker, along with [bugs](https://github.com/astral-sh/ruff/issues?q=is%3Aissue+is%3Aopen+label%3Abug)
 and [improvements](https://github.com/astral-sh/ruff/issues?q=is%3Aissue+is%3Aopen+label%3Aaccepted)
 that are ready for contributions.
 
-If you're looking for a place to start, we recommend implementing a new lint rule (see:
-[_Adding a new lint rule_](#example-adding-a-new-lint-rule), which will allow you to learn from and
-pattern-match against the examples in the existing codebase. Many lint rules are inspired by
-existing Python plugins, which can be used as a reference implementation.
-
-As a concrete example: consider taking on one of the rules from the [`flake8-pyi`](https://github.com/astral-sh/ruff/issues/848)
-plugin, and looking to the originating [Python source](https://github.com/PyCQA/flake8-pyi) for
-guidance.
-
 If you have suggestions on how we might improve the contributing documentation, [let us know](https://github.com/astral-sh/ruff/discussions/5693)!
 
 ### Prerequisites
@@ -63,7 +58,7 @@ You'll also need [Insta](https://insta.rs/docs/) to update snapshot tests:
 cargo install cargo-insta
 ```
 
-and pre-commit to run some validation checks:
+And you'll need pre-commit to run some validation checks:
 
 ```shell
 pipx install pre-commit  # or `pip install pre-commit` if you have a virtualenv
@@ -76,6 +71,16 @@ when making a commit:
 pre-commit install
 ```
 
+We recommend [nextest](https://nexte.st/) to run Ruff's test suite (via `cargo nextest run`),
+though it's not strictly necessary:
+
+```shell
+cargo install cargo-nextest --locked
+```
+
+Throughout this guide, any usages of `cargo test` can be replaced with `cargo nextest run`,
+if you choose to install `nextest`.
+
 ### Development
 
 After cloning the repository, run Ruff locally from the repository root with:
@@ -93,7 +98,7 @@ RUFF_UPDATE_SCHEMA=1 cargo test  # Rust testing and updating ruff.schema.json
 pre-commit run --all-files --show-diff-on-failure  # Rust and Python formatting, Markdown and Python linting, etc.
 ```
 
-These checks will run on GitHub Actions when you open your Pull Request, but running them locally
+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.
 
 Note that many code changes also require updating the snapshot tests, which is done interactively
@@ -103,7 +108,14 @@ after running `cargo test` like so:
 cargo insta review
 ```
 
-Your Pull Request will be reviewed by a maintainer, which may involve a few rounds of iteration
+If your pull request relates to a specific lint rule, include the category and rule code in the
+title, as in the following examples:
+
+- \[`flake8-bugbear`\] Avoid false positive for usage after `continue` (`B031`)
+- \[`flake8-simplify`\] Detect implicit `else` cases in `needless-bool` (`SIM103`)
+- \[`pycodestyle`\] Implement `redundant-backslash` (`E502`)
+
+Your pull request will be reviewed by a maintainer, which may involve a few rounds of iteration
 prior to merging.
 
 ### Project Structure
@@ -111,8 +123,8 @@ prior to merging.
 Ruff is structured as a monorepo with a [flat crate structure](https://matklad.github.io/2021/08/22/large-rust-workspaces.html),
 such that all crates are contained in a flat `crates` directory.
 
-The vast majority of the code, including all lint rules, lives in the `ruff` crate (located at
-`crates/ruff_linter`). As a contributor, that's the crate that'll be most relevant to you.
+The vast majority of the code, including all lint rules, lives in the `ruff_linter` crate (located
+at `crates/ruff_linter`). As a contributor, that's the crate that'll be most relevant to you.
 
 At the time of writing, the repository includes the following crates:
 
@@ -185,11 +197,14 @@ and calling out to lint rule analyzer functions as it goes.
 If you need to inspect the AST, you can run `cargo dev print-ast` with a Python file. Grep
 for the `Diagnostic::new` invocations to understand how other, similar rules are implemented.
 
-Once you're satisfied with your code, add tests for your rule. See [rule testing](#rule-testing-fixtures-and-snapshots)
-for more details.
+Once you're satisfied with your code, add tests for your rule
+(see: [rule testing](#rule-testing-fixtures-and-snapshots)), and regenerate the documentation and
+associated assets (like our JSON Schema) with `cargo dev generate-all`.
 
-Finally, regenerate the documentation and other generated assets (like our JSON Schema) with:
-`cargo dev generate-all`.
+Finally, submit a pull request, and include the category, rule name, and rule code in the title, as
+in:
+
+> \[`pycodestyle`\] Implement `redundant-backslash` (`E502`)
 
 #### Rule naming convention
 
@@ -302,7 +317,7 @@ To preview any changes to the documentation locally:
     ```
 
 The documentation should then be available locally at
-[http://127.0.0.1:8000/docs/](http://127.0.0.1:8000/docs/).
+[http://127.0.0.1:8000/ruff/](http://127.0.0.1:8000/ruff/).
 
 ## Release Process
 
@@ -315,13 +330,13 @@ even patch releases may contain [non-backwards-compatible changes](https://semve
 
 ### Creating a new release
 
-We use an experimental in-house tool for managing releases.
-
-1. Install `rooster`: `pip install git+https://github.com/zanieb/rooster@main`
-1. Run `rooster release`; this command will:
+1. Install `uv`: `curl -LsSf https://astral.sh/uv/install.sh | sh`
+1. Run `./scripts/release/bump.sh`; this command will:
+    - Generate a temporary virtual environment with `rooster`
     - Generate a changelog entry in `CHANGELOG.md`
     - Update versions in `pyproject.toml` and `Cargo.toml`
     - Update references to versions in the `README.md` and documentation
+    - Display contributors for the release
 1. The changelog should then be editorialized for consistency
     - Often labels will be missing from pull requests they will need to be manually organized into the proper section
     - Changes should be edited to be user-facing descriptions, avoiding internal details
@@ -345,7 +360,7 @@ We use an experimental in-house tool for managing releases.
     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. Generate the contributor list with `rooster contributors` and add to the release notes
+    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
         `git diff old-version-tag new-version-tag -- ruff.schema.json` returns a non-empty diff.
@@ -373,6 +388,11 @@ We have several ways of benchmarking and profiling Ruff:
 - Microbenchmarks which run the linter or the formatter on individual files. These run on pull requests.
 - Profiling the linter on either the microbenchmarks or entire projects
 
+> \[!NOTE\]
+> When running benchmarks, ensure that your CPU is otherwise idle (e.g., close any background
+> applications, like web browsers). You may also want to switch your CPU to a "performance"
+> mode, if it exists, especially when benchmarking short-lived processes.
+
 ### CPython Benchmark
 
 First, clone [CPython](https://github.com/python/cpython). It's a large and diverse Python codebase,
@@ -794,8 +814,8 @@ To understand Ruff's import categorization system, we first need to define two c
     "project root".)
 - "Package root": The top-most directory defining the Python package that includes a given Python
     file. To find the package root for a given Python file, traverse up its parent directories until
-    you reach a parent directory that doesn't contain an `__init__.py` file (and isn't marked as
-    a [namespace package](https://docs.astral.sh/ruff/settings/#namespace-packages)); take the directory
+    you reach a parent directory that doesn't contain an `__init__.py` file (and isn't in a subtree
+    marked as a [namespace package](https://docs.astral.sh/ruff/settings/#namespace-packages)); take the directory
     just before that, i.e., the first directory in the package.
 
 For example, given:

Cargo.toml

@@ -12,102 +12,113 @@ authors = ["Charlie Marsh <charlie.r.marsh@gmail.com>"]
 license = "MIT"
 
 [workspace.dependencies]
-aho-corasick = { version = "1.1.2" }
+aho-corasick = { version = "1.1.3" }
 annotate-snippets = { version = "0.9.2", features = ["color"] }
-anyhow = { version = "1.0.79" }
-argfile = { version = "0.1.6" }
-assert_cmd = { version = "2.0.13" }
+anyhow = { version = "1.0.80" }
+argfile = { version = "0.2.0" }
 bincode = { version = "1.3.3" }
-bitflags = { version = "2.4.1" }
+bitflags = { version = "2.5.0" }
+bstr = { version = "1.9.1" }
 cachedir = { version = "0.3.1" }
-chrono = { version = "0.4.33", default-features = false, features = ["clock"] }
-clap = { version = "4.4.18", features = ["derive"] }
+chrono = { version = "0.4.35", default-features = false, features = ["clock"] }
+clap = { version = "4.5.3", features = ["derive"] }
 clap_complete_command = { version = "0.5.1" }
-clearscreen = { version = "2.0.0" }
-codspeed-criterion-compat = { version = "2.3.3", default-features = false }
+clearscreen = { version = "3.0.0" }
+codspeed-criterion-compat = { version = "2.6.0", default-features = false }
 colored = { version = "2.1.0" }
-configparser = { version = "3.0.3" }
 console_error_panic_hook = { version = "0.1.7" }
 console_log = { version = "1.0.0" }
-countme = { version ="3.0.1"}
+countme = { version = "3.0.1" }
 criterion = { version = "0.5.1", default-features = false }
+crossbeam = { version = "0.8.4" }
+dashmap = { version = "5.5.3" }
 dirs = { version = "5.0.0" }
 drop_bomb = { version = "0.1.5" }
-env_logger = { version ="0.10.1"}
+env_logger = { version = "0.11.0" }
 fern = { version = "0.6.1" }
 filetime = { version = "0.2.23" }
-fs-err = { version ="2.11.0"}
+fs-err = { version = "2.11.0" }
 glob = { version = "0.3.1" }
 globset = { version = "0.4.14" }
-hexf-parse = { version ="0.2.1"}
+hashbrown = "0.14.3"
+hexf-parse = { version = "0.2.1" }
 ignore = { version = "0.4.22" }
-imara-diff ={ version = "0.1.5"}
+imara-diff = { version = "0.1.5" }
 imperative = { version = "1.0.4" }
-indicatif ={ version = "0.17.7"}
-indoc ={ version = "2.0.4"}
-insta = { version = "1.34.0", feature = ["filters", "glob"] }
-insta-cmd = { version = "0.4.0" }
+indexmap = { version = "2.2.6" }
+indicatif = { version = "0.17.8" }
+indoc = { version = "2.0.4" }
+insta = { version = "1.35.1", feature = ["filters", "glob"] }
+insta-cmd = { version = "0.6.0" }
 is-macro = { version = "0.3.5" }
 is-wsl = { version = "0.4.0" }
 itertools = { version = "0.12.1" }
-js-sys = { version = "0.3.67" }
-lalrpop-util = { version = "0.20.0", default-features = false }
+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"] }
+matchit = { version = "0.8.1" }
 memchr = { version = "2.7.1" }
-mimalloc = { version ="0.1.39"}
+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" }
 pathdiff = { version = "0.2.1" }
-pep440_rs = { version = "0.4.0", features = ["serde"] }
+parking_lot = "0.12.1"
+pep440_rs = { version = "0.6.0", features = ["serde"] }
 pretty_assertions = "1.3.0"
-proc-macro2 = { version = "1.0.78" }
-pyproject-toml = { version = "0.8.2" }
-quick-junit = { version = "0.3.5" }
+proc-macro2 = { version = "1.0.79" }
+pyproject-toml = { version = "0.9.0" }
+quick-junit = { version = "0.4.0" }
 quote = { version = "1.0.23" }
 rand = { version = "0.8.5" }
-rayon = { version = "1.8.1" }
+rayon = { version = "1.10.0" }
 regex = { version = "1.10.2" }
 result-like = { version = "0.5.0" }
 rustc-hash = { version = "1.1.0" }
 schemars = { version = "0.8.16" }
-seahash = { version ="4.1.0"}
-semver = { version = "1.0.21" }
-serde = { version = "1.0.196", features = ["derive"] }
-serde-wasm-bindgen = { version = "0.6.3" }
+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"] }
 shellexpand = { version = "3.0.0" }
-shlex = { version ="1.3.0"}
+shlex = { version = "1.3.0" }
 similar = { version = "2.4.0", features = ["inline"] }
-smallvec = { version = "1.13.1" }
+smallvec = { version = "1.13.2" }
 static_assertions = "1.1.0"
-strum = { version = "0.25.0", features = ["strum_macros"] }
-strum_macros = { version = "0.25.3" }
-syn = { version = "2.0.40" }
-tempfile = { version ="3.9.0"}
+strum = { version = "0.26.0", features = ["strum_macros"] }
+strum_macros = { version = "0.26.0" }
+syn = { version = "2.0.55" }
+tempfile = { version = "3.9.0" }
 test-case = { version = "3.3.1" }
-thiserror = { version = "1.0.51" }
-tikv-jemallocator = { version ="0.5.0"}
-toml = { version = "0.8.9" }
+thiserror = { version = "1.0.58" }
+tikv-jemallocator = { version = "0.5.0" }
+toml = { version = "0.8.11" }
 tracing = { version = "0.1.40" }
 tracing-indicatif = { version = "0.3.6" }
 tracing-subscriber = { version = "0.3.18", features = ["env-filter"] }
+tracing-tree = { version = "0.3.0" }
 typed-arena = { version = "2.0.2" }
-unic-ucd-category = { version ="0.9"}
+unic-ucd-category = { version = "0.9" }
 unicode-ident = { version = "1.0.12" }
 unicode-width = { version = "0.1.11" }
-unicode_names2 = { version = "1.2.1" }
-ureq = { version = "2.9.1" }
+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"] }
 walkdir = { version = "2.3.2" }
-wasm-bindgen = { version = "0.2.84" }
-wasm-bindgen-test = { version = "0.3.40" }
+wasm-bindgen = { version = "0.2.92" }
+wasm-bindgen-test = { version = "0.3.42" }
 wild = { version = "2" }
 
 [workspace.lints.rust]

README.md

@@ -4,11 +4,12 @@
 
 [![Ruff](https://img.shields.io/endpoint?url=https://raw.githubusercontent.com/astral-sh/ruff/main/assets/badge/v2.json)](https://github.com/astral-sh/ruff)
 [![image](https://img.shields.io/pypi/v/ruff.svg)](https://pypi.python.org/pypi/ruff)
-[![image](https://img.shields.io/pypi/l/ruff.svg)](https://pypi.python.org/pypi/ruff)
+[![image](https://img.shields.io/pypi/l/ruff.svg)](https://github.com/astral-sh/ruff/blob/main/LICENSE)
 [![image](https://img.shields.io/pypi/pyversions/ruff.svg)](https://pypi.python.org/pypi/ruff)
 [![Actions status](https://github.com/astral-sh/ruff/workflows/CI/badge.svg)](https://github.com/astral-sh/ruff/actions)
+[![Discord](https://img.shields.io/badge/Discord-%235865F2.svg?logo=discord&logoColor=white)](https://discord.com/invite/astral-sh)
 
-[**Discord**](https://discord.gg/c9MhzV8aU5) | [**Docs**](https://docs.astral.sh/ruff/) | [**Playground**](https://play.ruff.rs/)
+[**Docs**](https://docs.astral.sh/ruff/) | [**Playground**](https://play.ruff.rs/)
 
 An extremely fast Python linter and code formatter, written in Rust.
 
@@ -31,7 +32,7 @@ An extremely fast Python linter and code formatter, written in Rust.
 - ⚖️ 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)
-- 📏 Over [700 built-in rules](https://docs.astral.sh/ruff/rules/), with native re-implementations
+- 📏 Over [800 built-in rules](https://docs.astral.sh/ruff/rules/), with native re-implementations
     of popular Flake8 plugins, like flake8-bugbear
 - ⌨️ First-party [editor integrations](https://docs.astral.sh/ruff/integrations/) for
     [VS Code](https://github.com/astral-sh/ruff-vscode) and [more](https://github.com/astral-sh/ruff-lsp)
@@ -49,6 +50,7 @@ times faster than any individual tool.
 Ruff is extremely actively developed and used in major open-source projects like:
 
 - [Apache Airflow](https://github.com/apache/airflow)
+- [Apache Superset](https://github.com/apache/superset)
 - [FastAPI](https://github.com/tiangolo/fastapi)
 - [Hugging Face](https://github.com/huggingface/transformers)
 - [Pandas](https://github.com/pandas-dev/pandas)
@@ -128,7 +130,7 @@ and with [a variety of other package managers](https://docs.astral.sh/ruff/insta
 To run Ruff as a linter, try any of the following:
 
 ```shell
-ruff check .                        # Lint all files in the current directory (and any subdirectories).
+ruff check                          # Lint all files in the current directory (and any subdirectories).
 ruff check path/to/code/            # Lint all files in `/path/to/code` (and any subdirectories).
 ruff check path/to/code/*.py        # Lint all `.py` files in `/path/to/code`.
 ruff check path/to/code/to/file.py  # Lint `file.py`.
@@ -138,7 +140,7 @@ ruff check @arguments.txt           # Lint using an input file, treating its con
 Or, to run Ruff as a formatter:
 
 ```shell
-ruff format .                        # Format all files in the current directory (and any subdirectories).
+ruff format                          # Format all files in the current directory (and any subdirectories).
 ruff format path/to/code/            # Format all files in `/path/to/code` (and any subdirectories).
 ruff format path/to/code/*.py        # Format all `.py` files in `/path/to/code`.
 ruff format path/to/code/to/file.py  # Format `file.py`.
@@ -150,7 +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: v0.2.1
+  rev: v0.4.2
   hooks:
     # Run the linter.
     - id: ruff
@@ -172,7 +174,7 @@ jobs:
   ruff:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: chartboost/ruff-action@v1
 ```
 
@@ -182,10 +184,9 @@ Ruff can be configured through a `pyproject.toml`, `ruff.toml`, or `.ruff.toml`
 [_Configuration_](https://docs.astral.sh/ruff/configuration/), or [_Settings_](https://docs.astral.sh/ruff/settings/)
 for a complete list of all configuration options).
 
-If left unspecified, Ruff's default configuration is equivalent to:
+If left unspecified, Ruff's default configuration is equivalent to the following `ruff.toml` file:
 
 ```toml
-[tool.ruff]
 # Exclude a variety of commonly ignored directories.
 exclude = [
     ".bzr",
@@ -223,7 +224,7 @@ indent-width = 4
 # Assume Python 3.8
 target-version = "py38"
 
-[tool.ruff.lint]
+[lint]
 # Enable Pyflakes (`F`) and a subset of the pycodestyle (`E`)  codes by default.
 select = ["E4", "E7", "E9", "F"]
 ignore = []
@@ -235,7 +236,7 @@ unfixable = []
 # Allow unused variables when underscore-prefixed.
 dummy-variable-rgx = "^(_+|(_+[a-zA-Z0-9_]*[a-zA-Z0-9]+?))$"
 
-[tool.ruff.format]
+[format]
 # Like Black, use double quotes for strings.
 quote-style = "double"
 
@@ -249,11 +250,20 @@ skip-magic-trailing-comma = false
 line-ending = "auto"
 ```
 
-Some configuration options can be provided via the command-line, such as those related to
-rule enablement and disablement, file discovery, and logging level:
+Note that, in a `pyproject.toml`, each section header should be prefixed with `tool.ruff`. For
+example, `[lint]` should be replaced with `[tool.ruff.lint]`.
+
+Some configuration options can be provided via dedicated command-line arguments, such as those
+related to rule enablement and disablement, file discovery, and logging level:
 
 ```shell
-ruff check path/to/code/ --select F401 --select F403 --quiet
+ruff check --select F401 --select F403 --quiet
+```
+
+The remaining configuration options can be provided through a catch-all `--config` argument:
+
+```shell
+ruff check --config "lint.per-file-ignores = {'some_file.py' = ['F841']}"
 ```
 
 See `ruff help` for more on Ruff's top-level commands, or `ruff help check` and `ruff help format`
@@ -263,7 +273,7 @@ for more on the linting and formatting commands, respectively.
 
 <!-- Begin section: Rules -->
 
-**Ruff supports over 700 lint rules**, many of which are inspired by popular tools like Flake8,
+**Ruff supports over 800 lint rules**, many of which are inspired by popular tools like Flake8,
 isort, pyupgrade, and others. Regardless of the rule's origin, Ruff re-implements every rule in
 Rust as a first-party feature.
 
@@ -341,14 +351,14 @@ For a complete enumeration of the supported rules, see [_Rules_](https://docs.as
 Contributions are welcome and highly appreciated. To get started, check out the
 [**contributing guidelines**](https://docs.astral.sh/ruff/contributing/).
 
-You can also join us on [**Discord**](https://discord.gg/c9MhzV8aU5).
+You can also join us on [**Discord**](https://discord.com/invite/astral-sh).
 
 ## Support
 
 Having trouble? Check out the existing issues on [**GitHub**](https://github.com/astral-sh/ruff/issues),
 or feel free to [**open a new one**](https://github.com/astral-sh/ruff/issues/new).
 
-You can also ask for help on [**Discord**](https://discord.gg/c9MhzV8aU5).
+You can also ask for help on [**Discord**](https://discord.com/invite/astral-sh).
 
 ## Acknowledgements
 
@@ -378,6 +388,7 @@ Ruff is released under the MIT license.
 
 Ruff is used by a number of major open-source projects and companies, including:
 
+- [Albumentations](https://github.com/albumentations-team/albumentations)
 - Amazon ([AWS SAM](https://github.com/aws/serverless-application-model))
 - Anthropic ([Python SDK](https://github.com/anthropics/anthropic-sdk-python))
 - [Apache Airflow](https://github.com/apache/airflow)
@@ -404,6 +415,7 @@ Ruff is used by a number of major open-source projects and companies, including:
 - [Ibis](https://github.com/ibis-project/ibis)
 - [ivy](https://github.com/unifyai/ivy)
 - [Jupyter](https://github.com/jupyter-server/jupyter_server)
+- [Kraken Tech](https://kraken.tech/)
 - [LangChain](https://github.com/hwchase17/langchain)
 - [Litestar](https://litestar.dev/)
 - [LlamaIndex](https://github.com/jerryjliu/llama_index)
@@ -418,6 +430,7 @@ Ruff is used by a number of major open-source projects and companies, including:
 - [Mypy](https://github.com/python/mypy)
 - Netflix ([Dispatch](https://github.com/Netflix/dispatch))
 - [Neon](https://github.com/neondatabase/neon)
+- [Nokia](https://nokia.com/)
 - [NoneBot](https://github.com/nonebot/nonebot2)
 - [NumPyro](https://github.com/pyro-ppl/numpyro)
 - [ONNX](https://github.com/onnx/onnx)
@@ -486,7 +499,7 @@ If you're using Ruff, consider adding the Ruff badge to your project's `README.m
 
 ## License
 
-MIT
+This repository is licensed under the [MIT License](https://github.com/astral-sh/ruff/blob/main/LICENSE)
 
 <div align="center">
   <a target="_blank" href="https://astral.sh" style="background:none">

_typos.toml

@@ -3,9 +3,17 @@
 extend-exclude = ["**/resources/**/*", "**/snapshots/**/*"]
 
 [default.extend-words]
+"arange" = "arange"  # e.g. `numpy.arange`
 hel = "hel"
 whos = "whos"
 spawnve = "spawnve"
 ned = "ned"
+pn = "pn"  # `import panel as pd` is a thing
 poit = "poit"
 BA = "BA" # acronym for "Bad Allowed", used in testing.
+
+[default]
+extend-ignore-re = [
+  # Line ignore with trailing "spellchecker:disable-line"
+  "(?Rm)^.*#\\s*spellchecker:disable-line$"
+]

crates/red_knot/Cargo.toml

@@ -0,0 +1,47 @@
+[package]
+name = "red_knot"
+version = "0.1.0"
+edition.workspace = true
+rust-version.workspace = true
+homepage.workspace = true
+documentation.workspace = true
+repository.workspace = true
+authors.workspace = true
+license.workspace = true
+
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+
+[dependencies]
+ruff_formatter = { path = "../ruff_formatter" }
+ruff_index = { path = "../ruff_index" }
+ruff_notebook = { path = "../ruff_notebook" }
+ruff_python_ast = { path = "../ruff_python_ast" }
+ruff_python_formatter = { path = "../ruff_python_formatter" }
+ruff_python_parser = { path = "../ruff_python_parser" }
+ruff_python_trivia = { path = "../ruff_python_trivia" }
+ruff_text_size = { path = "../ruff_text_size" }
+
+anyhow = { workspace = true }
+bitflags = { workspace = true }
+ctrlc = "3.4.4"
+crossbeam = { workspace = true }
+dashmap = { workspace = true }
+hashbrown = { workspace = true }
+indexmap = { workspace = true }
+log = { workspace = true }
+notify = { workspace = true }
+parking_lot = { workspace = true }
+rayon = { workspace = true }
+rustc-hash = { workspace = true }
+smallvec = { workspace = true }
+smol_str = "0.2.1"
+tracing = { workspace = true }
+tracing-subscriber = { workspace = true }
+tracing-tree = { workspace = true }
+
+[dev-dependencies]
+textwrap = "0.16.1"
+tempfile = { workspace = true }
+
+[lints]
+workspace = true

crates/red_knot/src/ast_ids.rs

@@ -0,0 +1,415 @@
+use std::any::type_name;
+use std::fmt::{Debug, Formatter};
+use std::hash::{Hash, Hasher};
+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::{
+    AnyNodeRef, AstNode, ExceptHandler, ExceptHandlerExceptHandler, Expr, MatchCase, ModModule,
+    NodeKind, Parameter, Stmt, StmtAnnAssign, StmtAssign, StmtAugAssign, StmtClassDef,
+    StmtFunctionDef, StmtGlobal, StmtImport, StmtImportFrom, StmtNonlocal, StmtTypeAlias,
+    TypeParam, TypeParamParamSpec, TypeParamTypeVar, TypeParamTypeVarTuple, WithItem,
+};
+use ruff_text_size::{Ranged, TextRange};
+
+/// A type agnostic ID that uniquely identifies an AST node in a file.
+#[ruff_index::newtype_index]
+pub struct AstId;
+
+/// A typed ID that uniquely identifies an AST node in a file.
+///
+/// This is different from [`AstId`] in that it is a combination of ID and the type of the node the ID identifies.
+/// Typing the ID prevents mixing IDs of different node types and allows to restrict the API to only accept
+/// nodes for which an ID has been created (not all AST nodes get an ID).
+pub struct TypedAstId<N: HasAstId> {
+    erased: AstId,
+    _marker: PhantomData<fn() -> N>,
+}
+
+impl<N: HasAstId> TypedAstId<N> {
+    /// Upcasts this ID from a more specific node type to a more general node type.
+    pub fn upcast<M: HasAstId>(self) -> TypedAstId<M>
+    where
+        N: Into<M>,
+    {
+        TypedAstId {
+            erased: self.erased,
+            _marker: PhantomData,
+        }
+    }
+}
+
+impl<N: HasAstId> Copy for TypedAstId<N> {}
+impl<N: HasAstId> Clone for TypedAstId<N> {
+    fn clone(&self) -> Self {
+        *self
+    }
+}
+
+impl<N: HasAstId> PartialEq for TypedAstId<N> {
+    fn eq(&self, other: &Self) -> bool {
+        self.erased == other.erased
+    }
+}
+
+impl<N: HasAstId> Eq for TypedAstId<N> {}
+impl<N: HasAstId> Hash for TypedAstId<N> {
+    fn hash<H: Hasher>(&self, state: &mut H) {
+        self.erased.hash(state);
+    }
+}
+
+impl<N: HasAstId> Debug for TypedAstId<N> {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        f.debug_tuple("TypedAstId")
+            .field(&self.erased)
+            .field(&type_name::<N>())
+            .finish()
+    }
+}
+
+pub struct AstIds {
+    ids: IndexVec<AstId, NodeKey>,
+    reverse: FxHashMap<NodeKey, AstId>,
+}
+
+impl AstIds {
+    // TODO rust analyzer doesn't allocate an ID for every node. It only allocates ids for
+    //  nodes with a corresponding HIR element, that is nodes that are definitions.
+    pub fn from_module(module: &ModModule) -> Self {
+        let mut visitor = AstIdsVisitor::default();
+
+        // TODO: visit_module?
+        // Make sure we visit the root
+        visitor.create_id(module);
+        visitor.visit_body(&module.body);
+
+        while let Some(deferred) = visitor.deferred.pop() {
+            match deferred {
+                DeferredNode::FunctionDefinition(def) => {
+                    def.visit_preorder(&mut visitor);
+                }
+                DeferredNode::ClassDefinition(def) => def.visit_preorder(&mut visitor),
+            }
+        }
+
+        AstIds {
+            ids: visitor.ids,
+            reverse: visitor.reverse,
+        }
+    }
+
+    /// Returns the ID to the root node.
+    pub fn root(&self) -> NodeKey {
+        self.ids[AstId::new(0)]
+    }
+
+    /// Returns the [`TypedAstId`] for a node.
+    pub fn ast_id<N: HasAstId>(&self, node: &N) -> TypedAstId<N> {
+        let key = node.syntax_node_key();
+        TypedAstId {
+            erased: self.reverse.get(&key).copied().unwrap(),
+            _marker: PhantomData,
+        }
+    }
+
+    /// Returns the [`TypedAstId`] for the node identified with the given [`TypedNodeKey`].
+    pub fn ast_id_for_key<N: HasAstId>(&self, node: &TypedNodeKey<N>) -> TypedAstId<N> {
+        let ast_id = self.ast_id_for_node_key(node.inner);
+
+        TypedAstId {
+            erased: ast_id,
+            _marker: PhantomData,
+        }
+    }
+
+    /// Returns the untyped [`AstId`] for the node identified by the given `node` key.
+    pub fn ast_id_for_node_key(&self, node: NodeKey) -> AstId {
+        self.reverse
+            .get(&node)
+            .copied()
+            .expect("Can't find node in AstIds map.")
+    }
+
+    /// Returns the [`TypedNodeKey`] for the node identified by the given [`TypedAstId`].
+    pub fn key<N: HasAstId>(&self, id: TypedAstId<N>) -> TypedNodeKey<N> {
+        let syntax_key = self.ids[id.erased];
+
+        TypedNodeKey::new(syntax_key).unwrap()
+    }
+
+    pub fn node_key<H: HasAstId>(&self, id: TypedAstId<H>) -> NodeKey {
+        self.ids[id.erased]
+    }
+}
+
+impl std::fmt::Debug for AstIds {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        let mut map = f.debug_map();
+        for (key, value) in self.ids.iter_enumerated() {
+            map.entry(&key, &value);
+        }
+
+        map.finish()
+    }
+}
+
+impl PartialEq for AstIds {
+    fn eq(&self, other: &Self) -> bool {
+        self.ids == other.ids
+    }
+}
+
+impl Eq for AstIds {}
+
+#[derive(Default)]
+struct AstIdsVisitor<'a> {
+    ids: IndexVec<AstId, NodeKey>,
+    reverse: FxHashMap<NodeKey, AstId>,
+    deferred: Vec<DeferredNode<'a>>,
+}
+
+impl<'a> AstIdsVisitor<'a> {
+    fn create_id<A: HasAstId>(&mut self, node: &A) {
+        let node_key = node.syntax_node_key();
+
+        let id = self.ids.push(node_key);
+        self.reverse.insert(node_key, id);
+    }
+}
+
+impl<'a> PreorderVisitor<'a> for AstIdsVisitor<'a> {
+    fn visit_stmt(&mut self, stmt: &'a Stmt) {
+        match stmt {
+            Stmt::FunctionDef(def) => {
+                self.create_id(def);
+                self.deferred.push(DeferredNode::FunctionDefinition(def));
+                return;
+            }
+            // TODO defer visiting the assignment body, type alias parameters etc?
+            Stmt::ClassDef(def) => {
+                self.create_id(def);
+                self.deferred.push(DeferredNode::ClassDefinition(def));
+                return;
+            }
+            Stmt::Expr(_) => {
+                // Skip
+                return;
+            }
+            Stmt::Return(_) => {}
+            Stmt::Delete(_) => {}
+            Stmt::Assign(assignment) => self.create_id(assignment),
+            Stmt::AugAssign(assignment) => {
+                self.create_id(assignment);
+            }
+            Stmt::AnnAssign(assignment) => self.create_id(assignment),
+            Stmt::TypeAlias(assignment) => self.create_id(assignment),
+            Stmt::For(_) => {}
+            Stmt::While(_) => {}
+            Stmt::If(_) => {}
+            Stmt::With(_) => {}
+            Stmt::Match(_) => {}
+            Stmt::Raise(_) => {}
+            Stmt::Try(_) => {}
+            Stmt::Assert(_) => {}
+            Stmt::Import(import) => self.create_id(import),
+            Stmt::ImportFrom(import_from) => self.create_id(import_from),
+            Stmt::Global(global) => self.create_id(global),
+            Stmt::Nonlocal(non_local) => self.create_id(non_local),
+            Stmt::Pass(_) => {}
+            Stmt::Break(_) => {}
+            Stmt::Continue(_) => {}
+            Stmt::IpyEscapeCommand(_) => {}
+        }
+
+        preorder::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);
+    }
+
+    fn visit_except_handler(&mut self, except_handler: &'a ExceptHandler) {
+        match except_handler {
+            ExceptHandler::ExceptHandler(except_handler) => {
+                self.create_id(except_handler);
+            }
+        }
+
+        preorder::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);
+    }
+
+    fn visit_match_case(&mut self, match_case: &'a MatchCase) {
+        self.create_id(match_case);
+        preorder::walk_match_case(self, match_case);
+    }
+
+    fn visit_type_param(&mut self, type_param: &'a TypeParam) {
+        self.create_id(type_param);
+    }
+}
+
+enum DeferredNode<'a> {
+    FunctionDefinition(&'a StmtFunctionDef),
+    ClassDefinition(&'a StmtClassDef),
+}
+
+#[derive(Copy, Clone, Debug, Eq, PartialEq, Hash)]
+pub struct TypedNodeKey<N: AstNode> {
+    /// The type erased node key.
+    inner: NodeKey,
+    _marker: PhantomData<fn() -> N>,
+}
+
+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(),
+        };
+        Self {
+            inner,
+            _marker: PhantomData,
+        }
+    }
+
+    pub fn new(node_key: NodeKey) -> Option<Self> {
+        N::can_cast(node_key.kind).then_some(TypedNodeKey {
+            inner: node_key,
+            _marker: PhantomData,
+        })
+    }
+
+    pub fn resolve<'a>(&self, root: AnyNodeRef<'a>) -> Option<N::Ref<'a>> {
+        let node_ref = self.inner.resolve(root)?;
+
+        Some(N::cast_ref(node_ref).unwrap())
+    }
+
+    pub fn resolve_unwrap<'a>(&self, root: AnyNodeRef<'a>) -> N::Ref<'a> {
+        self.resolve(root).expect("node should resolve")
+    }
+
+    pub fn erased(&self) -> &NodeKey {
+        &self.inner
+    }
+}
+
+struct FindNodeKeyVisitor<'a> {
+    key: NodeKey,
+    result: Option<AnyNodeRef<'a>>,
+}
+
+impl<'a> PreorderVisitor<'a> for FindNodeKeyVisitor<'a> {
+    fn enter_node(&mut self, node: AnyNodeRef<'a>) -> TraversalSignal {
+        if self.result.is_some() {
+            return TraversalSignal::Skip;
+        }
+
+        if node.range() == self.key.range && node.kind() == self.key.kind {
+            self.result = Some(node);
+            TraversalSignal::Skip
+        } else if node.range().contains_range(self.key.range) {
+            TraversalSignal::Traverse
+        } else {
+            TraversalSignal::Skip
+        }
+    }
+
+    fn visit_body(&mut self, body: &'a [Stmt]) {
+        // TODO it would be more efficient to use binary search instead of linear
+        for stmt in body {
+            if stmt.range().start() > self.key.range.end() {
+                break;
+            }
+
+            self.visit_stmt(stmt);
+        }
+    }
+}
+
+// TODO an alternative to this is to have a `NodeId` on each node (in increasing order depending on the position).
+//  This would allow to reduce the size of this to a u32.
+//  What would be nice if we could use an `Arc::weak_ref` here but that only works if we use
+//   `Arc` internally
+// TODO: Implement the logic to resolve a node, given a db (and the correct file).
+#[derive(Copy, Clone, Debug, Eq, PartialEq, Hash)]
+pub struct NodeKey {
+    kind: NodeKind,
+    range: TextRange,
+}
+
+impl NodeKey {
+    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 {
+            key: *self,
+            result: None,
+        };
+
+        if visitor.enter_node(root) == TraversalSignal::Traverse {
+            root.visit_preorder(&mut visitor);
+        }
+
+        visitor.result
+    }
+}
+
+/// Marker trait implemented by AST nodes for which we extract the `AstId`.
+pub trait HasAstId: AstNode {
+    fn node_key(&self) -> TypedNodeKey<Self>
+    where
+        Self: Sized,
+    {
+        TypedNodeKey {
+            inner: self.syntax_node_key(),
+            _marker: PhantomData,
+        }
+    }
+
+    fn syntax_node_key(&self) -> NodeKey {
+        NodeKey {
+            kind: self.as_any_node_ref().kind(),
+            range: self.range(),
+        }
+    }
+}
+
+impl HasAstId for StmtFunctionDef {}
+impl HasAstId for StmtClassDef {}
+impl HasAstId for StmtAnnAssign {}
+impl HasAstId for StmtAugAssign {}
+impl HasAstId for StmtAssign {}
+impl HasAstId for StmtTypeAlias {}
+
+impl HasAstId for ModModule {}
+
+impl HasAstId for StmtImport {}
+
+impl HasAstId for StmtImportFrom {}
+
+impl HasAstId for Parameter {}
+
+impl HasAstId for TypeParam {}
+impl HasAstId for Stmt {}
+impl HasAstId for TypeParamTypeVar {}
+impl HasAstId for TypeParamTypeVarTuple {}
+impl HasAstId for TypeParamParamSpec {}
+impl HasAstId for StmtGlobal {}
+impl HasAstId for StmtNonlocal {}
+
+impl HasAstId for ExceptHandlerExceptHandler {}
+impl HasAstId for WithItem {}
+impl HasAstId for MatchCase {}

crates/red_knot/src/cache.rs

@@ -0,0 +1,165 @@
+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;
+
+/// Simple key value cache that locks on a per-key level.
+pub struct KeyValueCache<K, V> {
+    map: FxDashMap<K, V>,
+    statistics: CacheStatistics,
+}
+
+impl<K, V> KeyValueCache<K, V>
+where
+    K: Eq + Hash + Clone,
+    V: Clone,
+{
+    pub fn try_get(&self, key: &K) -> Option<V> {
+        if let Some(existing) = self.map.get(key) {
+            self.statistics.hit();
+            Some(existing.clone())
+        } else {
+            self.statistics.miss();
+            None
+        }
+    }
+
+    pub fn get<F>(&self, key: &K, compute: F) -> QueryResult<V>
+    where
+        F: FnOnce(&K) -> QueryResult<V>,
+    {
+        Ok(match self.map.entry(key.clone()) {
+            Entry::Occupied(cached) => {
+                self.statistics.hit();
+
+                cached.get().clone()
+            }
+            Entry::Vacant(vacant) => {
+                self.statistics.miss();
+
+                let value = compute(key)?;
+                vacant.insert(value.clone());
+                value
+            }
+        })
+    }
+
+    pub fn set(&mut self, key: K, value: V) {
+        self.map.insert(key, value);
+    }
+
+    pub fn remove(&mut self, key: &K) -> Option<V> {
+        self.map.remove(key).map(|(_, value)| value)
+    }
+
+    pub fn clear(&mut self) {
+        self.map.clear();
+        self.map.shrink_to_fit();
+    }
+
+    pub fn statistics(&self) -> Option<Statistics> {
+        self.statistics.to_statistics()
+    }
+}
+
+impl<K, V> Default for KeyValueCache<K, V>
+where
+    K: Eq + Hash,
+    V: Clone,
+{
+    fn default() -> Self {
+        Self {
+            map: FxDashMap::default(),
+            statistics: CacheStatistics::default(),
+        }
+    }
+}
+
+impl<K, V> std::fmt::Debug for KeyValueCache<K, V>
+where
+    K: std::fmt::Debug + Eq + Hash,
+    V: std::fmt::Debug,
+{
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        let mut debug = f.debug_map();
+
+        for entry in &self.map {
+            debug.entry(&entry.value(), &entry.key());
+        }
+
+        debug.finish()
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct Statistics {
+    pub hits: usize,
+    pub misses: usize,
+}
+
+impl Statistics {
+    #[allow(clippy::cast_precision_loss)]
+    pub fn hit_rate(&self) -> Option<f64> {
+        if self.hits + self.misses == 0 {
+            return None;
+        }
+
+        Some((self.hits as f64) / (self.hits + self.misses) as f64)
+    }
+}
+
+#[cfg(debug_assertions)]
+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 StatisticsRecorder for DebugStatistics {
+    // TODO figure out appropriate Ordering
+    fn hit(&self) {
+        self.hits.fetch_add(1, Ordering::SeqCst);
+    }
+
+    fn miss(&self) {
+        self.misses.fetch_add(1, Ordering::SeqCst);
+    }
+
+    fn to_statistics(&self) -> Option<Statistics> {
+        let hits = self.hits.load(Ordering::SeqCst);
+        let misses = self.misses.load(Ordering::SeqCst);
+
+        Some(Statistics { hits, misses })
+    }
+}
+
+#[derive(Debug, Default)]
+pub struct ReleaseStatistics;
+
+impl StatisticsRecorder for ReleaseStatistics {
+    #[inline]
+    fn hit(&self) {}
+
+    #[inline]
+    fn miss(&self) {}
+
+    #[inline]
+    fn to_statistics(&self) -> Option<Statistics> {
+        None
+    }
+}

crates/red_knot/src/cancellation.rs

@@ -0,0 +1,42 @@
+use std::sync::atomic::AtomicBool;
+use std::sync::Arc;
+
+#[derive(Debug, Clone, Default)]
+pub struct CancellationTokenSource {
+    signal: Arc<AtomicBool>,
+}
+
+impl CancellationTokenSource {
+    pub fn new() -> Self {
+        Self {
+            signal: Arc::new(AtomicBool::new(false)),
+        }
+    }
+
+    #[tracing::instrument(level = "trace", skip_all)]
+    pub fn cancel(&self) {
+        self.signal.store(true, std::sync::atomic::Ordering::SeqCst);
+    }
+
+    pub fn is_cancelled(&self) -> bool {
+        self.signal.load(std::sync::atomic::Ordering::SeqCst)
+    }
+
+    pub fn token(&self) -> CancellationToken {
+        CancellationToken {
+            signal: self.signal.clone(),
+        }
+    }
+}
+
+#[derive(Clone, Debug)]
+pub struct CancellationToken {
+    signal: Arc<AtomicBool>,
+}
+
+impl CancellationToken {
+    /// Returns `true` if cancellation has been requested.
+    pub fn is_cancelled(&self) -> bool {
+        self.signal.load(std::sync::atomic::Ordering::SeqCst)
+    }
+}

crates/red_knot/src/db.rs

@@ -0,0 +1,296 @@
+use std::path::Path;
+use std::sync::Arc;
+
+pub use jars::{HasJar, HasJars};
+pub use query::{QueryError, QueryResult};
+pub use runtime::DbRuntime;
+pub use storage::JarsStorage;
+
+use crate::files::FileId;
+use crate::lint::{Diagnostics, LintSemanticStorage, 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};
+
+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.
+    fn snapshot(&self) -> Snapshot<Self>;
+}
+
+/// 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
+    }
+}
+
+// Red knot specific databases code.
+
+pub trait SourceDb: Database {
+    // 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) -> QueryResult<Source>;
+
+    fn parse(&self, file_id: FileId) -> QueryResult<Parsed>;
+}
+
+pub trait SemanticDb: SourceDb {
+    // queries
+    fn resolve_module(&self, name: ModuleName) -> QueryResult<Option<Module>>;
+
+    fn file_to_module(&self, file_id: FileId) -> QueryResult<Option<Module>>;
+
+    fn path_to_module(&self, path: &Path) -> QueryResult<Option<Module>>;
+
+    fn symbol_table(&self, file_id: FileId) -> QueryResult<Arc<SymbolTable>>;
+
+    fn infer_symbol_type(&self, file_id: FileId, symbol_id: SymbolId) -> QueryResult<Type>;
+
+    // mutations
+
+    fn add_module(&mut self, path: &Path) -> Option<(Module, Vec<Arc<ModuleData>>)>;
+
+    fn set_module_search_paths(&mut self, paths: Vec<ModuleSearchPath>);
+}
+
+pub trait LintDb: SemanticDb {
+    fn lint_syntax(&self, file_id: FileId) -> QueryResult<Diagnostics>;
+
+    fn lint_semantic(&self, file_id: FileId) -> QueryResult<Diagnostics>;
+}
+
+pub trait Db: LintDb {}
+
+#[derive(Debug, Default)]
+pub struct SourceJar {
+    pub sources: SourceStorage,
+    pub parsed: ParsedStorage,
+}
+
+#[derive(Debug, Default)]
+pub struct SemanticJar {
+    pub module_resolver: ModuleResolver,
+    pub symbol_tables: SymbolTablesStorage,
+    pub type_store: TypeStore,
+}
+
+#[derive(Debug, Default)]
+pub struct LintJar {
+    pub lint_syntax: LintSyntaxStorage,
+    pub lint_semantic: LintSemanticStorage,
+}
+
+#[cfg(test)]
+pub(crate) mod tests {
+    use std::path::Path;
+    use std::sync::Arc;
+
+    use crate::db::{
+        Database, DbRuntime, HasJar, HasJars, JarsStorage, LintDb, LintJar, QueryResult, SourceDb,
+        SourceJar,
+    };
+    use crate::files::{FileId, Files};
+    use crate::lint::{lint_semantic, 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 super::{SemanticDb, SemanticJar};
+
+    // This can be a partial database used in a single crate for testing.
+    // It would hold fewer data than the full database.
+    #[derive(Debug, Default)]
+    pub(crate) struct TestDb {
+        files: Files,
+        jars: JarsStorage<Self>,
+    }
+
+    impl HasJar<SourceJar> for TestDb {
+        fn jar(&self) -> QueryResult<&SourceJar> {
+            Ok(&self.jars()?.0)
+        }
+
+        fn jar_mut(&mut self) -> &mut SourceJar {
+            &mut self.jars_mut().0
+        }
+    }
+
+    impl HasJar<SemanticJar> for TestDb {
+        fn jar(&self) -> QueryResult<&SemanticJar> {
+            Ok(&self.jars()?.1)
+        }
+
+        fn jar_mut(&mut self) -> &mut SemanticJar {
+            &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
+        }
+    }
+
+    impl SourceDb for TestDb {
+        fn file_id(&self, path: &Path) -> FileId {
+            self.files.intern(path)
+        }
+
+        fn file_path(&self, file_id: FileId) -> Arc<Path> {
+            self.files.path(file_id)
+        }
+
+        fn source(&self, file_id: FileId) -> QueryResult<Source> {
+            source_text(self, file_id)
+        }
+
+        fn parse(&self, file_id: FileId) -> QueryResult<Parsed> {
+            parse(self, file_id)
+        }
+    }
+
+    impl SemanticDb for TestDb {
+        fn resolve_module(&self, name: ModuleName) -> QueryResult<Option<Module>> {
+            resolve_module(self, name)
+        }
+
+        fn file_to_module(&self, file_id: FileId) -> QueryResult<Option<Module>> {
+            file_to_module(self, file_id)
+        }
+
+        fn path_to_module(&self, path: &Path) -> QueryResult<Option<Module>> {
+            path_to_module(self, path)
+        }
+
+        fn symbol_table(&self, file_id: FileId) -> QueryResult<Arc<SymbolTable>> {
+            symbol_table(self, file_id)
+        }
+
+        fn infer_symbol_type(&self, file_id: FileId, symbol_id: SymbolId) -> QueryResult<Type> {
+            infer_symbol_type(self, file_id, symbol_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);
+        }
+    }
+
+    impl LintDb for TestDb {
+        fn lint_syntax(&self, file_id: FileId) -> QueryResult<Diagnostics> {
+            lint_syntax(self, file_id)
+        }
+
+        fn lint_semantic(&self, file_id: FileId) -> QueryResult<Diagnostics> {
+            lint_semantic(self, file_id)
+        }
+    }
+
+    impl HasJars for TestDb {
+        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 Database for TestDb {
+        fn runtime(&self) -> &DbRuntime {
+            self.jars.runtime()
+        }
+
+        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 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) -> 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

@@ -0,0 +1,148 @@
+use std::fmt::{Debug, Formatter};
+use std::hash::{Hash, Hasher};
+use std::path::Path;
+use std::sync::Arc;
+
+use hashbrown::hash_map::RawEntryMut;
+use parking_lot::RwLock;
+use rustc_hash::FxHasher;
+
+use ruff_index::{newtype_index, IndexVec};
+
+type Map<K, V> = hashbrown::HashMap<K, V, ()>;
+
+#[newtype_index]
+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)]
+pub struct Files {
+    inner: Arc<RwLock<FilesInner>>,
+}
+
+impl Files {
+    #[tracing::instrument(level = "debug", skip(self))]
+    pub fn intern(&self, path: &Path) -> FileId {
+        self.inner.write().intern(path)
+    }
+
+    pub fn try_get(&self, path: &Path) -> Option<FileId> {
+        self.inner.read().try_get(path)
+    }
+
+    #[tracing::instrument(level = "debug", skip(self))]
+    pub fn path(&self, id: FileId) -> Arc<Path> {
+        self.inner.read().path(id)
+    }
+}
+
+impl Debug for Files {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        let files = self.inner.read();
+        let mut debug = f.debug_map();
+        for item in files.iter() {
+            debug.entry(&item.0, &item.1);
+        }
+
+        debug.finish()
+    }
+}
+
+impl PartialEq for Files {
+    fn eq(&self, other: &Self) -> bool {
+        self.inner.read().eq(&other.inner.read())
+    }
+}
+
+impl Eq for Files {}
+
+#[derive(Default)]
+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.
+    by_id: IndexVec<FileId, Arc<Path>>,
+}
+
+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 entry = self
+            .by_path
+            .raw_entry_mut()
+            .from_hash(hash, |existing_file| &*self.by_id[*existing_file] == path);
+
+        match entry {
+            RawEntryMut::Occupied(entry) => *entry.key(),
+            RawEntryMut::Vacant(entry) => {
+                let id = self.by_id.push(Arc::from(path));
+                entry.insert_with_hasher(hash, id, (), |_| hash);
+                id
+            }
+        }
+    }
+
+    pub(crate) fn try_get(&self, path: &Path) -> Option<FileId> {
+        let mut hasher = FxHasher::default();
+        path.hash(&mut hasher);
+        let hash = hasher.finish();
+
+        Some(
+            *self
+                .by_path
+                .raw_entry()
+                .from_hash(hash, |existing_file| &*self.by_id[*existing_file] == path)?
+                .0,
+        )
+    }
+
+    /// Returns the path for the file with the given id.
+    pub(crate) fn path(&self, id: FileId) -> Arc<Path> {
+        self.by_id[id].clone()
+    }
+
+    pub(crate) fn iter(&self) -> impl Iterator<Item = (FileId, Arc<Path>)> + '_ {
+        self.by_path.keys().map(|id| (*id, self.by_id[*id].clone()))
+    }
+}
+
+impl PartialEq for FilesInner {
+    fn eq(&self, other: &Self) -> bool {
+        self.by_id == other.by_id
+    }
+}
+
+impl Eq for FilesInner {}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::path::PathBuf;
+
+    #[test]
+    fn insert_path_twice_same_id() {
+        let files = Files::default();
+        let path = PathBuf::from("foo/bar");
+        let id1 = files.intern(&path);
+        let id2 = files.intern(&path);
+        assert_eq!(id1, id2);
+    }
+
+    #[test]
+    fn insert_different_paths_different_ids() {
+        let files = Files::default();
+        let path1 = PathBuf::from("foo/bar");
+        let path2 = PathBuf::from("foo/bar/baz");
+        let id1 = files.intern(&path1);
+        let id2 = files.intern(&path2);
+        assert_ne!(id1, id2);
+    }
+}

crates/red_knot/src/format.rs

@@ -0,0 +1,135 @@
+use std::ops::{Deref, DerefMut};
+
+use ruff_formatter::PrintedRange;
+use ruff_python_formatter::{FormatModuleError, PyFormatOptions};
+use ruff_text_size::TextRange;
+
+use crate::cache::KeyValueCache;
+use crate::db::{HasJar, QueryError, SourceDb};
+use crate::files::FileId;
+use crate::lint::Diagnostics;
+use crate::FxDashSet;
+
+pub(crate) trait FormatDb: SourceDb {
+    /// Formats a file and returns its formatted content or an indicator that it is unchanged.
+    fn format_file(&self, file_id: FileId) -> Result<FormattedFile, FormatError>;
+
+    /// Formats a range in a file.
+    fn format_file_range(
+        &self,
+        file_id: FileId,
+        range: TextRange,
+    ) -> Result<PrintedRange, FormatError>;
+
+    fn check_file_formatted(&self, file_id: FileId) -> Result<Diagnostics, FormatError>;
+}
+
+#[tracing::instrument(level = "trace", skip(db))]
+pub(crate) fn format_file<Db>(db: &Db, file_id: FileId) -> Result<FormattedFile, FormatError>
+where
+    Db: FormatDb + HasJar<FormatJar>,
+{
+    let formatted = &db.jar()?.formatted;
+
+    if formatted.contains(&file_id) {
+        return Ok(FormattedFile::Unchanged);
+    }
+
+    let source = db.source(file_id)?;
+
+    // TODO use the `format_module` method here to re-use the AST.
+    let printed =
+        ruff_python_formatter::format_module_source(source.text(), PyFormatOptions::default())?;
+
+    Ok(if printed.as_code() == source.text() {
+        formatted.insert(file_id);
+        FormattedFile::Unchanged
+    } else {
+        FormattedFile::Formatted(printed.into_code())
+    })
+}
+
+#[tracing::instrument(level = "trace", skip(db))]
+pub(crate) fn format_file_range<Db: FormatDb + HasJar<FormatJar>>(
+    db: &Db,
+    file_id: FileId,
+    range: TextRange,
+) -> Result<PrintedRange, FormatError> {
+    let formatted = &db.jar()?.formatted;
+    let source = db.source(file_id)?;
+
+    if formatted.contains(&file_id) {
+        return Ok(PrintedRange::new(source.text()[range].into(), range));
+    }
+
+    // TODO use the `format_module` method here to re-use the AST.
+
+    let result =
+        ruff_python_formatter::format_range(source.text(), range, PyFormatOptions::default())?;
+    Ok(result)
+}
+
+/// Checks if the file is correctly formatted. It creates a diagnostic for formatting issues.
+#[tracing::instrument(level = "trace", skip(db))]
+pub(crate) fn check_formatted<Db>(db: &Db, file_id: FileId) -> Result<Diagnostics, FormatError>
+where
+    Db: FormatDb + HasJar<FormatJar>,
+{
+    Ok(if db.format_file(file_id)?.is_unchanged() {
+        Diagnostics::Empty
+    } else {
+        Diagnostics::from(vec!["File is not formatted".to_string()])
+    })
+}
+
+#[derive(Debug)]
+pub(crate) enum FormatError {
+    Format(FormatModuleError),
+    Query(QueryError),
+}
+
+impl From<FormatModuleError> for FormatError {
+    fn from(value: FormatModuleError) -> Self {
+        Self::Format(value)
+    }
+}
+
+impl From<QueryError> for FormatError {
+    fn from(value: QueryError) -> Self {
+        Self::Query(value)
+    }
+}
+
+#[derive(Clone, Eq, PartialEq, Debug)]
+pub(crate) enum FormattedFile {
+    Formatted(String),
+    Unchanged,
+}
+
+impl FormattedFile {
+    pub(crate) const fn is_unchanged(&self) -> bool {
+        matches!(self, FormattedFile::Unchanged)
+    }
+}
+
+#[derive(Debug, Default)]
+pub struct FormatJar {
+    pub formatted: FxDashSet<FileId>,
+}
+
+#[derive(Default, Debug)]
+pub(crate) struct FormattedStorage(KeyValueCache<FileId, ()>);
+
+impl Deref for FormattedStorage {
+    type Target = KeyValueCache<FileId, ()>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+impl DerefMut for FormattedStorage {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.0
+    }
+}

crates/red_knot/src/hir.rs

@@ -0,0 +1,67 @@
+//! Key observations
+//!
+//! The HIR avoids allocations to large extends by:
+//! * Using an arena per node type
+//! * using ids and id ranges to reference items.
+//!
+//! Using separate arena per node type has the advantage that the IDs are relatively stable, because
+//! they only change when a node of the same kind has been added or removed. (What's unclear is if that matters or if
+//! it still triggers a re-compute because the AST-id in the node has changed).
+//!
+//! The HIR does not store all details. It mainly stores the *public* interface. There's a reference
+//! back to the AST node to get more details.
+//!
+//!
+
+use crate::ast_ids::{HasAstId, TypedAstId};
+use crate::files::FileId;
+use std::fmt::Formatter;
+use std::hash::{Hash, Hasher};
+
+pub struct HirAstId<N: HasAstId> {
+    file_id: FileId,
+    node_id: TypedAstId<N>,
+}
+
+impl<N: HasAstId> Copy for HirAstId<N> {}
+impl<N: HasAstId> Clone for HirAstId<N> {
+    fn clone(&self) -> Self {
+        *self
+    }
+}
+
+impl<N: HasAstId> PartialEq for HirAstId<N> {
+    fn eq(&self, other: &Self) -> bool {
+        self.file_id == other.file_id && self.node_id == other.node_id
+    }
+}
+
+impl<N: HasAstId> Eq for HirAstId<N> {}
+
+impl<N: HasAstId> std::fmt::Debug for HirAstId<N> {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("HirAstId")
+            .field("file_id", &self.file_id)
+            .field("node_id", &self.node_id)
+            .finish()
+    }
+}
+
+impl<N: HasAstId> Hash for HirAstId<N> {
+    fn hash<H: Hasher>(&self, state: &mut H) {
+        self.file_id.hash(state);
+        self.node_id.hash(state);
+    }
+}
+
+impl<N: HasAstId> HirAstId<N> {
+    pub fn upcast<M: HasAstId>(self) -> HirAstId<M>
+    where
+        N: Into<M>,
+    {
+        HirAstId {
+            file_id: self.file_id,
+            node_id: self.node_id.upcast(),
+        }
+    }
+}

crates/red_knot/src/hir/definition.rs

@@ -0,0 +1,556 @@
+use std::ops::{Index, Range};
+
+use ruff_index::{newtype_index, IndexVec};
+use ruff_python_ast::visitor::preorder;
+use ruff_python_ast::visitor::preorder::PreorderVisitor;
+use ruff_python_ast::{
+    Decorator, ExceptHandler, ExceptHandlerExceptHandler, Expr, MatchCase, ModModule, Stmt,
+    StmtAnnAssign, StmtAssign, StmtClassDef, StmtFunctionDef, StmtGlobal, StmtImport,
+    StmtImportFrom, StmtNonlocal, StmtTypeAlias, TypeParam, TypeParamParamSpec, TypeParamTypeVar,
+    TypeParamTypeVarTuple, WithItem,
+};
+
+use crate::ast_ids::{AstIds, HasAstId};
+use crate::files::FileId;
+use crate::hir::HirAstId;
+use crate::Name;
+
+#[newtype_index]
+pub struct FunctionId;
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct Function {
+    ast_id: HirAstId<StmtFunctionDef>,
+    name: Name,
+    parameters: Range<ParameterId>,
+    type_parameters: Range<TypeParameterId>, // TODO: type_parameters, return expression, decorators
+}
+
+#[newtype_index]
+pub struct ParameterId;
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct Parameter {
+    kind: ParameterKind,
+    name: Name,
+    default: Option<()>, // TODO use expression HIR
+    ast_id: HirAstId<ruff_python_ast::Parameter>,
+}
+
+// TODO or should `Parameter` be an enum?
+#[derive(Copy, Clone, Debug, Eq, PartialEq, Hash)]
+pub enum ParameterKind {
+    PositionalOnly,
+    Arguments,
+    Vararg,
+    KeywordOnly,
+    Kwarg,
+}
+
+#[newtype_index]
+pub struct ClassId;
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct Class {
+    name: Name,
+    ast_id: HirAstId<StmtClassDef>,
+    // TODO type parameters, inheritance, decorators, members
+}
+
+#[newtype_index]
+pub struct AssignmentId;
+
+// This can have more than one name...
+// but that means we can't implement `name()` on `ModuleItem`.
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct Assignment {
+    // TODO: Handle multiple names  / targets
+    name: Name,
+    ast_id: HirAstId<StmtAssign>,
+}
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct AnnotatedAssignment {
+    name: Name,
+    ast_id: HirAstId<StmtAnnAssign>,
+}
+
+#[newtype_index]
+pub struct AnnotatedAssignmentId;
+
+#[newtype_index]
+pub struct TypeAliasId;
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct TypeAlias {
+    name: Name,
+    ast_id: HirAstId<StmtTypeAlias>,
+    parameters: Range<TypeParameterId>,
+}
+
+#[newtype_index]
+pub struct TypeParameterId;
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub enum TypeParameter {
+    TypeVar(TypeParameterTypeVar),
+    ParamSpec(TypeParameterParamSpec),
+    TypeVarTuple(TypeParameterTypeVarTuple),
+}
+
+impl TypeParameter {
+    pub fn ast_id(&self) -> HirAstId<TypeParam> {
+        match self {
+            TypeParameter::TypeVar(type_var) => type_var.ast_id.upcast(),
+            TypeParameter::ParamSpec(param_spec) => param_spec.ast_id.upcast(),
+            TypeParameter::TypeVarTuple(type_var_tuple) => type_var_tuple.ast_id.upcast(),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct TypeParameterTypeVar {
+    name: Name,
+    ast_id: HirAstId<TypeParamTypeVar>,
+}
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct TypeParameterParamSpec {
+    name: Name,
+    ast_id: HirAstId<TypeParamParamSpec>,
+}
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct TypeParameterTypeVarTuple {
+    name: Name,
+    ast_id: HirAstId<TypeParamTypeVarTuple>,
+}
+
+#[newtype_index]
+pub struct GlobalId;
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct Global {
+    // TODO track names
+    ast_id: HirAstId<StmtGlobal>,
+}
+
+#[newtype_index]
+pub struct NonLocalId;
+
+#[derive(Debug, Clone, Eq, PartialEq)]
+pub struct NonLocal {
+    // TODO track names
+    ast_id: HirAstId<StmtNonlocal>,
+}
+
+pub enum DefinitionId {
+    Function(FunctionId),
+    Parameter(ParameterId),
+    Class(ClassId),
+    Assignment(AssignmentId),
+    AnnotatedAssignment(AnnotatedAssignmentId),
+    Global(GlobalId),
+    NonLocal(NonLocalId),
+    TypeParameter(TypeParameterId),
+    TypeAlias(TypeAlias),
+}
+
+pub enum DefinitionItem {
+    Function(Function),
+    Parameter(Parameter),
+    Class(Class),
+    Assignment(Assignment),
+    AnnotatedAssignment(AnnotatedAssignment),
+    Global(Global),
+    NonLocal(NonLocal),
+    TypeParameter(TypeParameter),
+    TypeAlias(TypeAlias),
+}
+
+// The closest is rust-analyzers item-tree. It only represents "Items" which make the public interface of a module
+// (it excludes any other statement or expressions). rust-analyzer uses it as the main input to the name resolution
+// algorithm
+// > It is the input to the name resolution algorithm, as well as to the queries defined in `adt.rs`,
+// > `data.rs`, and most things in `attr.rs`.
+//
+// > One important purpose of this layer is to provide an "invalidation barrier" for incremental
+// > computations: when typing inside an item body, the `ItemTree` of the modified file is typically
+// > unaffected, so we don't have to recompute name resolution results or item data (see `data.rs`).
+//
+// I haven't fully figured this out but I think that this composes the "public" interface of a module?
+// But maybe that's too optimistic.
+//
+//
+#[derive(Debug, Clone, Default, Eq, PartialEq)]
+pub struct Definitions {
+    functions: IndexVec<FunctionId, Function>,
+    parameters: IndexVec<ParameterId, Parameter>,
+    classes: IndexVec<ClassId, Class>,
+    assignments: IndexVec<AssignmentId, Assignment>,
+    annotated_assignments: IndexVec<AnnotatedAssignmentId, AnnotatedAssignment>,
+    type_aliases: IndexVec<TypeAliasId, TypeAlias>,
+    type_parameters: IndexVec<TypeParameterId, TypeParameter>,
+    globals: IndexVec<GlobalId, Global>,
+    non_locals: IndexVec<NonLocalId, NonLocal>,
+}
+
+impl Definitions {
+    pub fn from_module(module: &ModModule, ast_ids: &AstIds, file_id: FileId) -> Self {
+        let mut visitor = DefinitionsVisitor {
+            definitions: Definitions::default(),
+            ast_ids,
+            file_id,
+        };
+
+        visitor.visit_body(&module.body);
+
+        visitor.definitions
+    }
+}
+
+impl Index<FunctionId> for Definitions {
+    type Output = Function;
+
+    fn index(&self, index: FunctionId) -> &Self::Output {
+        &self.functions[index]
+    }
+}
+
+impl Index<ParameterId> for Definitions {
+    type Output = Parameter;
+
+    fn index(&self, index: ParameterId) -> &Self::Output {
+        &self.parameters[index]
+    }
+}
+
+impl Index<ClassId> for Definitions {
+    type Output = Class;
+
+    fn index(&self, index: ClassId) -> &Self::Output {
+        &self.classes[index]
+    }
+}
+
+impl Index<AssignmentId> for Definitions {
+    type Output = Assignment;
+
+    fn index(&self, index: AssignmentId) -> &Self::Output {
+        &self.assignments[index]
+    }
+}
+
+impl Index<AnnotatedAssignmentId> for Definitions {
+    type Output = AnnotatedAssignment;
+
+    fn index(&self, index: AnnotatedAssignmentId) -> &Self::Output {
+        &self.annotated_assignments[index]
+    }
+}
+
+impl Index<TypeAliasId> for Definitions {
+    type Output = TypeAlias;
+
+    fn index(&self, index: TypeAliasId) -> &Self::Output {
+        &self.type_aliases[index]
+    }
+}
+
+impl Index<GlobalId> for Definitions {
+    type Output = Global;
+
+    fn index(&self, index: GlobalId) -> &Self::Output {
+        &self.globals[index]
+    }
+}
+
+impl Index<NonLocalId> for Definitions {
+    type Output = NonLocal;
+
+    fn index(&self, index: NonLocalId) -> &Self::Output {
+        &self.non_locals[index]
+    }
+}
+
+impl Index<TypeParameterId> for Definitions {
+    type Output = TypeParameter;
+
+    fn index(&self, index: TypeParameterId) -> &Self::Output {
+        &self.type_parameters[index]
+    }
+}
+
+struct DefinitionsVisitor<'a> {
+    definitions: Definitions,
+    ast_ids: &'a AstIds,
+    file_id: FileId,
+}
+
+impl DefinitionsVisitor<'_> {
+    fn ast_id<N: HasAstId>(&self, node: &N) -> HirAstId<N> {
+        HirAstId {
+            file_id: self.file_id,
+            node_id: self.ast_ids.ast_id(node),
+        }
+    }
+
+    fn lower_function_def(&mut self, function: &StmtFunctionDef) -> FunctionId {
+        let name = Name::new(&function.name);
+
+        let first_type_parameter_id = self.definitions.type_parameters.next_index();
+        let mut last_type_parameter_id = first_type_parameter_id;
+
+        if let Some(type_params) = &function.type_params {
+            for parameter in &type_params.type_params {
+                let id = self.lower_type_parameter(parameter);
+                last_type_parameter_id = id;
+            }
+        }
+
+        let parameters = self.lower_parameters(&function.parameters);
+
+        self.definitions.functions.push(Function {
+            name,
+            ast_id: self.ast_id(function),
+            parameters,
+            type_parameters: first_type_parameter_id..last_type_parameter_id,
+        })
+    }
+
+    fn lower_parameters(&mut self, parameters: &ruff_python_ast::Parameters) -> Range<ParameterId> {
+        let first_parameter_id = self.definitions.parameters.next_index();
+        let mut last_parameter_id = first_parameter_id;
+
+        for parameter in &parameters.posonlyargs {
+            last_parameter_id = self.definitions.parameters.push(Parameter {
+                kind: ParameterKind::PositionalOnly,
+                name: Name::new(&parameter.parameter.name),
+                default: None,
+                ast_id: self.ast_id(&parameter.parameter),
+            });
+        }
+
+        if let Some(vararg) = &parameters.vararg {
+            last_parameter_id = self.definitions.parameters.push(Parameter {
+                kind: ParameterKind::Vararg,
+                name: Name::new(&vararg.name),
+                default: None,
+                ast_id: self.ast_id(vararg),
+            });
+        }
+
+        for parameter in &parameters.kwonlyargs {
+            last_parameter_id = self.definitions.parameters.push(Parameter {
+                kind: ParameterKind::KeywordOnly,
+                name: Name::new(&parameter.parameter.name),
+                default: None,
+                ast_id: self.ast_id(&parameter.parameter),
+            });
+        }
+
+        if let Some(kwarg) = &parameters.kwarg {
+            last_parameter_id = self.definitions.parameters.push(Parameter {
+                kind: ParameterKind::KeywordOnly,
+                name: Name::new(&kwarg.name),
+                default: None,
+                ast_id: self.ast_id(kwarg),
+            });
+        }
+
+        first_parameter_id..last_parameter_id
+    }
+
+    fn lower_class_def(&mut self, class: &StmtClassDef) -> ClassId {
+        let name = Name::new(&class.name);
+
+        self.definitions.classes.push(Class {
+            name,
+            ast_id: self.ast_id(class),
+        })
+    }
+
+    fn lower_assignment(&mut self, assignment: &StmtAssign) {
+        // FIXME handle multiple names
+        if let Some(Expr::Name(name)) = assignment.targets.first() {
+            self.definitions.assignments.push(Assignment {
+                name: Name::new(&name.id),
+                ast_id: self.ast_id(assignment),
+            });
+        }
+    }
+
+    fn lower_annotated_assignment(&mut self, annotated_assignment: &StmtAnnAssign) {
+        if let Expr::Name(name) = &*annotated_assignment.target {
+            self.definitions
+                .annotated_assignments
+                .push(AnnotatedAssignment {
+                    name: Name::new(&name.id),
+                    ast_id: self.ast_id(annotated_assignment),
+                });
+        }
+    }
+
+    fn lower_type_alias(&mut self, type_alias: &StmtTypeAlias) {
+        if let Expr::Name(name) = &*type_alias.name {
+            let name = Name::new(&name.id);
+
+            let lower_parameters_id = self.definitions.type_parameters.next_index();
+            let mut last_parameter_id = lower_parameters_id;
+
+            if let Some(type_params) = &type_alias.type_params {
+                for type_parameter in &type_params.type_params {
+                    let id = self.lower_type_parameter(type_parameter);
+                    last_parameter_id = id;
+                }
+            }
+
+            self.definitions.type_aliases.push(TypeAlias {
+                name,
+                ast_id: self.ast_id(type_alias),
+                parameters: lower_parameters_id..last_parameter_id,
+            });
+        }
+    }
+
+    fn lower_type_parameter(&mut self, type_parameter: &TypeParam) -> TypeParameterId {
+        match type_parameter {
+            TypeParam::TypeVar(type_var) => {
+                self.definitions
+                    .type_parameters
+                    .push(TypeParameter::TypeVar(TypeParameterTypeVar {
+                        name: Name::new(&type_var.name),
+                        ast_id: self.ast_id(type_var),
+                    }))
+            }
+            TypeParam::ParamSpec(param_spec) => {
+                self.definitions
+                    .type_parameters
+                    .push(TypeParameter::ParamSpec(TypeParameterParamSpec {
+                        name: Name::new(&param_spec.name),
+                        ast_id: self.ast_id(param_spec),
+                    }))
+            }
+            TypeParam::TypeVarTuple(type_var_tuple) => {
+                self.definitions
+                    .type_parameters
+                    .push(TypeParameter::TypeVarTuple(TypeParameterTypeVarTuple {
+                        name: Name::new(&type_var_tuple.name),
+                        ast_id: self.ast_id(type_var_tuple),
+                    }))
+            }
+        }
+    }
+
+    fn lower_import(&mut self, _import: &StmtImport) {
+        // TODO
+    }
+
+    fn lower_import_from(&mut self, _import_from: &StmtImportFrom) {
+        // TODO
+    }
+
+    fn lower_global(&mut self, global: &StmtGlobal) -> GlobalId {
+        self.definitions.globals.push(Global {
+            ast_id: self.ast_id(global),
+        })
+    }
+
+    fn lower_non_local(&mut self, non_local: &StmtNonlocal) -> NonLocalId {
+        self.definitions.non_locals.push(NonLocal {
+            ast_id: self.ast_id(non_local),
+        })
+    }
+
+    fn lower_except_handler(&mut self, _except_handler: &ExceptHandlerExceptHandler) {
+        // TODO
+    }
+
+    fn lower_with_item(&mut self, _with_item: &WithItem) {
+        // TODO
+    }
+
+    fn lower_match_case(&mut self, _match_case: &MatchCase) {
+        // TODO
+    }
+}
+
+impl PreorderVisitor<'_> for DefinitionsVisitor<'_> {
+    fn visit_stmt(&mut self, stmt: &Stmt) {
+        match stmt {
+            // Definition statements
+            Stmt::FunctionDef(definition) => {
+                self.lower_function_def(definition);
+                self.visit_body(&definition.body);
+            }
+            Stmt::ClassDef(definition) => {
+                self.lower_class_def(definition);
+                self.visit_body(&definition.body);
+            }
+            Stmt::Assign(assignment) => {
+                self.lower_assignment(assignment);
+            }
+            Stmt::AnnAssign(annotated_assignment) => {
+                self.lower_annotated_assignment(annotated_assignment);
+            }
+            Stmt::TypeAlias(type_alias) => {
+                self.lower_type_alias(type_alias);
+            }
+
+            Stmt::Import(import) => self.lower_import(import),
+            Stmt::ImportFrom(import_from) => self.lower_import_from(import_from),
+            Stmt::Global(global) => {
+                self.lower_global(global);
+            }
+            Stmt::Nonlocal(non_local) => {
+                self.lower_non_local(non_local);
+            }
+
+            // Visit the compound statement bodies because they can contain other definitions.
+            Stmt::For(_)
+            | Stmt::While(_)
+            | Stmt::If(_)
+            | Stmt::With(_)
+            | Stmt::Match(_)
+            | Stmt::Try(_) => {
+                preorder::walk_stmt(self, stmt);
+            }
+
+            // Skip over simple statements because they can't contain any other definitions.
+            Stmt::Return(_)
+            | Stmt::Delete(_)
+            | Stmt::AugAssign(_)
+            | Stmt::Raise(_)
+            | Stmt::Assert(_)
+            | Stmt::Expr(_)
+            | Stmt::Pass(_)
+            | Stmt::Break(_)
+            | Stmt::Continue(_)
+            | Stmt::IpyEscapeCommand(_) => {
+                // No op
+            }
+        }
+    }
+
+    fn visit_expr(&mut self, _: &'_ Expr) {}
+
+    fn visit_decorator(&mut self, _decorator: &'_ Decorator) {}
+
+    fn visit_except_handler(&mut self, except_handler: &'_ ExceptHandler) {
+        match except_handler {
+            ExceptHandler::ExceptHandler(except_handler) => {
+                self.lower_except_handler(except_handler);
+            }
+        }
+    }
+
+    fn visit_with_item(&mut self, with_item: &'_ WithItem) {
+        self.lower_with_item(with_item);
+    }
+
+    fn visit_match_case(&mut self, match_case: &'_ MatchCase) {
+        self.lower_match_case(match_case);
+        self.visit_body(&match_case.body);
+    }
+}

crates/red_knot/src/lib.rs

@@ -0,0 +1,110 @@
+use std::fmt::Formatter;
+use std::hash::BuildHasherDefault;
+use std::ops::Deref;
+use std::path::{Path, PathBuf};
+
+use rustc_hash::{FxHashSet, FxHasher};
+
+use crate::files::FileId;
+
+pub mod ast_ids;
+pub mod cache;
+pub mod cancellation;
+pub mod db;
+pub mod files;
+mod format;
+pub mod hir;
+pub mod lint;
+pub mod module;
+mod parse;
+pub mod program;
+pub mod source;
+mod symbols;
+mod types;
+pub mod watch;
+
+pub(crate) type FxDashMap<K, V> = dashmap::DashMap<K, V, BuildHasherDefault<FxHasher>>;
+#[allow(unused)]
+pub(crate) type FxDashSet<V> = dashmap::DashSet<V, BuildHasherDefault<FxHasher>>;
+pub(crate) type FxIndexSet<V> = indexmap::set::IndexSet<V, BuildHasherDefault<FxHasher>>;
+
+#[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.
+    root: PathBuf,
+    /// The files that are open in the workspace.
+    ///
+    /// * Editor: The files that are actively being edited in the editor (the user has a tab open with the file).
+    /// * CLI: The resolved files passed as arguments to the CLI.
+    open_files: FxHashSet<FileId>,
+}
+
+impl Workspace {
+    pub fn new(root: PathBuf) -> Self {
+        Self {
+            root,
+            open_files: FxHashSet::default(),
+        }
+    }
+
+    pub fn root(&self) -> &Path {
+        self.root.as_path()
+    }
+
+    // TODO having the content in workspace feels wrong.
+    pub fn open_file(&mut self, file_id: FileId) {
+        self.open_files.insert(file_id);
+    }
+
+    pub fn close_file(&mut self, file_id: FileId) {
+        self.open_files.remove(&file_id);
+    }
+
+    // TODO introduce an `OpenFile` type instead of using an anonymous tuple.
+    pub fn open_files(&self) -> impl Iterator<Item = FileId> + '_ {
+        self.open_files.iter().copied()
+    }
+
+    pub fn is_file_open(&self, file_id: FileId) -> bool {
+        self.open_files.contains(&file_id)
+    }
+}
+
+#[derive(Debug, Clone, Eq, PartialEq, Hash)]
+pub struct Name(smol_str::SmolStr);
+
+impl Name {
+    #[inline]
+    pub fn new(name: &str) -> Self {
+        Self(smol_str::SmolStr::new(name))
+    }
+
+    pub fn as_str(&self) -> &str {
+        self.0.as_str()
+    }
+}
+
+impl Deref for Name {
+    type Target = str;
+
+    #[inline]
+    fn deref(&self) -> &Self::Target {
+        self.as_str()
+    }
+}
+
+impl<T> From<T> for Name
+where
+    T: Into<smol_str::SmolStr>,
+{
+    fn from(value: T) -> Self {
+        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

@@ -0,0 +1,264 @@
+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::{ModModule, StringLiteral};
+
+use crate::cache::KeyValueCache;
+use crate::db::{HasJar, LintDb, LintJar, QueryResult, SemanticDb};
+use crate::files::FileId;
+use crate::parse::Parsed;
+use crate::source::Source;
+use crate::symbols::{Definition, SymbolId, SymbolTable};
+use crate::types::Type;
+
+#[tracing::instrument(level = "debug", skip(db))]
+pub(crate) fn lint_syntax<Db>(db: &Db, file_id: FileId) -> QueryResult<Diagnostics>
+where
+    Db: LintDb + HasJar<LintJar>,
+{
+    let storage = &db.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)?;
+        lint_lines(source.text(), &mut diagnostics);
+
+        let parsed = db.parse(*file_id)?;
+
+        if parsed.errors().is_empty() {
+            let ast = parsed.ast();
+
+            let mut visitor = SyntaxLintVisitor {
+                diagnostics,
+                source: source.text(),
+            };
+            visitor.visit_body(&ast.body);
+            diagnostics = visitor.diagnostics;
+        } else {
+            diagnostics.extend(parsed.errors().iter().map(std::string::ToString::to_string));
+        }
+
+        Ok(Diagnostics::from(diagnostics))
+    })
+}
+
+fn lint_lines(source: &str, diagnostics: &mut Vec<String>) {
+    for (line_number, line) in source.lines().enumerate() {
+        if line.len() < 88 {
+            continue;
+        }
+
+        let char_count = line.chars().count();
+        if char_count > 88 {
+            diagnostics.push(format!(
+                "Line {} is too long ({} characters)",
+                line_number + 1,
+                char_count
+            ));
+        }
+    }
+}
+
+#[tracing::instrument(level = "debug", skip(db))]
+pub(crate) fn lint_semantic<Db>(db: &Db, file_id: FileId) -> QueryResult<Diagnostics>
+where
+    Db: LintDb + HasJar<LintJar>,
+{
+    let storage = &db.jar()?.lint_semantic;
+
+    storage.get(&file_id, |file_id| {
+        let source = db.source(*file_id)?;
+        let parsed = db.parse(*file_id)?;
+        let symbols = db.symbol_table(*file_id)?;
+
+        let context = SemanticLintContext {
+            file_id: *file_id,
+            source,
+            parsed,
+            symbols,
+            db,
+            diagnostics: RefCell::new(Vec::new()),
+        };
+
+        lint_unresolved_imports(&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.symbols().all_definitions() {
+        match definition {
+            Definition::Import(import) => {
+                let ty = context.infer_symbol_type(symbol)?;
+
+                if ty.is_unknown() {
+                    context.push_diagnostic(format!("Unresolved module {}", import.module));
+                }
+            }
+            Definition::ImportFrom(import) => {
+                let ty = context.infer_symbol_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(())
+}
+
+pub struct SemanticLintContext<'a> {
+    file_id: FileId,
+    source: Source,
+    parsed: Parsed,
+    symbols: Arc<SymbolTable>,
+    db: &'a dyn SemanticDb,
+    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) -> &ModModule {
+        self.parsed.ast()
+    }
+
+    pub fn symbols(&self) -> &SymbolTable {
+        &self.symbols
+    }
+
+    pub fn infer_symbol_type(&self, symbol_id: SymbolId) -> QueryResult<Type> {
+        self.db.infer_symbol_type(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);
+    }
+}
+
+#[derive(Debug)]
+struct SyntaxLintVisitor<'a> {
+    diagnostics: Vec<String>,
+    source: &'a str,
+}
+
+impl Visitor<'_> for SyntaxLintVisitor<'_> {
+    fn visit_string_literal(&mut self, string_literal: &'_ StringLiteral) {
+        // A very naive implementation of use double quotes
+        let text = &self.source[string_literal.range];
+
+        if text.starts_with('\'') {
+            self.diagnostics
+                .push("Use double quotes for strings".to_string());
+        }
+    }
+}
+
+#[derive(Debug, Clone)]
+pub enum Diagnostics {
+    Empty,
+    List(Arc<Vec<String>>),
+}
+
+impl Diagnostics {
+    pub fn as_slice(&self) -> &[String] {
+        match self {
+            Diagnostics::Empty => &[],
+            Diagnostics::List(list) => list.as_slice(),
+        }
+    }
+}
+
+impl Deref for Diagnostics {
+    type Target = [String];
+    fn deref(&self) -> &Self::Target {
+        self.as_slice()
+    }
+}
+
+impl From<Vec<String>> for Diagnostics {
+    fn from(value: Vec<String>) -> Self {
+        if value.is_empty() {
+            Diagnostics::Empty
+        } else {
+            Diagnostics::List(Arc::new(value))
+        }
+    }
+}
+
+#[derive(Default, Debug)]
+pub struct LintSyntaxStorage(KeyValueCache<FileId, Diagnostics>);
+
+impl Deref for LintSyntaxStorage {
+    type Target = KeyValueCache<FileId, Diagnostics>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+impl DerefMut for LintSyntaxStorage {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &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

@@ -0,0 +1,446 @@
+#![allow(clippy::dbg_macro)]
+
+use std::collections::hash_map::Entry;
+use std::path::Path;
+use std::sync::Mutex;
+
+use crossbeam::channel as crossbeam_channel;
+use rustc_hash::FxHashMap;
+use tracing::subscriber::Interest;
+use tracing::{Level, Metadata};
+use tracing_subscriber::filter::LevelFilter;
+use tracing_subscriber::layer::{Context, Filter, SubscriberExt};
+use tracing_subscriber::{Layer, Registry};
+use tracing_tree::time::Uptime;
+
+use red_knot::db::{
+    Database, HasJar, ParallelDatabase, QueryError, SemanticDb, SourceDb, SourceJar,
+};
+use red_knot::files::FileId;
+use red_knot::module::{ModuleSearchPath, ModuleSearchPathKind};
+use red_knot::program::check::ExecutionMode;
+use red_knot::program::{FileChange, FileChangeKind, Program};
+use red_knot::watch::FileWatcher;
+use red_knot::Workspace;
+
+#[allow(clippy::print_stdout, clippy::unnecessary_wraps, clippy::print_stderr)]
+fn main() -> anyhow::Result<()> {
+    setup_tracing();
+
+    let arguments: Vec<_> = std::env::args().collect();
+
+    if arguments.len() < 2 {
+        eprintln!("Usage: red_knot <path>");
+        return Err(anyhow::anyhow!("Invalid arguments"));
+    }
+
+    let entry_point = Path::new(&arguments[1]);
+
+    if !entry_point.exists() {
+        eprintln!("The entry point does not exist.");
+        return Err(anyhow::anyhow!("Invalid arguments"));
+    }
+
+    if !entry_point.is_file() {
+        eprintln!("The entry point is not a file.");
+        return Err(anyhow::anyhow!("Invalid arguments"));
+    }
+
+    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);
+    program.set_module_search_paths(vec![workspace_search_path]);
+
+    let entry_id = program.file_id(entry_point);
+    program.workspace_mut().open_file(entry_id);
+
+    let (main_loop, main_loop_cancellation_token) = MainLoop::new();
+
+    // Listen to Ctrl+C and abort the watch mode.
+    let main_loop_cancellation_token = Mutex::new(Some(main_loop_cancellation_token));
+    ctrlc::set_handler(move || {
+        let mut lock = main_loop_cancellation_token.lock().unwrap();
+
+        if let Some(token) = lock.take() {
+            token.stop();
+        }
+    })?;
+
+    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(),
+    )?;
+
+    file_watcher.watch_folder(workspace_folder)?;
+
+    main_loop.run(&mut program);
+
+    let source_jar: &SourceJar = program.jar().unwrap();
+
+    dbg!(source_jar.parsed.statistics());
+    dbg!(source_jar.sources.statistics());
+
+    Ok(())
+}
+
+struct MainLoop {
+    orchestrator_sender: crossbeam_channel::Sender<OrchestratorMessage>,
+    main_loop_receiver: crossbeam_channel::Receiver<MainLoopMessage>,
+}
+
+impl MainLoop {
+    fn new() -> (Self, MainLoopCancellationToken) {
+        let (orchestrator_sender, orchestrator_receiver) = crossbeam_channel::bounded(1);
+        let (main_loop_sender, main_loop_receiver) = crossbeam_channel::bounded(1);
+
+        let mut orchestrator = Orchestrator {
+            receiver: orchestrator_receiver,
+            sender: main_loop_sender.clone(),
+            revision: 0,
+        };
+
+        std::thread::spawn(move || {
+            orchestrator.run();
+        });
+
+        (
+            Self {
+                orchestrator_sender,
+                main_loop_receiver,
+            },
+            MainLoopCancellationToken {
+                sender: main_loop_sender,
+            },
+        )
+    }
+
+    fn file_changes_notifier(&self) -> FileChangesNotifier {
+        FileChangesNotifier {
+            sender: self.orchestrator_sender.clone(),
+        }
+    }
+
+    fn run(self, program: &mut Program) {
+        self.orchestrator_sender
+            .send(OrchestratorMessage::Run)
+            .unwrap();
+
+        for message in &self.main_loop_receiver {
+            tracing::trace!("Main Loop: Tick");
+
+            match message {
+                MainLoopMessage::CheckProgram { revision } => {
+                    {
+                        let program = program.snapshot();
+                        let sender = self.orchestrator_sender.clone();
+
+                        // 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) => {}
+                        });
+                    }
+
+                    if !program.is_cancelled() {
+                        let _ = program.format();
+                    }
+                }
+                MainLoopMessage::ApplyChanges(changes) => {
+                    // Automatically cancels any pending queries and waits for them to complete.
+                    program.apply_changes(changes.iter());
+                }
+                MainLoopMessage::CheckCompleted(diagnostics) => {
+                    dbg!(diagnostics);
+                }
+                MainLoopMessage::Exit => {
+                    return;
+                }
+            }
+        }
+    }
+}
+
+impl Drop for MainLoop {
+    fn drop(&mut self) {
+        self.orchestrator_sender
+            .send(OrchestratorMessage::Shutdown)
+            .unwrap();
+    }
+}
+
+#[derive(Debug, Clone)]
+struct FileChangesNotifier {
+    sender: crossbeam_channel::Sender<OrchestratorMessage>,
+}
+
+impl FileChangesNotifier {
+    fn notify(&self, changes: Vec<FileChange>) {
+        self.sender
+            .send(OrchestratorMessage::FileChanges(changes))
+            .unwrap();
+    }
+}
+
+#[derive(Debug)]
+struct MainLoopCancellationToken {
+    sender: crossbeam_channel::Sender<MainLoopMessage>,
+}
+
+impl MainLoopCancellationToken {
+    fn stop(self) {
+        self.sender.send(MainLoopMessage::Exit).unwrap();
+    }
+}
+
+struct Orchestrator {
+    /// 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 {
+    fn run(&mut self) {
+        while let Ok(message) = self.receiver.recv() {
+            match message {
+                OrchestratorMessage::Run => {
+                    self.sender
+                        .send(MainLoopMessage::CheckProgram {
+                            revision: self.revision,
+                        })
+                        .unwrap();
+                }
+
+                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.
+
+                    self.revision += 1;
+                    self.debounce_changes(changes);
+                }
+                OrchestratorMessage::Shutdown => {
+                    return self.shutdown();
+                }
+            }
+        }
+    }
+
+    fn debounce_changes(&self, changes: Vec<FileChange>) {
+        let mut aggregated_changes = AggregatedChanges::default();
+        aggregated_changes.extend(changes);
+
+        loop {
+            // Consume possibly incoming file change messages before running a new analysis, but don't wait for more than 100ms.
+            crossbeam_channel::select! {
+                recv(self.receiver) -> message => {
+                    match message {
+                        Ok(OrchestratorMessage::Shutdown) => {
+                            return self.shutdown();
+                        }
+                        Ok(OrchestratorMessage::FileChanges(file_changes)) => {
+                            aggregated_changes.extend(file_changes);
+                        }
+
+                        Ok(OrchestratorMessage::CheckProgramCompleted { .. })=> {
+                            // disregard any outdated completion message.
+                        }
+                        Ok(OrchestratorMessage::Run) => unreachable!("The orchestrator is already running."),
+
+                        Err(_) => {
+                            // There are no more senders, no point in waiting for more messages
+                            return;
+                        }
+                    }
+                },
+                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(aggregated_changes)).unwrap();
+                    self.sender.send(MainLoopMessage::CheckProgram { revision: self.revision}).unwrap();
+                    return;
+                }
+            }
+        }
+    }
+
+    #[allow(clippy::unused_self)]
+    fn shutdown(&self) {
+        tracing::trace!("Shutting down orchestrator.");
+    }
+}
+
+/// Message sent from the orchestrator to the main loop.
+#[derive(Debug)]
+enum MainLoopMessage {
+    CheckProgram { revision: usize },
+    CheckCompleted(Vec<String>),
+    ApplyChanges(AggregatedChanges),
+    Exit,
+}
+
+#[derive(Debug)]
+enum OrchestratorMessage {
+    Run,
+    Shutdown,
+
+    CheckProgramCompleted {
+        diagnostics: Vec<String>,
+        revision: usize,
+    },
+
+    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))
+    }
+}
+
+fn setup_tracing() {
+    let subscriber = Registry::default().with(
+        tracing_tree::HierarchicalLayer::default()
+            .with_indent_lines(true)
+            .with_indent_amount(2)
+            .with_bracketed_fields(true)
+            .with_thread_ids(true)
+            .with_targets(true)
+            .with_writer(|| Box::new(std::io::stderr()))
+            .with_timer(Uptime::default())
+            .with_filter(LoggingFilter {
+                trace_level: Level::TRACE,
+            }),
+    );
+
+    tracing::subscriber::set_global_default(subscriber).unwrap();
+}
+
+struct LoggingFilter {
+    trace_level: Level,
+}
+
+impl LoggingFilter {
+    fn is_enabled(&self, meta: &Metadata<'_>) -> bool {
+        let filter = if meta.target().starts_with("red_knot") || meta.target().starts_with("ruff") {
+            self.trace_level
+        } else {
+            Level::INFO
+        };
+
+        meta.level() <= &filter
+    }
+}
+
+impl<S> Filter<S> for LoggingFilter {
+    fn enabled(&self, meta: &Metadata<'_>, _cx: &Context<'_, S>) -> bool {
+        self.is_enabled(meta)
+    }
+
+    fn callsite_enabled(&self, meta: &'static Metadata<'static>) -> Interest {
+        if self.is_enabled(meta) {
+            Interest::always()
+        } else {
+            Interest::never()
+        }
+    }
+
+    fn max_level_hint(&self) -> Option<LevelFilter> {
+        Some(LevelFilter::from_level(self.trace_level))
+    }
+}

crates/red_knot/src/parse.rs

@@ -0,0 +1,95 @@
+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 crate::cache::KeyValueCache;
+use crate::db::{HasJar, QueryResult, SourceDb, SourceJar};
+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
+    }
+}
+
+#[tracing::instrument(level = "debug", skip(db))]
+pub(crate) fn parse<Db>(db: &Db, file_id: FileId) -> QueryResult<Parsed>
+where
+    Db: SourceDb + HasJar<SourceJar>,
+{
+    let parsed = db.jar()?;
+
+    parsed.parsed.get(&file_id, |file_id| {
+        let source = db.source(*file_id)?;
+
+        Ok(Parsed::from_text(source.text()))
+    })
+}
+
+#[derive(Debug, Default)]
+pub struct ParsedStorage(KeyValueCache<FileId, Parsed>);
+
+impl Deref for ParsedStorage {
+    type Target = KeyValueCache<FileId, Parsed>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+impl DerefMut for ParsedStorage {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.0
+    }
+}

crates/red_knot/src/program/check.rs

@@ -0,0 +1,425 @@
+use rayon::{current_num_threads, yield_local};
+use rustc_hash::FxHashSet;
+
+use crate::db::{Database, LintDb, QueryError, QueryResult, SemanticDb};
+use crate::files::FileId;
+use crate::format::{FormatDb, FormatError};
+use crate::lint::Diagnostics;
+use crate::program::Program;
+use crate::symbols::Dependency;
+
+impl Program {
+    /// Checks all open files in the workspace and its dependencies.
+    #[tracing::instrument(level = "debug", skip_all)]
+    pub fn check(&self, mode: ExecutionMode) -> QueryResult<Vec<String>> {
+        self.cancelled()?;
+
+        let mut context = CheckContext::new(self);
+
+        match mode {
+            ExecutionMode::SingleThreaded => SingleThreadedExecutor.run(&mut context)?,
+            ExecutionMode::ThreadPool => ThreadPoolExecutor.run(&mut context)?,
+        };
+
+        Ok(context.finish())
+    }
+
+    #[tracing::instrument(level = "debug", skip(self, context))]
+    fn check_file(&self, file: FileId, context: &CheckFileContext) -> QueryResult<Diagnostics> {
+        self.cancelled()?;
+
+        let symbol_table = self.symbol_table(file)?;
+        let dependencies = symbol_table.dependencies();
+
+        if !dependencies.is_empty() {
+            let module = self.file_to_module(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 = 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_dependency(dependency.path(self)?.file());
+                        }
+                    }
+                }
+            }
+        }
+
+        let mut diagnostics = Vec::new();
+
+        if self.workspace().is_file_open(file) {
+            diagnostics.extend_from_slice(&self.lint_syntax(file)?);
+            diagnostics.extend_from_slice(&self.lint_semantic(file)?);
+
+            match self.check_file_formatted(file) {
+                Ok(format_diagnostics) => {
+                    diagnostics.extend_from_slice(&format_diagnostics);
+                }
+                Err(FormatError::Query(err)) => {
+                    return Err(err);
+                }
+                Err(FormatError::Format(error)) => {
+                    diagnostics.push(format!("Error formatting file: {error}"));
+                }
+            }
+        }
+
+        Ok(Diagnostics::from(diagnostics))
+    }
+}
+
+#[derive(Copy, Clone, Debug, PartialEq, Eq)]
+pub enum ExecutionMode {
+    SingleThreaded,
+    ThreadPool,
+}
+
+/// 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<'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
+    }
+}
+
+/// 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
+    T: Fn(FileId),
+{
+    fn schedule(&self, file_id: FileId) {
+        let f = self;
+        f(file_id);
+    }
+}
+
+/// 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, 'scheduler, S> CheckTaskContext<'a, 'scheduler, S>
+where
+    S: ScheduleDependency,
+{
+    fn as_file_context(&self) -> CheckFileContext<'scheduler> {
+        CheckFileContext {
+            dependency_scheduler: self.dependency_scheduler,
+        }
+    }
+}
+
+/// 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,
+}
+
+impl<'a> CheckFileContext<'a> {
+    fn schedule_dependency(&self, file_id: FileId) {
+        self.dependency_scheduler.schedule(file_id);
+    }
+}
+
+#[derive(Debug)]
+enum CheckFileTask {
+    OpenFile(CheckOpenFileTask),
+    Dependency(CheckDependencyTask),
+}
+
+impl CheckFileTask {
+    /// 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,
+        }
+    }
+}
+
+/// Task to check an open file.
+
+#[derive(Debug)]
+struct CheckOpenFileTask {
+    file_id: FileId,
+}
+
+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())
+    }
+}
+
+/// Task to check a dependency file.
+#[derive(Debug)]
+struct CheckDependencyTask {
+    file_id: FileId,
+}
+
+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())
+    }
+}
+
+/// Executor that schedules the checking of individual program files.
+trait CheckExecutor {
+    fn run(self, context: &mut CheckContext) -> QueryResult<()>;
+}
+
+/// 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;
+
+impl CheckExecutor for SingleThreadedExecutor {
+    fn run(self, context: &mut CheckContext) -> QueryResult<()> {
+        let mut queue = context.check_open_files();
+
+        let noop_schedule_dependency = |_| {};
+
+        while let Some(file) = queue.pop() {
+            context.program().cancelled()?;
+
+            let task_context = context.task_context(&noop_schedule_dependency);
+            context.push_diagnostics(&file.run(&task_context)?);
+        }
+
+        Ok(())
+    }
+}
+
+/// 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/format.rs

@@ -0,0 +1,44 @@
+use crate::db::{QueryResult, SourceDb};
+use crate::format::{FormatDb, FormatError, FormattedFile};
+use crate::program::Program;
+
+impl Program {
+    #[tracing::instrument(level = "trace", skip(self))]
+    pub fn format(&mut self) -> QueryResult<()> {
+        // Formats all open files
+
+        // TODO make `Executor` from `check` reusable.
+        for file in self.workspace.open_files() {
+            match self.format_file(file) {
+                Ok(FormattedFile::Formatted(content)) => {
+                    let path = self.file_path(file);
+
+                    // TODO: This is problematic because it immediately re-triggers the file watcher.
+                    //  A possible solution is to track the self "inflicted" changes inside of programs
+                    //  by tracking the file revision right after the write. It could then use the revision
+                    //  to determine which changes are safe to ignore (and in which context).
+                    //  An other alternative is to not write as part of the `format` command and instead
+                    //  return a Vec with the format results and leave the writing to the caller.
+                    //  I think that's undesired because a) we still need a way to tell the formatter
+                    //  that it won't be necessary to format the content again and
+                    //  b) it would reduce concurrency because the writing would need to wait for the file
+                    //  formatting to be complete, unless we use some form of communication channel.
+                    std::fs::write(path, content).expect("Unable to write file");
+                }
+                Ok(FormattedFile::Unchanged) => {
+                    // No op
+                }
+                Err(FormatError::Query(error)) => {
+                    return Err(error);
+                }
+                Err(FormatError::Format(error)) => {
+                    // TODO proper error handling. We should either propagate this error or
+                    //  emit a diagnostic (probably this).
+                    tracing::warn!("Failed to format file: {}", error);
+                }
+            }
+        }
+
+        Ok(())
+    }
+}

crates/red_knot/src/program/mod.rs

@@ -0,0 +1,252 @@
+use ruff_formatter::PrintedRange;
+use ruff_text_size::TextRange;
+use std::path::Path;
+use std::sync::Arc;
+
+use crate::db::{
+    Database, Db, DbRuntime, HasJar, HasJars, JarsStorage, LintDb, LintJar, ParallelDatabase,
+    QueryResult, SemanticDb, SemanticJar, Snapshot, SourceDb, SourceJar,
+};
+use crate::files::{FileId, Files};
+use crate::format::{
+    check_formatted, format_file, format_file_range, FormatDb, FormatError, FormatJar,
+    FormattedFile,
+};
+use crate::lint::{lint_semantic, 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 crate::Workspace;
+
+pub mod check;
+mod format;
+
+#[derive(Debug)]
+pub struct Program {
+    jars: JarsStorage<Program>,
+    files: Files,
+    workspace: Workspace,
+}
+
+impl Program {
+    pub fn new(workspace: Workspace) -> Self {
+        Self {
+            jars: JarsStorage::default(),
+            files: Files::default(),
+            workspace,
+        }
+    }
+
+    pub fn apply_changes<I>(&mut self, changes: I)
+    where
+        I: IntoIterator<Item = FileChange>,
+    {
+        let (source, semantic, lint, format) = self.jars_mut();
+        for change in changes {
+            semantic.module_resolver.remove_module(change.id);
+            semantic.symbol_tables.remove(&change.id);
+            source.sources.remove(&change.id);
+            source.parsed.remove(&change.id);
+            // TODO: remove all dependent modules as well
+            semantic.type_store.remove_module(change.id);
+            lint.lint_syntax.remove(&change.id);
+            lint.lint_semantic.remove(&change.id);
+            format.formatted.remove(&change.id);
+        }
+    }
+
+    pub fn files(&self) -> &Files {
+        &self.files
+    }
+
+    pub fn workspace(&self) -> &Workspace {
+        &self.workspace
+    }
+
+    pub fn workspace_mut(&mut self) -> &mut Workspace {
+        &mut self.workspace
+    }
+}
+
+impl SourceDb for Program {
+    fn file_id(&self, path: &Path) -> FileId {
+        self.files.intern(path)
+    }
+
+    fn file_path(&self, file_id: FileId) -> Arc<Path> {
+        self.files.path(file_id)
+    }
+
+    fn source(&self, file_id: FileId) -> QueryResult<Source> {
+        source_text(self, file_id)
+    }
+
+    fn parse(&self, file_id: FileId) -> QueryResult<Parsed> {
+        parse(self, file_id)
+    }
+}
+
+impl SemanticDb for Program {
+    fn resolve_module(&self, name: ModuleName) -> QueryResult<Option<Module>> {
+        resolve_module(self, name)
+    }
+
+    fn file_to_module(&self, file_id: FileId) -> QueryResult<Option<Module>> {
+        file_to_module(self, file_id)
+    }
+
+    fn path_to_module(&self, path: &Path) -> QueryResult<Option<Module>> {
+        path_to_module(self, path)
+    }
+
+    fn symbol_table(&self, file_id: FileId) -> QueryResult<Arc<SymbolTable>> {
+        symbol_table(self, file_id)
+    }
+
+    fn infer_symbol_type(&self, file_id: FileId, symbol_id: SymbolId) -> QueryResult<Type> {
+        infer_symbol_type(self, file_id, symbol_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);
+    }
+}
+
+impl LintDb for Program {
+    fn lint_syntax(&self, file_id: FileId) -> QueryResult<Diagnostics> {
+        lint_syntax(self, file_id)
+    }
+
+    fn lint_semantic(&self, file_id: FileId) -> QueryResult<Diagnostics> {
+        lint_semantic(self, file_id)
+    }
+}
+
+impl FormatDb for Program {
+    fn format_file(&self, file_id: FileId) -> Result<FormattedFile, FormatError> {
+        format_file(self, file_id)
+    }
+
+    fn format_file_range(
+        &self,
+        file_id: FileId,
+        range: TextRange,
+    ) -> Result<PrintedRange, FormatError> {
+        format_file_range(self, file_id, range)
+    }
+
+    fn check_file_formatted(&self, file_id: FileId) -> Result<Diagnostics, FormatError> {
+        check_formatted(self, file_id)
+    }
+}
+
+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.clone(),
+            workspace: self.workspace.clone(),
+        })
+    }
+}
+
+impl HasJars for Program {
+    type Jars = (SourceJar, SemanticJar, LintJar, FormatJar);
+
+    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) -> QueryResult<&SourceJar> {
+        Ok(&self.jars()?.0)
+    }
+
+    fn jar_mut(&mut self) -> &mut SourceJar {
+        &mut self.jars_mut().0
+    }
+}
+
+impl HasJar<SemanticJar> for Program {
+    fn jar(&self) -> QueryResult<&SemanticJar> {
+        Ok(&self.jars()?.1)
+    }
+
+    fn jar_mut(&mut self) -> &mut SemanticJar {
+        &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
+    }
+}
+
+impl HasJar<FormatJar> for Program {
+    fn jar(&self) -> QueryResult<&FormatJar> {
+        Ok(&self.jars()?.3)
+    }
+
+    fn jar_mut(&mut self) -> &mut FormatJar {
+        &mut self.jars_mut().3
+    }
+}
+
+#[derive(Copy, Clone, Debug)]
+pub 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 {
+        self.id
+    }
+
+    pub fn kind(&self) -> FileChangeKind {
+        self.kind
+    }
+}
+
+#[derive(Copy, Clone, Debug, Eq, PartialEq)]
+pub enum FileChangeKind {
+    Created,
+    Modified,
+    Deleted,
+}

crates/red_knot/src/source.rs

@@ -0,0 +1,96 @@
+use crate::cache::KeyValueCache;
+use crate::db::{HasJar, QueryResult, SourceDb, SourceJar};
+use ruff_notebook::Notebook;
+use ruff_python_ast::PySourceType;
+use std::ops::{Deref, DerefMut};
+use std::sync::Arc;
+
+use crate::files::FileId;
+
+#[tracing::instrument(level = "debug", skip(db))]
+pub(crate) fn source_text<Db>(db: &Db, file_id: FileId) -> QueryResult<Source>
+where
+    Db: SourceDb + HasJar<SourceJar>,
+{
+    let sources = &db.jar()?.sources;
+
+    sources.get(&file_id, |file_id| {
+        let path = db.file_path(*file_id);
+
+        let source_text = std::fs::read_to_string(&path).unwrap_or_else(|err| {
+            tracing::error!("Failed to read file '{path:?}: {err}'. Falling back to empty text");
+            String::new()
+        });
+
+        let python_ty = PySourceType::from(&path);
+
+        let kind = match python_ty {
+            PySourceType::Python => {
+                SourceKind::Python(Arc::from(source_text))
+            }
+            PySourceType::Stub => SourceKind::Stub(Arc::from(source_text)),
+            PySourceType::Ipynb => {
+                let notebook = Notebook::from_source_code(&source_text).unwrap_or_else(|err| {
+                    // TODO should this be changed to never fail?
+                    // or should we instead add a diagnostic somewhere? But what would we return in this case?
+                    tracing::error!(
+                        "Failed to parse notebook '{path:?}: {err}'. Falling back to an empty notebook"
+                    );
+                    Notebook::from_source_code("").unwrap()
+                });
+
+                SourceKind::IpyNotebook(Arc::new(notebook))
+            }
+        };
+
+        Ok(Source { kind })
+    })
+}
+
+#[derive(Debug, Clone, PartialEq)]
+pub enum SourceKind {
+    Python(Arc<str>),
+    Stub(Arc<str>),
+    IpyNotebook(Arc<Notebook>),
+}
+
+#[derive(Debug, Clone, PartialEq)]
+pub struct Source {
+    kind: SourceKind,
+}
+
+impl Source {
+    pub fn python<T: Into<Arc<str>>>(source: T) -> Self {
+        Self {
+            kind: SourceKind::Python(source.into()),
+        }
+    }
+    pub fn kind(&self) -> &SourceKind {
+        &self.kind
+    }
+
+    pub fn text(&self) -> &str {
+        match &self.kind {
+            SourceKind::Python(text) => text,
+            SourceKind::Stub(text) => text,
+            SourceKind::IpyNotebook(notebook) => notebook.source_code(),
+        }
+    }
+}
+
+#[derive(Debug, Default)]
+pub struct SourceStorage(pub(crate) KeyValueCache<FileId, Source>);
+
+impl Deref for SourceStorage {
+    type Target = KeyValueCache<FileId, Source>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+impl DerefMut for SourceStorage {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.0
+    }
+}

crates/red_knot/src/symbols.rs

@@ -0,0 +1,939 @@
+#![allow(dead_code)]
+
+use std::hash::{Hash, Hasher};
+use std::iter::{Copied, DoubleEndedIterator, FusedIterator};
+use std::num::NonZeroU32;
+use std::ops::{Deref, DerefMut};
+use std::sync::Arc;
+
+use bitflags::bitflags;
+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, QueryResult, SemanticDb, SemanticJar};
+use crate::files::FileId;
+use crate::module::ModuleName;
+use crate::Name;
+
+#[allow(unreachable_pub)]
+#[tracing::instrument(level = "debug", skip(db))]
+pub fn symbol_table<Db>(db: &Db, file_id: FileId) -> QueryResult<Arc<SymbolTable>>
+where
+    Db: SemanticDb + HasJar<SemanticJar>,
+{
+    let jar = db.jar()?;
+
+    jar.symbol_tables.get(&file_id, |_| {
+        let parsed = db.parse(file_id)?;
+        Ok(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) enum Kind {
+    FreeVar,
+    CellVar,
+    CellVarAssigned,
+    ExplicitGlobal,
+    ImplicitGlobal,
+}
+
+bitflags! {
+    #[derive(Copy,Clone,Debug)]
+    pub(crate) 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(crate) struct Symbol {
+    name: Name,
+    flags: SymbolFlags,
+    // kind: Kind,
+}
+
+impl Symbol {
+    pub(crate) fn name(&self) -> &str {
+        self.name.as_str()
+    }
+
+    /// Is the symbol used in its containing scope?
+    pub(crate) fn is_used(&self) -> bool {
+        self.flags.contains(SymbolFlags::IS_USED)
+    }
+
+    /// Is the symbol defined in its containing scope?
+    pub(crate) 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
+}
+
+// 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(Clone, 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(Clone, Debug)]
+pub(crate) struct ImportDefinition {
+    pub(crate) module: ModuleName,
+}
+
+#[derive(Clone, Debug)]
+pub(crate) struct ImportFromDefinition {
+    pub(crate) module: Option<ModuleName>,
+    pub(crate) name: Name,
+    pub(crate) level: u32,
+}
+
+impl ImportFromDefinition {
+    pub(crate) fn module(&self) -> Option<&ModuleName> {
+        self.module.as_ref()
+    }
+
+    pub(crate) fn name(&self) -> &Name {
+        &self.name
+    }
+
+    pub(crate) fn level(&self) -> u32 {
+        self.level
+    }
+}
+
+#[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>,
+    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],
+            current_definition: None,
+        };
+        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 definitions(&self, symbol_id: SymbolId) -> &[Definition] {
+        self.defs
+            .get(&symbol_id)
+            .map(std::vec::Vec::as_slice)
+            .unwrap_or_default()
+    }
+
+    pub(crate) 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 add_or_update_symbol(
+        &mut self,
+        scope_id: ScopeId,
+        name: &str,
+        flags: SymbolFlags,
+    ) -> 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) => {
+                if let Some(symbol) = self.symbols_by_id.get_mut(*entry.key()) {
+                    symbol.flags.insert(flags);
+                };
+                *entry.key()
+            }
+            RawEntryMut::Vacant(entry) => {
+                let id = self.symbols_by_id.push(Symbol { name, flags });
+                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>,
+    /// the definition whose target(s) we are currently walking
+    current_definition: Option<Definition>,
+}
+
+impl SymbolTableBuilder {
+    fn add_or_update_symbol(&mut self, identifier: &str, flags: SymbolFlags) -> SymbolId {
+        self.table
+            .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.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_or_update_symbol(name, SymbolFlags::IS_DEFINED);
+            }
+        }
+        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, ctx, .. }) = expr {
+            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::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_or_update_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_or_update_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 = ModuleName::new(&alias.name.id);
+
+                    let def = Definition::Import(ImportDefinition {
+                        module: module.clone(),
+                    });
+                    self.add_or_update_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| 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.table.dependencies.push(dependency);
+            }
+            ast::Stmt::Assign(node) => {
+                debug_assert!(self.current_definition.is_none());
+                self.current_definition =
+                    Some(Definition::Assignment(TypedNodeKey::from_node(node)));
+                ast::visitor::preorder::walk_stmt(self, stmt);
+                self.current_definition = None;
+            }
+            _ => {
+                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::{SymbolFlags, 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
+                    .definitions(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
+                    .definitions(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
+                    .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 = SymbolTable::from_ast(parsed.ast());
+            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 = 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
+                    .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 = 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
+                    .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 = 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
+                    .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 = 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"]);
+            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 insert_same_name_symbol_twice() {
+        let mut table = SymbolTable::new();
+        let root_scope_id = SymbolTable::root_scope_id();
+        let symbol_id_1 = table.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::IS_DEFINED);
+        let symbol_id_2 = table.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::IS_USED);
+        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 table = SymbolTable::new();
+        let root_scope_id = SymbolTable::root_scope_id();
+        let symbol_id_1 = table.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::empty());
+        let symbol_id_2 = table.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 table = SymbolTable::new();
+        let root_scope_id = SymbolTable::root_scope_id();
+        let foo_symbol_top = table.add_or_update_symbol(root_scope_id, "foo", SymbolFlags::empty());
+        let c_scope = table.add_child_scope(root_scope_id, "C", ScopeKind::Class);
+        let foo_symbol_inner = table.add_or_update_symbol(c_scope, "foo", SymbolFlags::empty());
+        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_or_update_symbol(root_scope_id, "foo", SymbolFlags::empty());
+        let symbol = foo_symbol_id.symbol(&table);
+        assert_eq!(symbol.name.as_str(), "foo");
+    }
+}

crates/red_knot/src/types.rs

@@ -0,0 +1,560 @@
+#![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 }
+    }
+
+    pub const fn is_unbound(&self) -> bool {
+        matches!(self, Type::Unbound)
+    }
+
+    pub const fn is_unknown(&self) -> bool {
+        matches!(self, Type::Unknown)
+    }
+}
+
+impl From<FunctionTypeId> for Type {
+    fn from(id: FunctionTypeId) -> Self {
+        Type::Function(id)
+    }
+}
+
+impl From<UnionTypeId> for Type {
+    fn from(id: UnionTypeId) -> Self {
+        Type::Union(id)
+    }
+}
+
+impl From<IntersectionTypeId> for Type {
+    fn from(id: IntersectionTypeId) -> Self {
+        Type::Intersection(id)
+    }
+}
+
+// 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(&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(&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(&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(&self, file_id: FileId, name: &str) -> FunctionTypeId {
+        self.add_or_get_module(file_id).add_function(name)
+    }
+
+    fn add_class(&self, file_id: FileId, name: &str, bases: Vec<Type>) -> ClassTypeId {
+        self.add_or_get_module(file_id).add_class(name, bases)
+    }
+
+    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, bases: Vec<Type>) -> ClassTypeId {
+        let class_id = self.classes.push(ClassType {
+            name: Name::new(name),
+            // TODO: if no bases are given, that should imply [object]
+            bases,
+        });
+        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,
+    bases: Vec<Type>,
+}
+
+impl ClassType {
+    fn name(&self) -> &str {
+        self.name.as_str()
+    }
+
+    fn bases(&self) -> &[Type] {
+        self.bases.as_slice()
+    }
+}
+
+#[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 store = TypeStore::default();
+        let files = Files::default();
+        let file_id = files.intern(Path::new("/foo"));
+        let id = store.add_class(file_id, "C", Vec::new());
+        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 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", Vec::new());
+        let c2 = store.add_class(file_id, "C2", Vec::new());
+        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", Vec::new());
+        let c2 = store.add_class(file_id, "C2", Vec::new());
+        let c3 = store.add_class(file_id, "C3", Vec::new());
+        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

@@ -0,0 +1,217 @@
+#![allow(dead_code)]
+
+use ruff_python_ast::AstNode;
+
+use crate::db::{HasJar, QueryResult, SemanticDb, SemanticJar};
+use crate::module::ModuleName;
+use crate::symbols::{Definition, ImportFromDefinition, SymbolId};
+use crate::types::Type;
+use crate::FileId;
+use ruff_python_ast as ast;
+
+// FIXME: Figure out proper dead-lock free synchronisation now that this takes `&db` instead of `&mut db`.
+#[tracing::instrument(level = "trace", skip(db))]
+pub fn infer_symbol_type<Db>(db: &Db, file_id: FileId, symbol_id: SymbolId) -> QueryResult<Type>
+where
+    Db: SemanticDb + HasJar<SemanticJar>,
+{
+    let symbols = db.symbol_table(file_id)?;
+    let defs = symbols.definitions(symbol_id);
+
+    if let Some(ty) = db
+        .jar()?
+        .type_store
+        .get_cached_symbol_type(file_id, symbol_id)
+    {
+        return Ok(ty);
+    }
+
+    // TODO handle multiple defs, conditional defs...
+    assert_eq!(defs.len(), 1);
+    let type_store = &db.jar()?.type_store;
+
+    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) = 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 mut bases = Vec::with_capacity(node.bases().len());
+
+                for base in node.bases() {
+                    bases.push(infer_expr_type(db, file_id, base)?);
+                }
+
+                let ty = Type::Class(type_store.add_class(file_id, &node.name.id, bases));
+                type_store.cache_node_type(file_id, *node_key.erased(), ty);
+                ty
+            }
+        }
+        Definition::FunctionDef(node_key) => {
+            if let Some(ty) = 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(ast.as_any_node_ref())
+                    .expect("node key should resolve");
+
+                let ty = type_store.add_function(file_id, &node.name.id).into();
+                type_store.cache_node_type(file_id, *node_key.erased(), ty);
+                ty
+            }
+        }
+        Definition::Assignment(node_key) => {
+            let parsed = db.parse(file_id)?;
+            let ast = parsed.ast();
+            let node = node_key.resolve_unwrap(ast.as_any_node_ref());
+            // TODO handle unpacking assignment correctly
+            infer_expr_type(db, file_id, &node.value)?
+        }
+        _ => todo!("other kinds of definitions"),
+    };
+
+    type_store.cache_symbol_type(file_id, symbol_id, ty);
+
+    // TODO record dependencies
+    Ok(ty)
+}
+
+fn infer_expr_type<Db>(db: &Db, file_id: FileId, expr: &ast::Expr) -> QueryResult<Type>
+where
+    Db: SemanticDb + HasJar<SemanticJar>,
+{
+    // TODO cache the resolution of the type on the node
+    let symbols = db.symbol_table(file_id)?;
+    match expr {
+        ast::Expr::Name(name) => {
+            if let Some(symbol_id) = symbols.root_symbol_id_by_name(&name.id) {
+                db.infer_symbol_type(file_id, symbol_id)
+            } else {
+                Ok(Type::Unknown)
+            }
+        }
+        _ => todo!("full expression type resolution"),
+    }
+}
+
+#[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() -> anyhow::Result<()> {
+        let case = create_test()?;
+        let db = &case.db;
+
+        let a_path = case.src.path().join("a.py");
+        let b_path = case.src.path().join("b.py");
+        std::fs::write(a_path, "from b import C as D; E = 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 e_sym = a_syms
+            .root_symbol_id_by_name("E")
+            .expect("E symbol should be found");
+
+        let ty = db.infer_symbol_type(a_file, e_sym)?;
+
+        let jar = HasJar::<SemanticJar>::jar(db)?;
+        assert!(matches!(ty, Type::Class(_)));
+        assert_eq!(format!("{}", ty.display(&jar.type_store)), "Literal[C]");
+
+        Ok(())
+    }
+
+    #[test]
+    fn resolve_base_class_by_name() -> anyhow::Result<()> {
+        let case = create_test()?;
+        let db = &case.db;
+
+        let path = case.src.path().join("mod.py");
+        std::fs::write(path, "class Base: pass\nclass Sub(Base): pass")?;
+        let file = db
+            .resolve_module(ModuleName::new("mod"))?
+            .expect("module should be found")
+            .path(db)?
+            .file();
+        let syms = db.symbol_table(file)?;
+        let sym = syms
+            .root_symbol_id_by_name("Sub")
+            .expect("Sub symbol should be found");
+
+        let ty = db.infer_symbol_type(file, sym)?;
+
+        let Type::Class(class_id) = ty else {
+            panic!("Sub is not a Class")
+        };
+        let jar = HasJar::<SemanticJar>::jar(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(())
+    }
+}

crates/red_knot/src/watch.rs

@@ -0,0 +1,78 @@
+use anyhow::Context;
+use std::path::Path;
+
+use crate::files::Files;
+use crate::program::{FileChange, FileChangeKind};
+use notify::event::{CreateKind, RemoveKind};
+use notify::{recommended_watcher, Event, EventKind, RecommendedWatcher, RecursiveMode, Watcher};
+
+pub struct FileWatcher {
+    watcher: RecommendedWatcher,
+}
+
+pub trait EventHandler: Send + 'static {
+    fn handle(&self, changes: Vec<FileChange>);
+}
+
+impl<F> EventHandler for F
+where
+    F: Fn(Vec<FileChange>) + Send + 'static,
+{
+    fn handle(&self, changes: Vec<FileChange>) {
+        let f = self;
+        f(changes);
+    }
+}
+
+impl FileWatcher {
+    pub fn new<E>(handler: E, files: Files) -> anyhow::Result<Self>
+    where
+        E: EventHandler,
+    {
+        Self::from_handler(Box::new(handler), files)
+    }
+
+    fn from_handler(handler: Box<dyn EventHandler>, files: Files) -> anyhow::Result<Self> {
+        let watcher = recommended_watcher(move |changes: notify::Result<Event>| {
+            match changes {
+                Ok(event) => {
+                    // TODO verify that this handles all events correctly
+                    let change_kind = match event.kind {
+                        EventKind::Create(CreateKind::File) => FileChangeKind::Created,
+                        EventKind::Modify(_) => FileChangeKind::Modified,
+                        EventKind::Remove(RemoveKind::File) => FileChangeKind::Deleted,
+                        _ => {
+                            return;
+                        }
+                    };
+
+                    let mut changes = Vec::new();
+
+                    for path in event.paths {
+                        if path.is_file() {
+                            let id = files.intern(&path);
+                            changes.push(FileChange::new(id, change_kind));
+                        }
+                    }
+
+                    if !changes.is_empty() {
+                        handler.handle(changes);
+                    }
+                }
+                // TODO proper error handling
+                Err(err) => {
+                    panic!("Error: {err}");
+                }
+            }
+        })
+        .context("Failed to create file watcher.")?;
+
+        Ok(Self { watcher })
+    }
+
+    pub fn watch_folder(&mut self, path: &Path) -> anyhow::Result<()> {
+        self.watcher.watch(path, RecursiveMode::Recursive)?;
+
+        Ok(())
+    }
+}

crates/ruff/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ruff"
-version = "0.2.1"
+version = "0.4.2"
 publish = false
 authors = { workspace = true }
 edition = { workspace = true }
@@ -20,6 +20,7 @@ ruff_macros = { path = "../ruff_macros" }
 ruff_notebook = { path = "../ruff_notebook" }
 ruff_python_ast = { path = "../ruff_python_ast" }
 ruff_python_formatter = { path = "../ruff_python_formatter" }
+ruff_server = { path = "../ruff_server" }
 ruff_source_file = { path = "../ruff_source_file" }
 ruff_text_size = { path = "../ruff_text_size" }
 ruff_workspace = { path = "../ruff_workspace" }
@@ -40,6 +41,7 @@ is-macro = { workspace = true }
 itertools = { workspace = true }
 log = { workspace = true }
 notify = { workspace = true }
+num_cpus = { workspace = true }
 path-absolutize = { workspace = true, features = ["once_cell_cache"] }
 rayon = { workspace = true }
 regex = { workspace = true }
@@ -48,17 +50,20 @@ serde = { workspace = true }
 serde_json = { workspace = true }
 shellexpand = { workspace = true }
 strum = { workspace = true, features = [] }
+tempfile = { workspace = true }
 thiserror = { workspace = true }
+toml = { workspace = true }
 tracing = { workspace = true, features = ["log"] }
+tracing-subscriber = { workspace = true, features = ["registry"] }
+tracing-tree = { workspace = true }
 walkdir = { workspace = true }
 wild = { workspace = true }
 
 [dev-dependencies]
 # Enable test rules during development
 ruff_linter = { path = "../ruff_linter", features = ["clap", "test-rules"] }
-assert_cmd = { workspace = true }
 # Avoid writing colored snapshots when running tests from the terminal
-colored = { workspace = true, features = ["no-color"]}
+colored = { workspace = true, features = ["no-color"] }
 insta = { workspace = true, features = ["filters", "json"] }
 insta-cmd = { workspace = true }
 tempfile = { workspace = true }

crates/ruff/src/args.rs

@@ -1,12 +1,18 @@
 use std::cmp::Ordering;
 use std::fmt::Formatter;
-use std::path::PathBuf;
+use std::ops::Deref;
+use std::path::{Path, PathBuf};
 use std::str::FromStr;
+use std::sync::Arc;
 
+use anyhow::bail;
+use clap::builder::{TypedValueParser, ValueParserFactory};
 use clap::{command, Parser};
 use colored::Colorize;
+use path_absolutize::path_dedot;
 use regex::Regex;
 use rustc_hash::FxHashMap;
+use toml;
 
 use ruff_linter::line_width::LineLength;
 use ruff_linter::logging::LogLevel;
@@ -19,9 +25,55 @@ use ruff_linter::{warn_user, RuleParser, RuleSelector, RuleSelectorParser};
 use ruff_source_file::{LineIndex, OneIndexed};
 use ruff_text_size::TextRange;
 use ruff_workspace::configuration::{Configuration, RuleSelection};
-use ruff_workspace::options::PycodestyleOptions;
+use ruff_workspace::options::{Options, PycodestyleOptions};
 use ruff_workspace::resolver::ConfigurationTransformer;
 
+/// All configuration options that can be passed "globally",
+/// i.e., can be passed to all subcommands
+#[derive(Debug, Default, Clone, clap::Args)]
+pub struct GlobalConfigArgs {
+    #[clap(flatten)]
+    log_level_args: LogLevelArgs,
+    /// Either a path to a TOML configuration file (`pyproject.toml` or `ruff.toml`),
+    /// or a TOML `<KEY> = <VALUE>` pair
+    /// (such as you might find in a `ruff.toml` configuration file)
+    /// overriding a specific configuration option.
+    /// Overrides of individual settings using this option always take precedence
+    /// over all configuration files, including configuration files that were also
+    /// specified using `--config`.
+    #[arg(
+        long,
+        action = clap::ArgAction::Append,
+        value_name = "CONFIG_OPTION",
+        value_parser = ConfigArgumentParser,
+        global = true,
+        help_heading = "Global options",
+    )]
+    pub config: Vec<SingleConfigArgument>,
+    /// Ignore all configuration files.
+    //
+    // Note: We can't mark this as conflicting with `--config` here
+    // as `--config` can be used for specifying configuration overrides
+    // as well as configuration files.
+    // Specifying a configuration file conflicts with `--isolated`;
+    // specifying a configuration override does not.
+    // If a user specifies `ruff check --isolated --config=ruff.toml`,
+    // we emit an error later on, after the initial parsing by clap.
+    #[arg(long, help_heading = "Global options", global = true)]
+    pub isolated: bool,
+}
+
+impl GlobalConfigArgs {
+    pub fn log_level(&self) -> LogLevel {
+        LogLevel::from(&self.log_level_args)
+    }
+
+    #[must_use]
+    fn partition(self) -> (LogLevel, Vec<SingleConfigArgument>, bool) {
+        (self.log_level(), self.config, self.isolated)
+    }
+}
+
 #[derive(Debug, Parser)]
 #[command(
     author,
@@ -32,9 +84,9 @@ use ruff_workspace::resolver::ConfigurationTransformer;
 #[command(version)]
 pub struct Args {
     #[command(subcommand)]
-    pub command: Command,
+    pub(crate) command: Command,
     #[clap(flatten)]
-    pub log_level_args: LogLevelArgs,
+    pub(crate) global_options: GlobalConfigArgs,
 }
 
 #[allow(clippy::large_enum_variant)]
@@ -57,10 +109,6 @@ pub enum Command {
         /// Output format
         #[arg(long, value_enum, default_value = "text")]
         output_format: HelpFormat,
-
-        /// Output format (Deprecated: Use `--output-format` instead).
-        #[arg(long, value_enum, conflicts_with = "output_format", hide = true)]
-        format: Option<HelpFormat>,
     },
     /// List or describe the available configuration options.
     Config { option: Option<String> },
@@ -69,10 +117,6 @@ pub enum Command {
         /// Output format
         #[arg(long, value_enum, default_value = "text")]
         output_format: HelpFormat,
-
-        /// Output format (Deprecated: Use `--output-format` instead).
-        #[arg(long, value_enum, conflicts_with = "output_format", hide = true)]
-        format: Option<HelpFormat>,
     },
     /// Clear any caches in the current directory and any subdirectories.
     #[clap(alias = "--clean")]
@@ -82,6 +126,8 @@ pub enum Command {
     GenerateShellCompletion { shell: clap_complete_command::Shell },
     /// Run the Ruff formatter on the given files or directories.
     Format(FormatCommand),
+    /// Run the language server.
+    Server(ServerCommand),
     /// Display Ruff's version
     Version {
         #[arg(long, value_enum, default_value = "text")]
@@ -144,7 +190,7 @@ pub struct CheckCommand {
     pub output_format: Option<SerializationFormat>,
 
     /// Specify file to write the linter output to (default: stdout).
-    #[arg(short, long)]
+    #[arg(short, long, env = "RUFF_OUTPUT_FILE")]
     pub output_file: Option<PathBuf>,
     /// The minimum Python version that should be supported.
     #[arg(long, value_enum)]
@@ -155,10 +201,6 @@ pub struct CheckCommand {
     preview: bool,
     #[clap(long, overrides_with("preview"), hide = true)]
     no_preview: bool,
-    /// Path to the `pyproject.toml` or `ruff.toml` file to use for
-    /// configuration.
-    #[arg(long, conflicts_with = "isolated")]
-    pub config: Option<PathBuf>,
     /// Comma-separated list of rule codes to enable (or ALL, to enable all rules).
     #[arg(
         long,
@@ -290,9 +332,6 @@ pub struct CheckCommand {
     /// Disable cache reads.
     #[arg(short, long, env = "RUFF_NO_CACHE", help_heading = "Miscellaneous")]
     pub no_cache: bool,
-    /// Ignore all configuration files.
-    #[arg(long, conflicts_with = "config", help_heading = "Miscellaneous")]
-    pub isolated: bool,
     /// Path to the cache directory.
     #[arg(long, env = "RUFF_CACHE_DIR", help_heading = "Miscellaneous")]
     pub cache_dir: Option<PathBuf>,
@@ -384,9 +423,6 @@ pub struct FormatCommand {
     /// difference between the current file and how the formatted file would look like.
     #[arg(long)]
     pub diff: bool,
-    /// Path to the `pyproject.toml` or `ruff.toml` file to use for configuration.
-    #[arg(long, conflicts_with = "isolated")]
-    pub config: Option<PathBuf>,
 
     /// Disable cache reads.
     #[arg(short, long, env = "RUFF_NO_CACHE", help_heading = "Miscellaneous")]
@@ -427,9 +463,6 @@ pub struct FormatCommand {
     /// Set the line-length.
     #[arg(long, help_heading = "Format configuration")]
     pub line_length: Option<LineLength>,
-    /// Ignore all configuration files.
-    #[arg(long, conflicts_with = "config", help_heading = "Miscellaneous")]
-    pub isolated: bool,
     /// The name of the file when passing it through stdin.
     #[arg(long, help_heading = "Miscellaneous")]
     pub stdin_filename: Option<PathBuf>,
@@ -463,6 +496,13 @@ pub struct FormatCommand {
     pub range: Option<FormatRange>,
 }
 
+#[derive(Copy, Clone, Debug, clap::Parser)]
+pub struct ServerCommand {
+    /// Enable preview mode; required for regular operation
+    #[arg(long)]
+    pub(crate) preview: bool,
+}
+
 #[derive(Debug, Clone, Copy, clap::ValueEnum)]
 pub enum HelpFormat {
     Text,
@@ -470,7 +510,7 @@ pub enum HelpFormat {
 }
 
 #[allow(clippy::module_name_repetitions)]
-#[derive(Debug, clap::Args)]
+#[derive(Debug, Default, Clone, clap::Args)]
 pub struct LogLevelArgs {
     /// Enable verbose logging.
     #[arg(
@@ -515,101 +555,191 @@ impl From<&LogLevelArgs> for LogLevel {
     }
 }
 
+/// Configuration-related arguments passed via the CLI.
+#[derive(Default)]
+pub struct ConfigArguments {
+    /// Whether the user specified --isolated on the command line
+    pub(crate) isolated: bool,
+    /// The logging level to be used, derived from command-line arguments passed
+    pub(crate) log_level: LogLevel,
+    /// Path to a pyproject.toml or ruff.toml configuration file (etc.).
+    /// Either 0 or 1 configuration file paths may be provided on the command line.
+    config_file: Option<PathBuf>,
+    /// Overrides provided via the `--config "KEY=VALUE"` option.
+    /// An arbitrary number of these overrides may be provided on the command line.
+    /// These overrides take precedence over all configuration files,
+    /// even configuration files that were also specified using `--config`.
+    overrides: Configuration,
+    /// Overrides provided via dedicated flags such as `--line-length` etc.
+    /// These overrides take precedence over all configuration files,
+    /// and also over all overrides specified using any `--config "KEY=VALUE"` flags.
+    per_flag_overrides: ExplicitConfigOverrides,
+}
+
+impl ConfigArguments {
+    pub fn config_file(&self) -> Option<&Path> {
+        self.config_file.as_deref()
+    }
+
+    fn from_cli_arguments(
+        global_options: GlobalConfigArgs,
+        per_flag_overrides: ExplicitConfigOverrides,
+    ) -> anyhow::Result<Self> {
+        let (log_level, config_options, isolated) = global_options.partition();
+        let mut config_file: Option<PathBuf> = None;
+        let mut overrides = Configuration::default();
+
+        for option in config_options {
+            match option {
+                SingleConfigArgument::SettingsOverride(overridden_option) => {
+                    let overridden_option = Arc::try_unwrap(overridden_option)
+                        .unwrap_or_else(|option| option.deref().clone());
+                    overrides = overrides.combine(Configuration::from_options(
+                        overridden_option,
+                        None,
+                        &path_dedot::CWD,
+                    )?);
+                }
+                SingleConfigArgument::FilePath(path) => {
+                    if isolated {
+                        bail!(
+                            "\
+The argument `--config={}` cannot be used with `--isolated`
+
+  tip: You cannot specify a configuration file and also specify `--isolated`,
+       as `--isolated` causes ruff to ignore all configuration files.
+       For more information, try `--help`.
+",
+                            path.display()
+                        );
+                    }
+                    if let Some(ref config_file) = config_file {
+                        let (first, second) = (config_file.display(), path.display());
+                        bail!(
+                            "\
+You cannot specify more than one configuration file on the command line.
+
+  tip: remove either `--config={first}` or `--config={second}`.
+       For more information, try `--help`.
+"
+                        );
+                    }
+                    config_file = Some(path);
+                }
+            }
+        }
+        Ok(Self {
+            isolated,
+            log_level,
+            config_file,
+            overrides,
+            per_flag_overrides,
+        })
+    }
+}
+
+impl ConfigurationTransformer for ConfigArguments {
+    fn transform(&self, config: Configuration) -> Configuration {
+        let with_config_overrides = self.overrides.clone().combine(config);
+        self.per_flag_overrides.transform(with_config_overrides)
+    }
+}
+
 impl CheckCommand {
     /// Partition the CLI into command-line arguments and configuration
     /// overrides.
-    pub fn partition(self) -> (CheckArguments, CliOverrides) {
-        (
-            CheckArguments {
-                add_noqa: self.add_noqa,
-                config: self.config,
-                diff: self.diff,
-                ecosystem_ci: self.ecosystem_ci,
-                exit_non_zero_on_fix: self.exit_non_zero_on_fix,
-                exit_zero: self.exit_zero,
-                files: self.files,
-                ignore_noqa: self.ignore_noqa,
-                isolated: self.isolated,
-                no_cache: self.no_cache,
-                output_file: self.output_file,
-                show_files: self.show_files,
-                show_settings: self.show_settings,
-                statistics: self.statistics,
-                stdin_filename: self.stdin_filename,
-                watch: self.watch,
-            },
-            CliOverrides {
-                dummy_variable_rgx: self.dummy_variable_rgx,
-                exclude: self.exclude,
-                extend_exclude: self.extend_exclude,
-                extend_fixable: self.extend_fixable,
-                extend_ignore: self.extend_ignore,
-                extend_per_file_ignores: self.extend_per_file_ignores,
-                extend_select: self.extend_select,
-                extend_unfixable: self.extend_unfixable,
-                fixable: self.fixable,
-                ignore: self.ignore,
-                line_length: self.line_length,
-                per_file_ignores: self.per_file_ignores,
-                preview: resolve_bool_arg(self.preview, self.no_preview).map(PreviewMode::from),
-                respect_gitignore: resolve_bool_arg(
-                    self.respect_gitignore,
-                    self.no_respect_gitignore,
-                ),
-                select: self.select,
-                target_version: self.target_version,
-                unfixable: self.unfixable,
-                // TODO(charlie): Included in `pyproject.toml`, but not inherited.
-                cache_dir: self.cache_dir,
-                fix: resolve_bool_arg(self.fix, self.no_fix),
-                fix_only: resolve_bool_arg(self.fix_only, self.no_fix_only),
-                unsafe_fixes: resolve_bool_arg(self.unsafe_fixes, self.no_unsafe_fixes)
-                    .map(UnsafeFixes::from),
-                force_exclude: resolve_bool_arg(self.force_exclude, self.no_force_exclude),
-                output_format: resolve_output_format(
-                    self.output_format,
-                    resolve_bool_arg(self.show_source, self.no_show_source),
-                    resolve_bool_arg(self.preview, self.no_preview).unwrap_or_default(),
-                ),
-                show_fixes: resolve_bool_arg(self.show_fixes, self.no_show_fixes),
-                extension: self.extension,
-            },
-        )
+    pub fn partition(
+        self,
+        global_options: GlobalConfigArgs,
+    ) -> anyhow::Result<(CheckArguments, ConfigArguments)> {
+        let check_arguments = CheckArguments {
+            add_noqa: self.add_noqa,
+            diff: self.diff,
+            ecosystem_ci: self.ecosystem_ci,
+            exit_non_zero_on_fix: self.exit_non_zero_on_fix,
+            exit_zero: self.exit_zero,
+            files: self.files,
+            ignore_noqa: self.ignore_noqa,
+            no_cache: self.no_cache,
+            output_file: self.output_file,
+            show_files: self.show_files,
+            show_settings: self.show_settings,
+            statistics: self.statistics,
+            stdin_filename: self.stdin_filename,
+            watch: self.watch,
+        };
+
+        let cli_overrides = ExplicitConfigOverrides {
+            dummy_variable_rgx: self.dummy_variable_rgx,
+            exclude: self.exclude,
+            extend_exclude: self.extend_exclude,
+            extend_fixable: self.extend_fixable,
+            extend_ignore: self.extend_ignore,
+            extend_per_file_ignores: self.extend_per_file_ignores,
+            extend_select: self.extend_select,
+            extend_unfixable: self.extend_unfixable,
+            fixable: self.fixable,
+            ignore: self.ignore,
+            line_length: self.line_length,
+            per_file_ignores: self.per_file_ignores,
+            preview: resolve_bool_arg(self.preview, self.no_preview).map(PreviewMode::from),
+            respect_gitignore: resolve_bool_arg(self.respect_gitignore, self.no_respect_gitignore),
+            select: self.select,
+            target_version: self.target_version,
+            unfixable: self.unfixable,
+            // TODO(charlie): Included in `pyproject.toml`, but not inherited.
+            cache_dir: self.cache_dir,
+            fix: resolve_bool_arg(self.fix, self.no_fix),
+            fix_only: resolve_bool_arg(self.fix_only, self.no_fix_only),
+            unsafe_fixes: resolve_bool_arg(self.unsafe_fixes, self.no_unsafe_fixes)
+                .map(UnsafeFixes::from),
+            force_exclude: resolve_bool_arg(self.force_exclude, self.no_force_exclude),
+            output_format: resolve_output_format(
+                self.output_format,
+                resolve_bool_arg(self.show_source, self.no_show_source),
+                resolve_bool_arg(self.preview, self.no_preview).unwrap_or_default(),
+            ),
+            show_fixes: resolve_bool_arg(self.show_fixes, self.no_show_fixes),
+            extension: self.extension,
+        };
+
+        let config_args = ConfigArguments::from_cli_arguments(global_options, cli_overrides)?;
+        Ok((check_arguments, config_args))
     }
 }
 
 impl FormatCommand {
     /// Partition the CLI into command-line arguments and configuration
     /// overrides.
-    pub fn partition(self) -> (FormatArguments, CliOverrides) {
-        (
-            FormatArguments {
-                check: self.check,
-                diff: self.diff,
-                config: self.config,
-                files: self.files,
-                isolated: self.isolated,
-                no_cache: self.no_cache,
-                stdin_filename: self.stdin_filename,
-                range: self.range,
-            },
-            CliOverrides {
-                line_length: self.line_length,
-                respect_gitignore: resolve_bool_arg(
-                    self.respect_gitignore,
-                    self.no_respect_gitignore,
-                ),
-                exclude: self.exclude,
-                preview: resolve_bool_arg(self.preview, self.no_preview).map(PreviewMode::from),
-                force_exclude: resolve_bool_arg(self.force_exclude, self.no_force_exclude),
-                target_version: self.target_version,
-                cache_dir: self.cache_dir,
-                extension: self.extension,
+    pub fn partition(
+        self,
+        global_options: GlobalConfigArgs,
+    ) -> anyhow::Result<(FormatArguments, ConfigArguments)> {
+        let format_arguments = FormatArguments {
+            check: self.check,
+            diff: self.diff,
+            files: self.files,
+            no_cache: self.no_cache,
+            stdin_filename: self.stdin_filename,
+            range: self.range,
+        };
 
-                // Unsupported on the formatter CLI, but required on `Overrides`.
-                ..CliOverrides::default()
-            },
-        )
+        let cli_overrides = ExplicitConfigOverrides {
+            line_length: self.line_length,
+            respect_gitignore: resolve_bool_arg(self.respect_gitignore, self.no_respect_gitignore),
+            exclude: self.exclude,
+            preview: resolve_bool_arg(self.preview, self.no_preview).map(PreviewMode::from),
+            force_exclude: resolve_bool_arg(self.force_exclude, self.no_force_exclude),
+            target_version: self.target_version,
+            cache_dir: self.cache_dir,
+            extension: self.extension,
+
+            // Unsupported on the formatter CLI, but required on `Overrides`.
+            ..ExplicitConfigOverrides::default()
+        };
+
+        let config_args = ConfigArguments::from_cli_arguments(global_options, cli_overrides)?;
+        Ok((format_arguments, config_args))
     }
 }
 
@@ -622,6 +752,180 @@ fn resolve_bool_arg(yes: bool, no: bool) -> Option<bool> {
     }
 }
 
+/// Enumeration of various ways in which a --config CLI flag
+/// could be invalid
+#[derive(Debug)]
+enum InvalidConfigFlagReason {
+    InvalidToml(toml::de::Error),
+    /// It was valid TOML, but not a valid ruff config file.
+    /// E.g. the user tried to select a rule that doesn't exist,
+    /// or tried to enable a setting that doesn't exist
+    ValidTomlButInvalidRuffSchema(toml::de::Error),
+    /// It was a valid ruff config file, but the user tried to pass a
+    /// value for `extend` as part of the config override.
+    // `extend` is special, because it affects which config files we look at
+    /// in the first place. We currently only parse --config overrides *after*
+    /// we've combined them with all the arguments from the various config files
+    /// that we found, so trying to override `extend` as part of a --config
+    /// override is forbidden.
+    ExtendPassedViaConfigFlag,
+}
+
+impl InvalidConfigFlagReason {
+    const fn description(&self) -> &'static str {
+        match self {
+            Self::InvalidToml(_) => "The supplied argument is not valid TOML",
+            Self::ValidTomlButInvalidRuffSchema(_) => {
+                "Could not parse the supplied argument as a `ruff.toml` configuration option"
+            }
+            Self::ExtendPassedViaConfigFlag => "Cannot include `extend` in a --config flag value",
+        }
+    }
+}
+
+/// Enumeration to represent a single `--config` argument
+/// passed via the CLI.
+///
+/// Using the `--config` flag, users may pass 0 or 1 paths
+/// to configuration files and an arbitrary number of
+/// "inline TOML" overrides for specific settings.
+///
+/// For example:
+///
+/// ```sh
+/// ruff check --config "path/to/ruff.toml" --config "extend-select=['E501', 'F841']" --config "lint.per-file-ignores = {'some_file.py' = ['F841']}"
+/// ```
+#[derive(Clone, Debug)]
+pub enum SingleConfigArgument {
+    FilePath(PathBuf),
+    SettingsOverride(Arc<Options>),
+}
+
+#[derive(Clone)]
+pub struct ConfigArgumentParser;
+
+impl ValueParserFactory for SingleConfigArgument {
+    type Parser = ConfigArgumentParser;
+
+    fn value_parser() -> Self::Parser {
+        ConfigArgumentParser
+    }
+}
+
+impl TypedValueParser for ConfigArgumentParser {
+    type Value = SingleConfigArgument;
+
+    fn parse_ref(
+        &self,
+        cmd: &clap::Command,
+        arg: Option<&clap::Arg>,
+        value: &std::ffi::OsStr,
+    ) -> Result<Self::Value, clap::Error> {
+        // Convert to UTF-8.
+        let Some(value) = value.to_str() else {
+            // But respect non-UTF-8 paths.
+            let path_to_config_file = PathBuf::from(value);
+            if path_to_config_file.is_file() {
+                return Ok(SingleConfigArgument::FilePath(path_to_config_file));
+            }
+            return Err(clap::Error::new(clap::error::ErrorKind::InvalidUtf8));
+        };
+
+        // Expand environment variables and tildes.
+        if let Ok(path_to_config_file) =
+            shellexpand::full(value).map(|config| PathBuf::from(&*config))
+        {
+            if path_to_config_file.is_file() {
+                return Ok(SingleConfigArgument::FilePath(path_to_config_file));
+            }
+        }
+
+        let config_parse_error = match toml::Table::from_str(value) {
+            Ok(table) => match table.try_into::<Options>() {
+                Ok(option) => {
+                    if option.extend.is_none() {
+                        return Ok(SingleConfigArgument::SettingsOverride(Arc::new(option)));
+                    }
+                    InvalidConfigFlagReason::ExtendPassedViaConfigFlag
+                }
+                Err(underlying_error) => {
+                    InvalidConfigFlagReason::ValidTomlButInvalidRuffSchema(underlying_error)
+                }
+            },
+            Err(underlying_error) => InvalidConfigFlagReason::InvalidToml(underlying_error),
+        };
+
+        let mut new_error = clap::Error::new(clap::error::ErrorKind::ValueValidation).with_cmd(cmd);
+        if let Some(arg) = arg {
+            new_error.insert(
+                clap::error::ContextKind::InvalidArg,
+                clap::error::ContextValue::String(arg.to_string()),
+            );
+        }
+        new_error.insert(
+            clap::error::ContextKind::InvalidValue,
+            clap::error::ContextValue::String(value.to_string()),
+        );
+
+        let underlying_error = match &config_parse_error {
+            InvalidConfigFlagReason::ExtendPassedViaConfigFlag => {
+                let tip = config_parse_error.description().into();
+                new_error.insert(
+                    clap::error::ContextKind::Suggested,
+                    clap::error::ContextValue::StyledStrs(vec![tip]),
+                );
+                return Err(new_error);
+            }
+            InvalidConfigFlagReason::InvalidToml(underlying_error)
+            | InvalidConfigFlagReason::ValidTomlButInvalidRuffSchema(underlying_error) => {
+                underlying_error
+            }
+        };
+
+        // small hack so that multiline tips
+        // have the same indent on the left-hand side:
+        let tip_indent = " ".repeat("  tip: ".len());
+
+        let mut tip = format!(
+            "\
+A `--config` flag must either be a path to a `.toml` configuration file
+{tip_indent}or a TOML `<KEY> = <VALUE>` pair overriding a specific configuration
+{tip_indent}option"
+        );
+
+        // Here we do some heuristics to try to figure out whether
+        // the user was trying to pass in a path to a configuration file
+        // or some inline TOML.
+        // We want to display the most helpful error to the user as possible.
+        if std::path::Path::new(value)
+            .extension()
+            .map_or(false, |ext| ext.eq_ignore_ascii_case("toml"))
+        {
+            if !value.contains('=') {
+                tip.push_str(&format!(
+                    "
+
+It looks like you were trying to pass a path to a configuration file.
+The path `{value}` does not point to a configuration file"
+                ));
+            }
+        } else if value.contains('=') {
+            tip.push_str(&format!(
+                "\n\n{}:\n\n{underlying_error}",
+                config_parse_error.description()
+            ));
+        }
+        let tip = tip.trim_end().to_owned().into();
+
+        new_error.insert(
+            clap::error::ContextKind::Suggested,
+            clap::error::ContextValue::StyledStrs(vec![tip]),
+        );
+
+        Err(new_error)
+    }
+}
+
 fn resolve_output_format(
     output_format: Option<SerializationFormat>,
     show_sources: Option<bool>,
@@ -664,14 +968,12 @@ fn resolve_output_format(
 #[allow(clippy::struct_excessive_bools)]
 pub struct CheckArguments {
     pub add_noqa: bool,
-    pub config: Option<PathBuf>,
     pub diff: bool,
     pub ecosystem_ci: bool,
     pub exit_non_zero_on_fix: bool,
     pub exit_zero: bool,
     pub files: Vec<PathBuf>,
     pub ignore_noqa: bool,
-    pub isolated: bool,
     pub no_cache: bool,
     pub output_file: Option<PathBuf>,
     pub show_files: bool,
@@ -688,9 +990,7 @@ pub struct FormatArguments {
     pub check: bool,
     pub no_cache: bool,
     pub diff: bool,
-    pub config: Option<PathBuf>,
     pub files: Vec<PathBuf>,
-    pub isolated: bool,
     pub stdin_filename: Option<PathBuf>,
     pub range: Option<FormatRange>,
 }
@@ -884,39 +1184,40 @@ impl LineColumnParseError {
     }
 }
 
-/// CLI settings that function as configuration overrides.
+/// Configuration overrides provided via dedicated CLI flags:
+/// `--line-length`, `--respect-gitignore`, etc.
 #[derive(Clone, Default)]
 #[allow(clippy::struct_excessive_bools)]
-pub struct CliOverrides {
-    pub dummy_variable_rgx: Option<Regex>,
-    pub exclude: Option<Vec<FilePattern>>,
-    pub extend_exclude: Option<Vec<FilePattern>>,
-    pub extend_fixable: Option<Vec<RuleSelector>>,
-    pub extend_ignore: Option<Vec<RuleSelector>>,
-    pub extend_select: Option<Vec<RuleSelector>>,
-    pub extend_unfixable: Option<Vec<RuleSelector>>,
-    pub fixable: Option<Vec<RuleSelector>>,
-    pub ignore: Option<Vec<RuleSelector>>,
-    pub line_length: Option<LineLength>,
-    pub per_file_ignores: Option<Vec<PatternPrefixPair>>,
-    pub extend_per_file_ignores: Option<Vec<PatternPrefixPair>>,
-    pub preview: Option<PreviewMode>,
-    pub respect_gitignore: Option<bool>,
-    pub select: Option<Vec<RuleSelector>>,
-    pub target_version: Option<PythonVersion>,
-    pub unfixable: Option<Vec<RuleSelector>>,
+struct ExplicitConfigOverrides {
+    dummy_variable_rgx: Option<Regex>,
+    exclude: Option<Vec<FilePattern>>,
+    extend_exclude: Option<Vec<FilePattern>>,
+    extend_fixable: Option<Vec<RuleSelector>>,
+    extend_ignore: Option<Vec<RuleSelector>>,
+    extend_select: Option<Vec<RuleSelector>>,
+    extend_unfixable: Option<Vec<RuleSelector>>,
+    fixable: Option<Vec<RuleSelector>>,
+    ignore: Option<Vec<RuleSelector>>,
+    line_length: Option<LineLength>,
+    per_file_ignores: Option<Vec<PatternPrefixPair>>,
+    extend_per_file_ignores: Option<Vec<PatternPrefixPair>>,
+    preview: Option<PreviewMode>,
+    respect_gitignore: Option<bool>,
+    select: Option<Vec<RuleSelector>>,
+    target_version: Option<PythonVersion>,
+    unfixable: Option<Vec<RuleSelector>>,
     // TODO(charlie): Captured in pyproject.toml as a default, but not part of `Settings`.
-    pub cache_dir: Option<PathBuf>,
-    pub fix: Option<bool>,
-    pub fix_only: Option<bool>,
-    pub unsafe_fixes: Option<UnsafeFixes>,
-    pub force_exclude: Option<bool>,
-    pub output_format: Option<SerializationFormat>,
-    pub show_fixes: Option<bool>,
-    pub extension: Option<Vec<ExtensionPair>>,
+    cache_dir: Option<PathBuf>,
+    fix: Option<bool>,
+    fix_only: Option<bool>,
+    unsafe_fixes: Option<UnsafeFixes>,
+    force_exclude: Option<bool>,
+    output_format: Option<SerializationFormat>,
+    show_fixes: Option<bool>,
+    extension: Option<Vec<ExtensionPair>>,
 }
 
-impl ConfigurationTransformer for CliOverrides {
+impl ConfigurationTransformer for ExplicitConfigOverrides {
     fn transform(&self, mut config: Configuration) -> Configuration {
         if let Some(cache_dir) = &self.cache_dir {
             config.cache_dir = Some(cache_dir.clone());

crates/ruff/src/cache.rs

@@ -1,7 +1,7 @@
 use std::fmt::Debug;
 use std::fs::{self, File};
 use std::hash::Hasher;
-use std::io::{self, BufReader, BufWriter, Write};
+use std::io::{self, BufReader, Write};
 use std::path::{Path, PathBuf};
 use std::sync::atomic::{AtomicU64, Ordering};
 use std::sync::Mutex;
@@ -15,6 +15,7 @@ use rayon::iter::ParallelIterator;
 use rayon::iter::{IntoParallelIterator, ParallelBridge};
 use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
+use tempfile::NamedTempFile;
 
 use ruff_cache::{CacheKey, CacheKeyHasher};
 use ruff_diagnostics::{DiagnosticKind, Fix};
@@ -165,15 +166,29 @@ impl Cache {
             return Ok(());
         }
 
-        let file = File::create(&self.path)
-            .with_context(|| format!("Failed to create cache file '{}'", self.path.display()))?;
-        let writer = BufWriter::new(file);
-        bincode::serialize_into(writer, &self.package).with_context(|| {
+        // Write the cache to a temporary file first and then rename it for an "atomic" write.
+        // Protects against data loss if the process is killed during the write and races between different ruff
+        // processes, resulting in a corrupted cache file. https://github.com/astral-sh/ruff/issues/8147#issuecomment-1943345964
+        let mut temp_file =
+            NamedTempFile::new_in(self.path.parent().expect("Write path must have a parent"))
+                .context("Failed to create temporary file")?;
+
+        // Serialize to in-memory buffer because hyperfine benchmark showed that it's faster than
+        // using a `BufWriter` and our cache files are small enough that streaming isn't necessary.
+        let serialized =
+            bincode::serialize(&self.package).context("Failed to serialize cache data")?;
+        temp_file
+            .write_all(&serialized)
+            .context("Failed to write serialized cache to temporary file.")?;
+
+        temp_file.persist(&self.path).with_context(|| {
             format!(
-                "Failed to serialise cache to file '{}'",
+                "Failed to rename temporary cache file to {}",
                 self.path.display()
             )
-        })
+        })?;
+
+        Ok(())
     }
 
     /// Applies the pending changes without storing the cache to disk.
@@ -360,15 +375,17 @@ pub(crate) fn init(path: &Path) -> Result<()> {
     fs::create_dir_all(path.join(VERSION))?;
 
     // Add the CACHEDIR.TAG.
-    if !cachedir::is_tagged(path)? {
-        cachedir::add_tag(path)?;
-    }
+    cachedir::ensure_tag(path)?;
 
     // Add the .gitignore.
-    let gitignore_path = path.join(".gitignore");
-    if !gitignore_path.exists() {
-        let mut file = fs::File::create(gitignore_path)?;
-        file.write_all(b"*")?;
+    match fs::OpenOptions::new()
+        .write(true)
+        .create_new(true)
+        .open(path.join(".gitignore"))
+    {
+        Ok(mut file) => file.write_all(b"# Automatically created by ruff.\n*\n")?,
+        Err(err) if err.kind() == io::ErrorKind::AlreadyExists => (),
+        Err(err) => return Err(err.into()),
     }
 
     Ok(())

crates/ruff/src/commands/add_noqa.rs

@@ -12,17 +12,17 @@ use ruff_linter::warn_user_once;
 use ruff_python_ast::{PySourceType, SourceType};
 use ruff_workspace::resolver::{python_files_in_path, PyprojectConfig, ResolvedFile};
 
-use crate::args::CliOverrides;
+use crate::args::ConfigArguments;
 
 /// Add `noqa` directives to a collection of files.
 pub(crate) fn add_noqa(
     files: &[PathBuf],
     pyproject_config: &PyprojectConfig,
-    overrides: &CliOverrides,
+    config_arguments: &ConfigArguments,
 ) -> Result<usize> {
     // Collect all the files to check.
     let start = Instant::now();
-    let (paths, resolver) = python_files_in_path(files, pyproject_config, overrides)?;
+    let (paths, resolver) = python_files_in_path(files, pyproject_config, config_arguments)?;
     let duration = start.elapsed();
     debug!("Identified files to lint in: {:?}", duration);
 

crates/ruff/src/commands/check.rs

@@ -24,7 +24,7 @@ use ruff_workspace::resolver::{
     match_exclusion, python_files_in_path, PyprojectConfig, ResolvedFile,
 };
 
-use crate::args::CliOverrides;
+use crate::args::ConfigArguments;
 use crate::cache::{Cache, PackageCacheMap, PackageCaches};
 use crate::diagnostics::Diagnostics;
 use crate::panic::catch_unwind;
@@ -34,7 +34,7 @@ use crate::panic::catch_unwind;
 pub(crate) fn check(
     files: &[PathBuf],
     pyproject_config: &PyprojectConfig,
-    overrides: &CliOverrides,
+    config_arguments: &ConfigArguments,
     cache: flags::Cache,
     noqa: flags::Noqa,
     fix_mode: flags::FixMode,
@@ -42,7 +42,7 @@ pub(crate) fn check(
 ) -> Result<Diagnostics> {
     // Collect all the Python files to check.
     let start = Instant::now();
-    let (paths, resolver) = python_files_in_path(files, pyproject_config, overrides)?;
+    let (paths, resolver) = python_files_in_path(files, pyproject_config, config_arguments)?;
     debug!("Identified files to lint in: {:?}", start.elapsed());
 
     if paths.is_empty() {
@@ -233,7 +233,7 @@ mod test {
     use ruff_workspace::resolver::{PyprojectConfig, PyprojectDiscoveryStrategy};
     use ruff_workspace::Settings;
 
-    use crate::args::CliOverrides;
+    use crate::args::ConfigArguments;
 
     use super::check;
 
@@ -252,6 +252,7 @@ mod test {
         for file in [&pyproject_toml, &python_file, &notebook] {
             fs::OpenOptions::new()
                 .create(true)
+                .truncate(true)
                 .write(true)
                 .mode(0o000)
                 .open(file)?;
@@ -272,7 +273,7 @@ mod test {
             // Notebooks are not included by default
             &[tempdir.path().to_path_buf(), notebook],
             &pyproject_config,
-            &CliOverrides::default(),
+            &ConfigArguments::default(),
             flags::Cache::Disabled,
             flags::Noqa::Disabled,
             flags::FixMode::Generate,

crates/ruff/src/commands/check_stdin.rs

@@ -6,7 +6,7 @@ use ruff_linter::packaging;
 use ruff_linter::settings::flags;
 use ruff_workspace::resolver::{match_exclusion, python_file_at_path, PyprojectConfig, Resolver};
 
-use crate::args::CliOverrides;
+use crate::args::ConfigArguments;
 use crate::diagnostics::{lint_stdin, Diagnostics};
 use crate::stdin::{parrot_stdin, read_from_stdin};
 
@@ -14,7 +14,7 @@ use crate::stdin::{parrot_stdin, read_from_stdin};
 pub(crate) fn check_stdin(
     filename: Option<&Path>,
     pyproject_config: &PyprojectConfig,
-    overrides: &CliOverrides,
+    overrides: &ConfigArguments,
     noqa: flags::Noqa,
     fix_mode: flags::FixMode,
 ) -> Result<Diagnostics> {

crates/ruff/src/commands/format.rs

@@ -29,7 +29,7 @@ use ruff_text_size::{TextLen, TextRange, TextSize};
 use ruff_workspace::resolver::{match_exclusion, python_files_in_path, ResolvedFile, Resolver};
 use ruff_workspace::FormatterSettings;
 
-use crate::args::{CliOverrides, FormatArguments, FormatRange};
+use crate::args::{ConfigArguments, FormatArguments, FormatRange};
 use crate::cache::{Cache, FileCacheKey, PackageCacheMap, PackageCaches};
 use crate::panic::{catch_unwind, PanicError};
 use crate::resolve::resolve;
@@ -60,18 +60,12 @@ impl FormatMode {
 /// Format a set of files, and return the exit status.
 pub(crate) fn format(
     cli: FormatArguments,
-    overrides: &CliOverrides,
-    log_level: LogLevel,
+    config_arguments: &ConfigArguments,
 ) -> Result<ExitStatus> {
-    let pyproject_config = resolve(
-        cli.isolated,
-        cli.config.as_deref(),
-        overrides,
-        cli.stdin_filename.as_deref(),
-    )?;
+    let pyproject_config = resolve(config_arguments, cli.stdin_filename.as_deref())?;
     let mode = FormatMode::from_cli(&cli);
     let files = resolve_default_files(cli.files, false);
-    let (paths, resolver) = python_files_in_path(&files, &pyproject_config, overrides)?;
+    let (paths, resolver) = python_files_in_path(&files, &pyproject_config, config_arguments)?;
 
     if paths.is_empty() {
         warn_user_once!("No Python files found under the given path(s)");
@@ -203,7 +197,7 @@ pub(crate) fn format(
     }
 
     // Report on the formatting changes.
-    if log_level >= LogLevel::Default {
+    if config_arguments.log_level >= LogLevel::Default {
         if mode.is_diff() {
             // Allow piping the diff to e.g. a file by writing the summary to stderr
             results.write_summary(&mut stderr().lock())?;
@@ -533,7 +527,7 @@ impl<'a> FormatResults<'a> {
             })
             .sorted_unstable_by_key(|(path, _, _)| *path)
         {
-            unformatted.diff(formatted, Some(path), f)?;
+            write!(f, "{}", unformatted.diff(formatted, Some(path)).unwrap())?;
         }
 
         Ok(())

crates/ruff/src/commands/format_stdin.rs

@@ -9,7 +9,7 @@ use ruff_python_ast::{PySourceType, SourceType};
 use ruff_workspace::resolver::{match_exclusion, python_file_at_path, Resolver};
 use ruff_workspace::FormatterSettings;
 
-use crate::args::{CliOverrides, FormatArguments, FormatRange};
+use crate::args::{ConfigArguments, FormatArguments, FormatRange};
 use crate::commands::format::{
     format_source, warn_incompatible_formatter_settings, FormatCommandError, FormatMode,
     FormatResult, FormattedSource,
@@ -19,13 +19,11 @@ use crate::stdin::{parrot_stdin, read_from_stdin};
 use crate::ExitStatus;
 
 /// Run the formatter over a single file, read from `stdin`.
-pub(crate) fn format_stdin(cli: &FormatArguments, overrides: &CliOverrides) -> Result<ExitStatus> {
-    let pyproject_config = resolve(
-        cli.isolated,
-        cli.config.as_deref(),
-        overrides,
-        cli.stdin_filename.as_deref(),
-    )?;
+pub(crate) fn format_stdin(
+    cli: &FormatArguments,
+    config_arguments: &ConfigArguments,
+) -> Result<ExitStatus> {
+    let pyproject_config = resolve(config_arguments, cli.stdin_filename.as_deref())?;
 
     let mut resolver = Resolver::new(&pyproject_config);
     warn_incompatible_formatter_settings(&resolver);
@@ -34,7 +32,7 @@ pub(crate) fn format_stdin(cli: &FormatArguments, overrides: &CliOverrides) -> R
 
     if resolver.force_exclude() {
         if let Some(filename) = cli.stdin_filename.as_deref() {
-            if !python_file_at_path(filename, &mut resolver, overrides)? {
+            if !python_file_at_path(filename, &mut resolver, config_arguments)? {
                 if mode.is_write() {
                     parrot_stdin()?;
                 }
@@ -120,9 +118,13 @@ fn format_source_code(
             }
             FormatMode::Check => {}
             FormatMode::Diff => {
-                source_kind
-                    .diff(formatted, path, &mut stdout().lock())
-                    .map_err(|err| FormatCommandError::Diff(path.map(Path::to_path_buf), err))?;
+                use std::io::Write;
+                write!(
+                    &mut stdout().lock(),
+                    "{}",
+                    source_kind.diff(formatted, path).unwrap()
+                )
+                .map_err(|err| FormatCommandError::Diff(path.map(Path::to_path_buf), err))?;
             }
         },
         FormattedSource::Unchanged => {

crates/ruff/src/commands/mod.rs

@@ -7,6 +7,7 @@ pub(crate) mod format;
 pub(crate) mod format_stdin;
 pub(crate) mod linter;
 pub(crate) mod rule;
+pub(crate) mod server;
 pub(crate) mod show_files;
 pub(crate) mod show_settings;
 pub(crate) mod version;

crates/ruff/src/commands/server.rs

@@ -0,0 +1,79 @@
+use std::num::NonZeroUsize;
+
+use crate::ExitStatus;
+use anyhow::Result;
+use ruff_linter::logging::LogLevel;
+use ruff_server::Server;
+use tracing::{level_filters::LevelFilter, metadata::Level, subscriber::Interest, Metadata};
+use tracing_subscriber::{
+    layer::{Context, Filter, SubscriberExt},
+    Layer, Registry,
+};
+use tracing_tree::time::Uptime;
+
+pub(crate) fn run_server(
+    preview: bool,
+    worker_threads: NonZeroUsize,
+    log_level: LogLevel,
+) -> Result<ExitStatus> {
+    if !preview {
+        tracing::error!("--preview needs to be provided as a command line argument while the server is still unstable.\nFor example: `ruff server --preview`");
+        return Ok(ExitStatus::Error);
+    }
+    let trace_level = if log_level == LogLevel::Verbose {
+        Level::TRACE
+    } else {
+        Level::DEBUG
+    };
+
+    let subscriber = Registry::default().with(
+        tracing_tree::HierarchicalLayer::default()
+            .with_indent_lines(true)
+            .with_indent_amount(2)
+            .with_bracketed_fields(true)
+            .with_targets(true)
+            .with_writer(|| Box::new(std::io::stderr()))
+            .with_timer(Uptime::default())
+            .with_filter(LoggingFilter { trace_level }),
+    );
+
+    tracing::subscriber::set_global_default(subscriber)?;
+
+    let server = Server::new(worker_threads)?;
+
+    server.run().map(|()| ExitStatus::Success)
+}
+
+struct LoggingFilter {
+    trace_level: Level,
+}
+
+impl LoggingFilter {
+    fn is_enabled(&self, meta: &Metadata<'_>) -> bool {
+        let filter = if meta.target().starts_with("ruff") {
+            self.trace_level
+        } else {
+            Level::INFO
+        };
+
+        meta.level() <= &filter
+    }
+}
+
+impl<S> Filter<S> for LoggingFilter {
+    fn enabled(&self, meta: &Metadata<'_>, _cx: &Context<'_, S>) -> bool {
+        self.is_enabled(meta)
+    }
+
+    fn callsite_enabled(&self, meta: &'static Metadata<'static>) -> Interest {
+        if self.is_enabled(meta) {
+            Interest::always()
+        } else {
+            Interest::never()
+        }
+    }
+
+    fn max_level_hint(&self) -> Option<LevelFilter> {
+        Some(LevelFilter::from_level(self.trace_level))
+    }
+}

crates/ruff/src/commands/show_files.rs

@@ -7,17 +7,17 @@ use itertools::Itertools;
 use ruff_linter::warn_user_once;
 use ruff_workspace::resolver::{python_files_in_path, PyprojectConfig, ResolvedFile};
 
-use crate::args::CliOverrides;
+use crate::args::ConfigArguments;
 
 /// Show the list of files to be checked based on current settings.
 pub(crate) fn show_files(
     files: &[PathBuf],
     pyproject_config: &PyprojectConfig,
-    overrides: &CliOverrides,
+    config_arguments: &ConfigArguments,
     writer: &mut impl Write,
 ) -> Result<()> {
     // Collect all files in the hierarchy.
-    let (paths, _resolver) = python_files_in_path(files, pyproject_config, overrides)?;
+    let (paths, _resolver) = python_files_in_path(files, pyproject_config, config_arguments)?;
 
     if paths.is_empty() {
         warn_user_once!("No Python files found under the given path(s)");

crates/ruff/src/commands/show_settings.rs

@@ -6,17 +6,17 @@ use itertools::Itertools;
 
 use ruff_workspace::resolver::{python_files_in_path, PyprojectConfig, ResolvedFile};
 
-use crate::args::CliOverrides;
+use crate::args::ConfigArguments;
 
 /// Print the user-facing configuration settings.
 pub(crate) fn show_settings(
     files: &[PathBuf],
     pyproject_config: &PyprojectConfig,
-    overrides: &CliOverrides,
+    config_arguments: &ConfigArguments,
     writer: &mut impl Write,
 ) -> Result<()> {
     // Collect all files in the hierarchy.
-    let (paths, resolver) = python_files_in_path(files, pyproject_config, overrides)?;
+    let (paths, resolver) = python_files_in_path(files, pyproject_config, config_arguments)?;
 
     // Print the list of files.
     let Some(path) = paths
@@ -31,9 +31,9 @@ pub(crate) fn show_settings(
 
     let settings = resolver.resolve(&path);
 
-    writeln!(writer, "Resolved settings for: {path:?}")?;
+    writeln!(writer, "Resolved settings for: \"{}\"", path.display())?;
     if let Some(settings_path) = pyproject_config.path.as_ref() {
-        writeln!(writer, "Settings path: {settings_path:?}")?;
+        writeln!(writer, "Settings path: \"{}\"", settings_path.display())?;
     }
     write!(writer, "{settings}")?;
 

crates/ruff/src/diagnostics.rs

@@ -3,6 +3,7 @@
 use std::borrow::Cow;
 use std::fs::File;
 use std::io;
+use std::io::Write;
 use std::ops::{Add, AddAssign};
 use std::path::Path;
 
@@ -289,10 +290,10 @@ pub(crate) fn lint_path(
                 match fix_mode {
                     flags::FixMode::Apply => transformed.write(&mut File::create(path)?)?,
                     flags::FixMode::Diff => {
-                        source_kind.diff(
-                            transformed.as_ref(),
-                            Some(path),
+                        write!(
                             &mut io::stdout().lock(),
+                            "{}",
+                            source_kind.diff(&transformed, Some(path)).unwrap()
                         )?;
                     }
                     flags::FixMode::Generate => {}
@@ -442,7 +443,11 @@ pub(crate) fn lint_stdin(
                 flags::FixMode::Diff => {
                     // But only write a diff if it's non-empty.
                     if !fixed.is_empty() {
-                        source_kind.diff(transformed.as_ref(), path, &mut io::stdout().lock())?;
+                        write!(
+                            &mut io::stdout().lock(),
+                            "{}",
+                            source_kind.diff(&transformed, path).unwrap()
+                        )?;
                     }
                 }
                 flags::FixMode::Generate => {}

crates/ruff/src/lib.rs

@@ -2,11 +2,13 @@
 
 use std::fs::File;
 use std::io::{self, stdout, BufWriter, Write};
+use std::num::NonZeroUsize;
 use std::path::{Path, PathBuf};
 use std::process::ExitCode;
 use std::sync::mpsc::channel;
 
 use anyhow::Result;
+use args::{GlobalConfigArgs, ServerCommand};
 use clap::CommandFactory;
 use colored::Colorize;
 use log::warn;
@@ -18,7 +20,7 @@ use ruff_linter::settings::types::SerializationFormat;
 use ruff_linter::{fs, warn_user, warn_user_once};
 use ruff_workspace::Settings;
 
-use crate::args::{Args, CheckCommand, Command, FormatCommand, HelpFormat};
+use crate::args::{Args, CheckCommand, Command, FormatCommand};
 use crate::printer::{Flags as PrinterFlags, Printer};
 
 pub mod args;
@@ -114,20 +116,12 @@ fn resolve_default_files(files: Vec<PathBuf>, is_stdin: bool) -> Vec<PathBuf> {
     }
 }
 
-/// Get the actual value of the `format` desired from either `output_format`
-/// or `format`, and warn the user if they're using the deprecated form.
-fn resolve_help_output_format(output_format: HelpFormat, format: Option<HelpFormat>) -> HelpFormat {
-    if format.is_some() {
-        warn_user!("The `--format` argument is deprecated. Use `--output-format` instead.");
-    }
-    format.unwrap_or(output_format)
-}
-
 pub fn run(
     Args {
         command,
-        log_level_args,
+        global_options,
     }: Args,
+    deprecated_alias_warning: Option<&'static str>,
 ) -> Result<ExitStatus> {
     {
         let default_panic_hook = std::panic::take_hook();
@@ -155,8 +149,18 @@ pub fn run(
     #[cfg(windows)]
     assert!(colored::control::set_virtual_terminal(true).is_ok());
 
-    let log_level = LogLevel::from(&log_level_args);
-    set_up_logging(&log_level)?;
+    // support FORCE_COLOR env var
+    if let Some(force_color) = std::env::var_os("FORCE_COLOR") {
+        if force_color.len() > 0 {
+            colored::control::set_override(true);
+        }
+    }
+
+    set_up_logging(global_options.log_level())?;
+
+    if let Some(deprecated_alias_warning) = deprecated_alias_warning {
+        warn_user!("{}", deprecated_alias_warning);
+    }
 
     match command {
         Command::Version { output_format } => {
@@ -166,10 +170,8 @@ pub fn run(
         Command::Rule {
             rule,
             all,
-            format,
-            mut output_format,
+            output_format,
         } => {
-            output_format = resolve_help_output_format(output_format, format);
             if all {
                 commands::rule::rules(output_format)?;
             }
@@ -182,48 +184,51 @@ pub fn run(
             commands::config::config(option.as_deref())?;
             Ok(ExitStatus::Success)
         }
-        Command::Linter {
-            format,
-            mut output_format,
-        } => {
-            output_format = resolve_help_output_format(output_format, format);
+        Command::Linter { output_format } => {
             commands::linter::linter(output_format)?;
             Ok(ExitStatus::Success)
         }
         Command::Clean => {
-            commands::clean::clean(log_level)?;
+            commands::clean::clean(global_options.log_level())?;
             Ok(ExitStatus::Success)
         }
         Command::GenerateShellCompletion { shell } => {
             shell.generate(&mut Args::command(), &mut stdout());
             Ok(ExitStatus::Success)
         }
-        Command::Check(args) => check(args, log_level),
-        Command::Format(args) => format(args, log_level),
+        Command::Check(args) => check(args, global_options),
+        Command::Format(args) => format(args, global_options),
+        Command::Server(args) => server(args, global_options.log_level()),
     }
 }
 
-fn format(args: FormatCommand, log_level: LogLevel) -> Result<ExitStatus> {
-    let (cli, overrides) = args.partition();
+fn format(args: FormatCommand, global_options: GlobalConfigArgs) -> Result<ExitStatus> {
+    let (cli, config_arguments) = args.partition(global_options)?;
 
     if is_stdin(&cli.files, cli.stdin_filename.as_deref()) {
-        commands::format_stdin::format_stdin(&cli, &overrides)
+        commands::format_stdin::format_stdin(&cli, &config_arguments)
     } else {
-        commands::format::format(cli, &overrides, log_level)
+        commands::format::format(cli, &config_arguments)
     }
 }
 
-pub fn check(args: CheckCommand, log_level: LogLevel) -> Result<ExitStatus> {
-    let (cli, overrides) = args.partition();
+fn server(args: ServerCommand, log_level: LogLevel) -> Result<ExitStatus> {
+    let ServerCommand { preview } = args;
+    // by default, we set the number of worker threads to `num_cpus`, with a maximum of 4.
+    let worker_threads = num_cpus::get().max(4);
+    commands::server::run_server(
+        preview,
+        NonZeroUsize::try_from(worker_threads).expect("a non-zero worker thread count"),
+        log_level,
+    )
+}
+
+pub fn check(args: CheckCommand, global_options: GlobalConfigArgs) -> Result<ExitStatus> {
+    let (cli, config_arguments) = args.partition(global_options)?;
 
     // Construct the "default" settings. These are used when no `pyproject.toml`
     // files are present, or files are injected from outside of the hierarchy.
-    let pyproject_config = resolve::resolve(
-        cli.isolated,
-        cli.config.as_deref(),
-        &overrides,
-        cli.stdin_filename.as_deref(),
-    )?;
+    let pyproject_config = resolve::resolve(&config_arguments, cli.stdin_filename.as_deref())?;
 
     let mut writer: Box<dyn Write> = match cli.output_file {
         Some(path) if !cli.watch => {
@@ -239,11 +244,21 @@ pub fn check(args: CheckCommand, log_level: LogLevel) -> Result<ExitStatus> {
     let files = resolve_default_files(cli.files, is_stdin);
 
     if cli.show_settings {
-        commands::show_settings::show_settings(&files, &pyproject_config, &overrides, &mut writer)?;
+        commands::show_settings::show_settings(
+            &files,
+            &pyproject_config,
+            &config_arguments,
+            &mut writer,
+        )?;
         return Ok(ExitStatus::Success);
     }
     if cli.show_files {
-        commands::show_files::show_files(&files, &pyproject_config, &overrides, &mut writer)?;
+        commands::show_files::show_files(
+            &files,
+            &pyproject_config,
+            &config_arguments,
+            &mut writer,
+        )?;
         return Ok(ExitStatus::Success);
     }
 
@@ -302,8 +317,9 @@ pub fn check(args: CheckCommand, log_level: LogLevel) -> Result<ExitStatus> {
         if !fix_mode.is_generate() {
             warn_user!("--fix is incompatible with --add-noqa.");
         }
-        let modifications = commands::add_noqa::add_noqa(&files, &pyproject_config, &overrides)?;
-        if modifications > 0 && log_level >= LogLevel::Default {
+        let modifications =
+            commands::add_noqa::add_noqa(&files, &pyproject_config, &config_arguments)?;
+        if modifications > 0 && config_arguments.log_level >= LogLevel::Default {
             let s = if modifications == 1 { "" } else { "s" };
             #[allow(clippy::print_stderr)]
             {
@@ -315,7 +331,7 @@ pub fn check(args: CheckCommand, log_level: LogLevel) -> Result<ExitStatus> {
 
     let printer = Printer::new(
         output_format,
-        log_level,
+        config_arguments.log_level,
         fix_mode,
         unsafe_fixes,
         printer_flags,
@@ -352,7 +368,7 @@ pub fn check(args: CheckCommand, log_level: LogLevel) -> Result<ExitStatus> {
         let messages = commands::check::check(
             &files,
             &pyproject_config,
-            &overrides,
+            &config_arguments,
             cache.into(),
             noqa.into(),
             fix_mode,
@@ -372,12 +388,8 @@ pub fn check(args: CheckCommand, log_level: LogLevel) -> Result<ExitStatus> {
                     };
 
                     if matches!(change_kind, ChangeKind::Configuration) {
-                        pyproject_config = resolve::resolve(
-                            cli.isolated,
-                            cli.config.as_deref(),
-                            &overrides,
-                            cli.stdin_filename.as_deref(),
-                        )?;
+                        pyproject_config =
+                            resolve::resolve(&config_arguments, cli.stdin_filename.as_deref())?;
                     }
                     Printer::clear_screen()?;
                     printer.write_to_user("File change detected...\n");
@@ -385,7 +397,7 @@ pub fn check(args: CheckCommand, log_level: LogLevel) -> Result<ExitStatus> {
                     let messages = commands::check::check(
                         &files,
                         &pyproject_config,
-                        &overrides,
+                        &config_arguments,
                         cache.into(),
                         noqa.into(),
                         fix_mode,
@@ -402,7 +414,7 @@ pub fn check(args: CheckCommand, log_level: LogLevel) -> Result<ExitStatus> {
             commands::check_stdin::check_stdin(
                 cli.stdin_filename.map(fs::normalize_path).as_deref(),
                 &pyproject_config,
-                &overrides,
+                &config_arguments,
                 noqa.into(),
                 fix_mode,
             )?
@@ -410,7 +422,7 @@ pub fn check(args: CheckCommand, log_level: LogLevel) -> Result<ExitStatus> {
             commands::check::check(
                 &files,
                 &pyproject_config,
-                &overrides,
+                &config_arguments,
                 cache.into(),
                 noqa.into(),
                 fix_mode,

crates/ruff/src/main.rs

@@ -27,26 +27,42 @@ pub fn main() -> ExitCode {
     let mut args =
         argfile::expand_args_from(args, argfile::parse_fromfile, argfile::PREFIX).unwrap();
 
-    // Clap doesn't support default subcommands but we want to run `check` by
-    // default for convenience and backwards-compatibility, so we just
-    // preprocess the arguments accordingly before passing them to Clap.
-    if let Some(arg) = args.get(1) {
-        if arg
-            .to_str()
-            .is_some_and(|arg| !Command::has_subcommand(rewrite_legacy_subcommand(arg)))
+    // We can't use `warn_user` here because logging isn't set up at this point
+    // and we also don't know if the user runs ruff with quiet.
+    // Keep the message and pass it to `run` that is responsible for emitting the warning.
+    let deprecated_alias_warning = match args.get(1).and_then(|arg| arg.to_str()) {
+        // Deprecated aliases that are handled by clap
+        Some("--explain") => {
+            Some("`ruff --explain <RULE>` is deprecated. Use `ruff rule <RULE>` instead.")
+        }
+        Some("--clean") => {
+            Some("`ruff --clean` is deprecated. Use `ruff clean` instead.")
+        }
+        Some("--generate-shell-completion") => {
+            Some("`ruff --generate-shell-completion <SHELL>` is deprecated. Use `ruff generate-shell-completion <SHELL>` instead.")
+        }
+        // Deprecated `ruff` alias to `ruff check`
+        // Clap doesn't support default subcommands but we want to run `check` by
+        // default for convenience and backwards-compatibility, so we just
+        // preprocess the arguments accordingly before passing them to Clap.
+        Some(arg) if !Command::has_subcommand(arg)
             && arg != "-h"
             && arg != "--help"
             && arg != "-V"
             && arg != "--version"
-            && arg != "help"
-        {
-            args.insert(1, "check".into());
-        }
-    }
+            && arg != "help" => {
+
+            {
+                args.insert(1, "check".into());
+                Some("`ruff <path>` is deprecated. Use `ruff check <path>` instead.")
+            }
+        },
+        _ => None
+    };
 
     let args = Args::parse_from(args);
 
-    match run(args) {
+    match run(args, deprecated_alias_warning) {
         Ok(code) => code.into(),
         Err(err) => {
             #[allow(clippy::print_stderr)]
@@ -65,12 +81,3 @@ pub fn main() -> ExitCode {
         }
     }
 }
-
-fn rewrite_legacy_subcommand(cmd: &str) -> &str {
-    match cmd {
-        "--explain" => "rule",
-        "--clean" => "clean",
-        "--generate-shell-completion" => "generate-shell-completion",
-        cmd => cmd,
-    }
-}

crates/ruff/src/panic.rs

@@ -16,7 +16,7 @@ impl std::fmt::Display for PanicError {
 }
 
 thread_local! {
-    static LAST_PANIC: std::cell::Cell<Option<PanicError>> = std::cell::Cell::new(None);
+    static LAST_PANIC: std::cell::Cell<Option<PanicError>> = const { std::cell::Cell::new(None) };
 }
 
 /// [`catch_unwind`](std::panic::catch_unwind) wrapper that sets a custom [`set_hook`](std::panic::set_hook)

crates/ruff/src/printer.rs

@@ -118,6 +118,8 @@ impl Printer {
                 } else if remaining > 0 {
                     let s = if remaining == 1 { "" } else { "s" };
                     writeln!(writer, "Found {remaining} error{s}.")?;
+                } else if remaining == 0 {
+                    writeln!(writer, "All checks passed!")?;
                 }
 
                 if let Some(fixables) = fixables {

crates/ruff/src/resolve.rs

@@ -1,4 +1,4 @@
-use std::path::{Path, PathBuf};
+use std::path::Path;
 
 use anyhow::Result;
 use log::debug;
@@ -11,19 +11,17 @@ use ruff_workspace::resolver::{
     Relativity,
 };
 
-use crate::args::CliOverrides;
+use crate::args::ConfigArguments;
 
 /// Resolve the relevant settings strategy and defaults for the current
 /// invocation.
 pub fn resolve(
-    isolated: bool,
-    config: Option<&Path>,
-    overrides: &CliOverrides,
+    config_arguments: &ConfigArguments,
     stdin_filename: Option<&Path>,
 ) -> Result<PyprojectConfig> {
     // First priority: if we're running in isolated mode, use the default settings.
-    if isolated {
-        let config = overrides.transform(Configuration::default());
+    if config_arguments.isolated {
+        let config = config_arguments.transform(Configuration::default());
         let settings = config.into_settings(&path_dedot::CWD)?;
         debug!("Isolated mode, not reading any pyproject.toml");
         return Ok(PyprojectConfig::new(
@@ -36,12 +34,8 @@ pub fn resolve(
     // Second priority: the user specified a `pyproject.toml` file. Use that
     // `pyproject.toml` for _all_ configuration, and resolve paths relative to the
     // current working directory. (This matches ESLint's behavior.)
-    if let Some(pyproject) = config
-        .map(|config| config.display().to_string())
-        .map(|config| shellexpand::full(&config).map(|config| PathBuf::from(config.as_ref())))
-        .transpose()?
-    {
-        let settings = resolve_root_settings(&pyproject, Relativity::Cwd, overrides)?;
+    if let Some(pyproject) = config_arguments.config_file() {
+        let settings = resolve_root_settings(pyproject, Relativity::Cwd, config_arguments)?;
         debug!(
             "Using user-specified configuration file at: {}",
             pyproject.display()
@@ -49,7 +43,7 @@ pub fn resolve(
         return Ok(PyprojectConfig::new(
             PyprojectDiscoveryStrategy::Fixed,
             settings,
-            Some(pyproject),
+            Some(pyproject.to_path_buf()),
         ));
     }
 
@@ -67,7 +61,7 @@ pub fn resolve(
             "Using configuration file (via parent) at: {}",
             pyproject.display()
         );
-        let settings = resolve_root_settings(&pyproject, Relativity::Parent, overrides)?;
+        let settings = resolve_root_settings(&pyproject, Relativity::Parent, config_arguments)?;
         return Ok(PyprojectConfig::new(
             PyprojectDiscoveryStrategy::Hierarchical,
             settings,
@@ -84,7 +78,7 @@ pub fn resolve(
             "Using configuration file (via cwd) at: {}",
             pyproject.display()
         );
-        let settings = resolve_root_settings(&pyproject, Relativity::Cwd, overrides)?;
+        let settings = resolve_root_settings(&pyproject, Relativity::Cwd, config_arguments)?;
         return Ok(PyprojectConfig::new(
             PyprojectDiscoveryStrategy::Hierarchical,
             settings,
@@ -97,7 +91,7 @@ pub fn resolve(
     // "closest" `pyproject.toml` file for every Python file later on, so these act
     // as the "default" settings.)
     debug!("Using Ruff default settings");
-    let config = overrides.transform(Configuration::default());
+    let config = config_arguments.transform(Configuration::default());
     let settings = config.into_settings(&path_dedot::CWD)?;
     Ok(PyprojectConfig::new(
         PyprojectDiscoveryStrategy::Hierarchical,

crates/ruff/src/version.rs

@@ -70,7 +70,7 @@ pub(crate) fn version() -> VersionInfo {
 
 #[cfg(test)]
 mod tests {
-    use insta::{assert_display_snapshot, assert_json_snapshot};
+    use insta::{assert_json_snapshot, assert_snapshot};
 
     use super::{CommitInfo, VersionInfo};
 
@@ -80,7 +80,7 @@ mod tests {
             version: "0.0.0".to_string(),
             commit_info: None,
         };
-        assert_display_snapshot!(version);
+        assert_snapshot!(version);
     }
 
     #[test]
@@ -95,7 +95,7 @@ mod tests {
                 commits_since_last_tag: 0,
             }),
         };
-        assert_display_snapshot!(version);
+        assert_snapshot!(version);
     }
 
     #[test]
@@ -110,7 +110,7 @@ mod tests {
                 commits_since_last_tag: 24,
             }),
         };
-        assert_display_snapshot!(version);
+        assert_snapshot!(version);
     }
 
     #[test]

crates/ruff/tests/deprecation.rs

@@ -12,9 +12,10 @@ const STDIN: &str = "l = 1";
 fn ruff_check(show_source: Option<bool>, output_format: Option<String>) -> Command {
     let mut cmd = Command::new(get_cargo_bin(BIN_NAME));
     let output_format = output_format.unwrap_or(format!("{}", SerializationFormat::default(false)));
-    cmd.arg("--output-format");
-    cmd.arg(output_format);
-    cmd.arg("--no-cache");
+    cmd.arg("check")
+        .arg("--output-format")
+        .arg(output_format)
+        .arg("--no-cache");
     match show_source {
         Some(true) => {
             cmd.arg("--show-source");

crates/ruff/tests/format.rs

@@ -7,10 +7,15 @@ use std::str;
 
 use anyhow::Result;
 use insta_cmd::{assert_cmd_snapshot, get_cargo_bin};
+use regex::escape;
 use tempfile::TempDir;
 
 const BIN_NAME: &str = "ruff";
 
+fn tempdir_filter(tempdir: &TempDir) -> String {
+    format!(r"{}\\?/?", escape(tempdir.path().to_str().unwrap()))
+}
+
 #[test]
 fn default_options() {
     assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
@@ -18,7 +23,7 @@ fn default_options() {
         .arg("-")
         .pass_stdin(r#"
 def foo(arg1, arg2,):
-    print('Should\'t change quotes')
+    print('Shouldn\'t change quotes')
 
 
 if condition:
@@ -33,7 +38,7 @@ if condition:
         arg1,
         arg2,
     ):
-        print("Should't change quotes")
+        print("Shouldn't change quotes")
 
 
     if condition:
@@ -90,6 +95,182 @@ fn format_warn_stdin_filename_with_files() {
     "###);
 }
 
+#[test]
+fn nonexistent_config_file() {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(["format", "--config", "foo.toml", "."]), @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    error: invalid value 'foo.toml' for '--config <CONFIG_OPTION>'
+
+      tip: A `--config` flag must either be a path to a `.toml` configuration file
+           or a TOML `<KEY> = <VALUE>` pair overriding a specific configuration
+           option
+
+    It looks like you were trying to pass a path to a configuration file.
+    The path `foo.toml` does not point to a configuration file
+
+    For more information, try '--help'.
+    "###);
+}
+
+#[test]
+fn config_override_rejected_if_invalid_toml() {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(["format", "--config", "foo = bar", "."]), @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    error: invalid value 'foo = bar' for '--config <CONFIG_OPTION>'
+
+      tip: A `--config` flag must either be a path to a `.toml` configuration file
+           or a TOML `<KEY> = <VALUE>` pair overriding a specific configuration
+           option
+
+    The supplied argument is not valid TOML:
+
+    TOML parse error at line 1, column 7
+      |
+    1 | foo = bar
+      |       ^
+    invalid string
+    expected `"`, `'`
+
+    For more information, try '--help'.
+    "###);
+}
+
+#[test]
+fn too_many_config_files() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_dot_toml = tempdir.path().join("ruff.toml");
+    let ruff2_dot_toml = tempdir.path().join("ruff2.toml");
+    fs::File::create(&ruff_dot_toml)?;
+    fs::File::create(&ruff2_dot_toml)?;
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/")]
+    }, {
+        assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .arg("format")
+        .arg("--config")
+        .arg(&ruff_dot_toml)
+        .arg("--config")
+        .arg(&ruff2_dot_toml)
+        .arg("."), @r###"
+            success: false
+            exit_code: 2
+            ----- stdout -----
+
+            ----- stderr -----
+            ruff failed
+              Cause: You cannot specify more than one configuration file on the command line.
+
+              tip: remove either `--config=[TMP]/ruff.toml` or `--config=[TMP]/ruff2.toml`.
+                   For more information, try `--help`.
+
+            "###);
+    });
+    Ok(())
+}
+
+#[test]
+fn config_file_and_isolated() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_dot_toml = tempdir.path().join("ruff.toml");
+    fs::File::create(&ruff_dot_toml)?;
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/")]
+    }, {
+        assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .arg("format")
+        .arg("--config")
+        .arg(&ruff_dot_toml)
+        .arg("--isolated")
+        .arg("."), @r###"
+        success: false
+        exit_code: 2
+        ----- stdout -----
+
+        ----- stderr -----
+        ruff failed
+          Cause: The argument `--config=[TMP]/ruff.toml` cannot be used with `--isolated`
+
+          tip: You cannot specify a configuration file and also specify `--isolated`,
+               as `--isolated` causes ruff to ignore all configuration files.
+               For more information, try `--help`.
+
+        "###);
+    });
+    Ok(())
+}
+
+#[test]
+fn config_override_via_cli() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(&ruff_toml, "line-length = 100")?;
+    let fixture = r#"
+def foo():
+    print("looooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooong string")
+
+    "#;
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .arg("format")
+        .arg("--config")
+        .arg(&ruff_toml)
+        // This overrides the long line length set in the config file
+        .args(["--config", "line-length=80"])
+        .arg("-")
+        .pass_stdin(fixture), @r###"
+    success: true
+    exit_code: 0
+    ----- stdout -----
+    def foo():
+        print(
+            "looooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooong string"
+        )
+
+    ----- stderr -----
+    "###);
+    Ok(())
+}
+
+#[test]
+fn config_doubly_overridden_via_cli() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(&ruff_toml, "line-length = 70")?;
+    let fixture = r#"
+def foo():
+    print("looooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooong string")
+
+    "#;
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .arg("format")
+        .arg("--config")
+        .arg(&ruff_toml)
+        // This overrides the long line length set in the config file...
+        .args(["--config", "line-length=80"])
+        // ...but this overrides them both:
+        .args(["--line-length", "100"])
+        .arg("-")
+        .pass_stdin(fixture), @r###"
+    success: true
+    exit_code: 0
+    ----- stdout -----
+    def foo():
+        print("looooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooong string")
+
+    ----- stderr -----
+    "###);
+    Ok(())
+}
+
 #[test]
 fn format_options() -> Result<()> {
     let tempdir = TempDir::new()?;
@@ -177,58 +358,52 @@ def f(x):
     '''
     pass
 "#), @r###"
-success: true
-exit_code: 0
------ stdout -----
-def f(x):
-    """
-    Something about `f`. And an example:
+    success: true
+    exit_code: 0
+    ----- stdout -----
+    def f(x):
+        """
+        Something about `f`. And an example:
 
-    .. code-block:: python
+        .. code-block:: python
 
-        (
-            foo,
-            bar,
-            quux,
-        ) = this_is_a_long_line(
-            lion,
-            hippo,
-            lemur,
-            bear,
+            foo, bar, quux = (
+                this_is_a_long_line(
+                    lion,
+                    hippo,
+                    lemur,
+                    bear,
+                )
+            )
+
+        Another example:
+
+        ```py
+        foo, bar, quux = (
+            this_is_a_long_line(
+                lion,
+                hippo,
+                lemur,
+                bear,
+            )
         )
+        ```
 
-    Another example:
+        And another:
 
-    ```py
-    (
-        foo,
-        bar,
-        quux,
-    ) = this_is_a_long_line(
-        lion,
-        hippo,
-        lemur,
-        bear,
-    )
-    ```
+        >>> foo, bar, quux = (
+        ...     this_is_a_long_line(
+        ...         lion,
+        ...         hippo,
+        ...         lemur,
+        ...         bear,
+        ...     )
+        ... )
+        """
+        pass
 
-    And another:
-
-    >>> (
-    ...     foo,
-    ...     bar,
-    ...     quux,
-    ... ) = this_is_a_long_line(
-    ...     lion,
-    ...     hippo,
-    ...     lemur,
-    ...     bear,
-    ... )
-    """
-    pass
-
------ stderr -----
-"###);
+    ----- stderr -----
+    "###);
     Ok(())
 }
 
@@ -348,7 +523,7 @@ from module import =
     ----- stdout -----
 
     ----- stderr -----
-    error: Failed to parse main.py:2:20: Unexpected token '='
+    error: Failed to parse main.py:2:20: Expected an import name
     "###);
 
     Ok(())

crates/ruff/tests/integration_test.rs

@@ -71,6 +71,7 @@ impl<'a> RuffCheck<'a> {
     /// Generate a [`Command`] for the `ruff check` command.
     fn build(self) -> Command {
         let mut cmd = ruff_cmd();
+        cmd.arg("check");
         if let Some(output_format) = self.output_format {
             cmd.args(["--output-format", output_format]);
         }
@@ -100,6 +101,7 @@ fn stdin_success() {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     "###);
@@ -221,6 +223,7 @@ fn stdin_source_type_pyi() {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     "###);
@@ -589,6 +592,7 @@ fn stdin_fix_when_no_issues_should_still_print_contents() {
     print(sys.version)
 
     ----- stderr -----
+    All checks passed!
     "###);
 }
 
@@ -727,11 +731,11 @@ fn stdin_parse_error() {
     success: false
     exit_code: 1
     ----- stdout -----
-    -:1:17: E999 SyntaxError: Unexpected token '='
+    -:1:17: E999 SyntaxError: Expected an import name
     Found 1 error.
 
     ----- stderr -----
-    error: Failed to parse at 1:17: Unexpected token '='
+    error: Failed to parse at 1:17: Expected an import name
     "###);
 }
 
@@ -805,13 +809,13 @@ fn full_output_format() {
 }
 
 #[test]
-fn explain_status_codes_f401() {
-    assert_cmd_snapshot!(ruff_cmd().args(["--explain", "F401"]));
+fn rule_f401() {
+    assert_cmd_snapshot!(ruff_cmd().args(["rule", "F401"]));
 }
 
 #[test]
-fn explain_status_codes_ruf404() {
-    assert_cmd_snapshot!(ruff_cmd().args(["--explain", "RUF404"]), @r###"
+fn rule_invalid_rule_name() {
+    assert_cmd_snapshot!(ruff_cmd().args(["rule", "RUF404"]), @r###"
     success: false
     exit_code: 2
     ----- stdout -----
@@ -1022,6 +1026,7 @@ fn preview_disabled_direct() {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     warning: Selection `RUF911` has no effect because preview is not enabled.
@@ -1038,6 +1043,7 @@ fn preview_disabled_prefix_empty() {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     warning: Selection `RUF91` has no effect because preview is not enabled.
@@ -1054,6 +1060,7 @@ fn preview_disabled_does_not_warn_for_empty_ignore_selections() {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     "###);
@@ -1069,6 +1076,7 @@ fn preview_disabled_does_not_warn_for_empty_fixable_selections() {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     "###);
@@ -1174,6 +1182,7 @@ fn removed_indirect() {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     "###);
@@ -1204,6 +1213,7 @@ fn redirect_indirect() {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     "###);
@@ -1306,6 +1316,7 @@ fn deprecated_indirect_preview_enabled() {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     "###);
@@ -1342,13 +1353,14 @@ fn unreadable_pyproject_toml() -> Result<()> {
     // Create an empty file with 000 permissions
     fs::OpenOptions::new()
         .create(true)
+        .truncate(true)
         .write(true)
         .mode(0o000)
         .open(pyproject_toml)?;
 
     // Don't `--isolated` since the configuration discovery is where the error happens
     let args = Args::parse_from(["", "check", "--no-cache", tempdir.path().to_str().unwrap()]);
-    let err = run(args).err().context("Unexpected success")?;
+    let err = run(args, None).err().context("Unexpected success")?;
     assert_eq!(
         err.chain()
             .map(std::string::ToString::to_string)
@@ -1382,6 +1394,7 @@ fn unreadable_dir() -> Result<()> {
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     warning: Encountered error: Permission denied (os error 13)
@@ -1896,6 +1909,7 @@ def log(x, base) -> float:
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     "###

crates/ruff/tests/lint.rs

@@ -12,7 +12,7 @@ use insta_cmd::{assert_cmd_snapshot, get_cargo_bin};
 use tempfile::TempDir;
 
 const BIN_NAME: &str = "ruff";
-const STDIN_BASE_OPTIONS: &[&str] = &["--no-cache", "--output-format", "concise"];
+const STDIN_BASE_OPTIONS: &[&str] = &["check", "--no-cache", "--output-format", "concise"];
 
 fn tempdir_filter(tempdir: &TempDir) -> String {
     format!(r"{}\\?/?", escape(tempdir.path().to_str().unwrap()))
@@ -246,7 +246,6 @@ OTHER = "OTHER"
     }, {
     assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
         .current_dir(tempdir.path())
-        .arg("check")
         .args(STDIN_BASE_OPTIONS)
         .args(["--config", &ruff_toml.file_name().unwrap().to_string_lossy()])
         // Explicitly pass test.py, should be linted regardless of it being excluded by lint.exclude
@@ -293,7 +292,6 @@ inline-quotes = "single"
     }, {
     assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
         .current_dir(tempdir.path())
-        .arg("check")
         .args(STDIN_BASE_OPTIONS)
         .args(["--config", &ruff_toml.file_name().unwrap().to_string_lossy()])
         .args(["--stdin-filename", "generated.py"])
@@ -386,7 +384,6 @@ inline-quotes = "single"
     }, {
     assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
         .current_dir(tempdir.path())
-        .arg("check")
         .args(STDIN_BASE_OPTIONS)
         .args(["--config", &ruff_toml.file_name().unwrap().to_string_lossy()])
         .args(["--stdin-filename", "generated.py"])
@@ -435,7 +432,6 @@ inline-quotes = "single"
     }, {
     assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
         .current_dir(tempdir.path())
-        .arg("check")
         .args(STDIN_BASE_OPTIONS)
         .args(["--config", &ruff_toml.file_name().unwrap().to_string_lossy()])
         .args(["--stdin-filename", "generated.py"])
@@ -495,12 +491,12 @@ ignore = ["D203", "D212"]
     }, {
     assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
         .current_dir(sub_dir)
-        .arg("check")
         .args(STDIN_BASE_OPTIONS)
         , @r###"
     success: true
     exit_code: 0
     ----- stdout -----
+    All checks passed!
 
     ----- stderr -----
     warning: No Python files found under the given path(s)
@@ -510,6 +506,360 @@ ignore = ["D203", "D212"]
     Ok(())
 }
 
+#[test]
+fn nonexistent_config_file() {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .args(["--config", "foo.toml", "."]), @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    error: invalid value 'foo.toml' for '--config <CONFIG_OPTION>'
+
+      tip: A `--config` flag must either be a path to a `.toml` configuration file
+           or a TOML `<KEY> = <VALUE>` pair overriding a specific configuration
+           option
+
+    It looks like you were trying to pass a path to a configuration file.
+    The path `foo.toml` does not point to a configuration file
+
+    For more information, try '--help'.
+    "###);
+}
+
+#[test]
+fn config_override_rejected_if_invalid_toml() {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .args(["--config", "foo = bar", "."]), @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    error: invalid value 'foo = bar' for '--config <CONFIG_OPTION>'
+
+      tip: A `--config` flag must either be a path to a `.toml` configuration file
+           or a TOML `<KEY> = <VALUE>` pair overriding a specific configuration
+           option
+
+    The supplied argument is not valid TOML:
+
+    TOML parse error at line 1, column 7
+      |
+    1 | foo = bar
+      |       ^
+    invalid string
+    expected `"`, `'`
+
+    For more information, try '--help'.
+    "###);
+}
+
+#[test]
+fn too_many_config_files() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_dot_toml = tempdir.path().join("ruff.toml");
+    let ruff2_dot_toml = tempdir.path().join("ruff2.toml");
+    fs::File::create(&ruff_dot_toml)?;
+    fs::File::create(&ruff2_dot_toml)?;
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/")]
+    }, {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_dot_toml)
+        .arg("--config")
+        .arg(&ruff2_dot_toml)
+        .arg("."), @r###"
+        success: false
+        exit_code: 2
+        ----- stdout -----
+
+        ----- stderr -----
+        ruff failed
+          Cause: You cannot specify more than one configuration file on the command line.
+
+          tip: remove either `--config=[TMP]/ruff.toml` or `--config=[TMP]/ruff2.toml`.
+               For more information, try `--help`.
+
+        "###);
+    });
+    Ok(())
+}
+
+#[test]
+fn extend_passed_via_config_argument() {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .args(["--config", "extend = 'foo.toml'", "."]), @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    error: invalid value 'extend = 'foo.toml'' for '--config <CONFIG_OPTION>'
+
+      tip: Cannot include `extend` in a --config flag value
+
+    For more information, try '--help'.
+    "###);
+}
+
+#[test]
+fn config_file_and_isolated() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_dot_toml = tempdir.path().join("ruff.toml");
+    fs::File::create(&ruff_dot_toml)?;
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/")]
+    }, {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_dot_toml)
+        .arg("--isolated")
+        .arg("."), @r###"
+        success: false
+        exit_code: 2
+        ----- stdout -----
+
+        ----- stderr -----
+        ruff failed
+          Cause: The argument `--config=[TMP]/ruff.toml` cannot be used with `--isolated`
+
+          tip: You cannot specify a configuration file and also specify `--isolated`,
+               as `--isolated` causes ruff to ignore all configuration files.
+               For more information, try `--help`.
+
+        "###);
+    });
+    Ok(())
+}
+
+#[test]
+fn config_override_via_cli() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        r#"
+line-length = 100
+
+[lint]
+select = ["I"]
+
+[lint.isort]
+combine-as-imports = true
+        "#,
+    )?;
+    let fixture = r#"
+from foo import (
+    aaaaaaaaaaaaaaaaaaa,
+    bbbbbbbbbbb as bbbbbbbbbbbbbbbb,
+    cccccccccccccccc,
+    ddddddddddd as ddddddddddddd,
+    eeeeeeeeeeeeeee,
+    ffffffffffff as ffffffffffffff,
+    ggggggggggggg,
+    hhhhhhh as hhhhhhhhhhh,
+    iiiiiiiiiiiiii,
+    jjjjjjjjjjjjj as jjjjjj,
+)
+
+x = "longer_than_90_charactersssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss"
+"#;
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_toml)
+        .args(["--config", "line-length=90"])
+        .args(["--config", "lint.extend-select=['E501', 'F841']"])
+        .args(["--config", "lint.isort.combine-as-imports = false"])
+        .arg("-")
+        .pass_stdin(fixture), @r###"
+    success: false
+    exit_code: 1
+    ----- stdout -----
+    -:2:1: I001 [*] Import block is un-sorted or un-formatted
+    -:15:91: E501 Line too long (97 > 90)
+    Found 2 errors.
+    [*] 1 fixable with the `--fix` option.
+
+    ----- stderr -----
+    "###);
+    Ok(())
+}
+
+#[test]
+fn valid_toml_but_nonexistent_option_provided_via_config_argument() {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .args([".", "--config", "extend-select=['F481']"]),  // No such code as F481!
+        @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    error: invalid value 'extend-select=['F481']' for '--config <CONFIG_OPTION>'
+
+      tip: A `--config` flag must either be a path to a `.toml` configuration file
+           or a TOML `<KEY> = <VALUE>` pair overriding a specific configuration
+           option
+
+    Could not parse the supplied argument as a `ruff.toml` configuration option:
+
+    Unknown rule selector: `F481`
+
+    For more information, try '--help'.
+    "###);
+}
+
+#[test]
+fn each_toml_option_requires_a_new_flag_1() {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        // commas can't be used to delimit different config overrides;
+        // you need a new --config flag for each override
+        .args([".", "--config", "extend-select=['F841'], line-length=90"]),
+        @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    error: invalid value 'extend-select=['F841'], line-length=90' for '--config <CONFIG_OPTION>'
+
+      tip: A `--config` flag must either be a path to a `.toml` configuration file
+           or a TOML `<KEY> = <VALUE>` pair overriding a specific configuration
+           option
+
+    The supplied argument is not valid TOML:
+
+    TOML parse error at line 1, column 23
+      |
+    1 | extend-select=['F841'], line-length=90
+      |                       ^
+    expected newline, `#`
+
+    For more information, try '--help'.
+    "###);
+}
+
+#[test]
+fn each_toml_option_requires_a_new_flag_2() {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        // spaces *also* can't be used to delimit different config overrides;
+        // you need a new --config flag for each override
+        .args([".", "--config", "extend-select=['F841'] line-length=90"]),
+        @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    error: invalid value 'extend-select=['F841'] line-length=90' for '--config <CONFIG_OPTION>'
+
+      tip: A `--config` flag must either be a path to a `.toml` configuration file
+           or a TOML `<KEY> = <VALUE>` pair overriding a specific configuration
+           option
+
+    The supplied argument is not valid TOML:
+
+    TOML parse error at line 1, column 24
+      |
+    1 | extend-select=['F841'] line-length=90
+      |                        ^
+    expected newline, `#`
+
+    For more information, try '--help'.
+    "###);
+}
+
+#[test]
+fn config_doubly_overridden_via_cli() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        r#"
+line-length = 100
+
+[lint]
+select=["E501"]
+"#,
+    )?;
+    let fixture = "x = 'longer_than_90_charactersssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssss'";
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        // The --line-length flag takes priority over both the config file
+        // and the `--config="line-length=110"` flag,
+        // despite them both being specified after this flag on the command line:
+        .args(["--line-length", "90"])
+        .arg("--config")
+        .arg(&ruff_toml)
+        .args(["--config", "line-length=110"])
+        .arg("-")
+        .pass_stdin(fixture), @r###"
+    success: false
+    exit_code: 1
+    ----- stdout -----
+    -:1:91: E501 Line too long (97 > 90)
+    Found 1 error.
+
+    ----- stderr -----
+    "###);
+    Ok(())
+}
+
+#[test]
+fn complex_config_setting_overridden_via_cli() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(&ruff_toml, "lint.select = ['N801']")?;
+    let fixture = "class violates_n801: pass";
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_toml)
+        .args(["--config", "lint.per-file-ignores = {'generated.py' = ['N801']}"])
+        .args(["--stdin-filename", "generated.py"])
+        .arg("-")
+        .pass_stdin(fixture), @r###"
+    success: true
+    exit_code: 0
+    ----- stdout -----
+    All checks passed!
+
+    ----- stderr -----
+    "###);
+    Ok(())
+}
+
+#[test]
+fn deprecated_config_option_overridden_via_cli() {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .args(["--config", "select=['N801']", "-"])
+        .pass_stdin("class lowercase: ..."),
+        @r###"
+    success: false
+    exit_code: 1
+    ----- stdout -----
+    -:1:7: N801 Class name `lowercase` should use CapWords convention
+    Found 1 error.
+
+    ----- stderr -----
+    warning: The top-level linter settings are deprecated in favour of their counterparts in the `lint` section. Please update the following options in your `--config` CLI arguments:
+      - 'select' -> 'lint.select'
+    "###);
+}
+
 #[test]
 fn extension() -> Result<()> {
     let tempdir = TempDir::new()?;
@@ -568,7 +918,6 @@ include = ["*.ipy"]
     }, {
     assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
         .current_dir(tempdir.path())
-        .arg("check")
         .args(STDIN_BASE_OPTIONS)
         .args(["--config", &ruff_toml.file_name().unwrap().to_string_lossy()])
         .args(["--extension", "ipy:ipynb"])
@@ -586,3 +935,398 @@ include = ["*.ipy"]
 
     Ok(())
 }
+
+#[test]
+fn file_noqa_external() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        r#"
+[lint]
+external = ["AAA"]
+"#,
+    )?;
+
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/")]
+    }, {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_toml)
+        .arg("-")
+        .pass_stdin(r#"
+# flake8: noqa: AAA101, BBB102
+import os
+"#), @r###"
+    success: false
+    exit_code: 1
+    ----- stdout -----
+    -:3:8: F401 [*] `os` imported but unused
+    Found 1 error.
+    [*] 1 fixable with the `--fix` option.
+
+    ----- stderr -----
+    warning: Invalid rule code provided to `# ruff: noqa` at -:2: BBB102
+    "###);
+    });
+
+    Ok(())
+}
+
+#[test]
+fn required_version_exact_mismatch() -> Result<()> {
+    let version = env!("CARGO_PKG_VERSION");
+
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        r#"
+required-version = "0.1.0"
+"#,
+    )?;
+
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/"), (version, "[VERSION]")]
+    }, {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_toml)
+        .arg("-")
+        .pass_stdin(r#"
+import os
+"#), @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    ruff failed
+      Cause: Required version `==0.1.0` does not match the running version `[VERSION]`
+    "###);
+    });
+
+    Ok(())
+}
+
+#[test]
+fn required_version_exact_match() -> Result<()> {
+    let version = env!("CARGO_PKG_VERSION");
+
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        format!(
+            r#"
+required-version = "{version}"
+"#
+        ),
+    )?;
+
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/"), (version, "[VERSION]")]
+    }, {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_toml)
+        .arg("-")
+        .pass_stdin(r#"
+import os
+"#), @r###"
+    success: false
+    exit_code: 1
+    ----- stdout -----
+    -:2:8: F401 [*] `os` imported but unused
+    Found 1 error.
+    [*] 1 fixable with the `--fix` option.
+
+    ----- stderr -----
+    "###);
+    });
+
+    Ok(())
+}
+
+#[test]
+fn required_version_bound_mismatch() -> Result<()> {
+    let version = env!("CARGO_PKG_VERSION");
+
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        format!(
+            r#"
+required-version = ">{version}"
+"#
+        ),
+    )?;
+
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/"), (version, "[VERSION]")]
+    }, {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_toml)
+        .arg("-")
+        .pass_stdin(r#"
+import os
+"#), @r###"
+    success: false
+    exit_code: 2
+    ----- stdout -----
+
+    ----- stderr -----
+    ruff failed
+      Cause: Required version `>[VERSION]` does not match the running version `[VERSION]`
+    "###);
+    });
+
+    Ok(())
+}
+
+#[test]
+fn required_version_bound_match() -> Result<()> {
+    let version = env!("CARGO_PKG_VERSION");
+
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        r#"
+required-version = ">=0.1.0"
+"#,
+    )?;
+
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/"), (version, "[VERSION]")]
+    }, {
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_toml)
+        .arg("-")
+        .pass_stdin(r#"
+import os
+"#), @r###"
+    success: false
+    exit_code: 1
+    ----- stdout -----
+    -:2:8: F401 [*] `os` imported but unused
+    Found 1 error.
+    [*] 1 fixable with the `--fix` option.
+
+    ----- stderr -----
+    "###);
+    });
+
+    Ok(())
+}
+
+/// Expand environment variables in `--config` paths provided via the CLI.
+#[test]
+fn config_expand() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        ruff_toml,
+        r#"
+[lint]
+select = ["F"]
+ignore = ["F841"]
+"#,
+    )?;
+
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg("${NAME}.toml")
+        .env("NAME", "ruff")
+        .arg("-")
+        .current_dir(tempdir.path())
+        .pass_stdin(r#"
+import os
+
+def func():
+    x = 1
+"#), @r###"
+    success: false
+    exit_code: 1
+    ----- stdout -----
+    -:2:8: F401 [*] `os` imported but unused
+    Found 1 error.
+    [*] 1 fixable with the `--fix` option.
+
+    ----- stderr -----
+    "###);
+
+    Ok(())
+}
+
+/// Per-file selects via ! negation in per-file-ignores
+#[test]
+fn negated_per_file_ignores() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        r#"
+[lint.per-file-ignores]
+"!selected.py" = ["RUF"]
+"#,
+    )?;
+    let selected = tempdir.path().join("selected.py");
+    fs::write(selected, "")?;
+    let ignored = tempdir.path().join("ignored.py");
+    fs::write(ignored, "")?;
+
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_toml)
+        .arg("--select")
+        .arg("RUF901")
+        .current_dir(&tempdir)
+        , @r###"
+    success: false
+    exit_code: 1
+    ----- stdout -----
+    selected.py:1:1: RUF901 [*] Hey this is a stable test rule with a safe fix.
+    Found 1 error.
+    [*] 1 fixable with the `--fix` option.
+
+    ----- stderr -----
+    "###);
+    Ok(())
+}
+
+#[test]
+fn negated_per_file_ignores_absolute() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        r#"
+[lint.per-file-ignores]
+"!src/**.py" = ["RUF"]
+"#,
+    )?;
+    let src_dir = tempdir.path().join("src");
+    fs::create_dir(&src_dir)?;
+    let selected = src_dir.join("selected.py");
+    fs::write(selected, "")?;
+    let ignored = tempdir.path().join("ignored.py");
+    fs::write(ignored, "")?;
+
+    insta::with_settings!({filters => vec![
+        // Replace windows paths
+        (r"\\", "/"),
+    ]}, {
+        assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+            .args(STDIN_BASE_OPTIONS)
+            .arg("--config")
+            .arg(&ruff_toml)
+            .arg("--select")
+            .arg("RUF901")
+            .current_dir(&tempdir)
+            , @r###"
+        success: false
+        exit_code: 1
+        ----- stdout -----
+        src/selected.py:1:1: RUF901 [*] Hey this is a stable test rule with a safe fix.
+        Found 1 error.
+        [*] 1 fixable with the `--fix` option.
+
+        ----- stderr -----
+        "###);
+    });
+    Ok(())
+}
+
+/// patterns are additive, can't use negative patterns to "un-ignore"
+#[test]
+fn negated_per_file_ignores_overlap() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        r#"
+[lint.per-file-ignores]
+"*.py" = ["RUF"]
+"!foo.py" = ["RUF"]
+"#,
+    )?;
+    let foo_file = tempdir.path().join("foo.py");
+    fs::write(foo_file, "")?;
+    let bar_file = tempdir.path().join("bar.py");
+    fs::write(bar_file, "")?;
+
+    assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+        .args(STDIN_BASE_OPTIONS)
+        .arg("--config")
+        .arg(&ruff_toml)
+        .arg("--select")
+        .arg("RUF901")
+        .current_dir(&tempdir)
+        , @r###"
+    success: true
+    exit_code: 0
+    ----- stdout -----
+    All checks passed!
+
+    ----- stderr -----
+    "###);
+    Ok(())
+}
+
+#[test]
+fn unused_interaction() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_toml = tempdir.path().join("ruff.toml");
+    fs::write(
+        &ruff_toml,
+        r#"
+[lint]
+select = ["F"]
+"#,
+    )?;
+
+    insta::with_settings!({
+        filters => vec![(tempdir_filter(&tempdir).as_str(), "[TMP]/")]
+    }, {
+        assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
+            .args(STDIN_BASE_OPTIONS)
+            .arg("--config")
+            .arg(&ruff_toml)
+            .args(["--stdin-filename", "test.py"])
+            .arg("--fix")
+            .arg("-")
+            .pass_stdin(r#"
+import os  # F401
+
+def function():
+    import os  # F811
+    print(os.name)
+"#), @r###"
+        success: true
+        exit_code: 0
+        ----- stdout -----
+
+        import os  # F401
+
+        def function():
+            print(os.name)
+
+        ----- stderr -----
+        Found 1 error (1 fixed, 0 remaining).
+        "###);
+    });
+
+    Ok(())
+}

crates/ruff/tests/show_settings.rs

@@ -4,25 +4,29 @@ use std::process::Command;
 
 const BIN_NAME: &str = "ruff";
 
-#[cfg(not(target_os = "windows"))]
-const TEST_FILTERS: &[(&str, &str)] = &[
-    ("\"[^\\*\"]*/pyproject.toml", "\"[BASEPATH]/pyproject.toml"),
-    ("\".*/crates", "\"[BASEPATH]/crates"),
-    ("\".*/\\.ruff_cache", "\"[BASEPATH]/.ruff_cache"),
-    ("\".*/ruff\"", "\"[BASEPATH]\""),
-];
-#[cfg(target_os = "windows")]
-const TEST_FILTERS: &[(&str, &str)] = &[
-    (r#""[^\*"]*\\pyproject.toml"#, "\"[BASEPATH]/pyproject.toml"),
-    (r#"".*\\crates"#, "\"[BASEPATH]/crates"),
-    (r#"".*\\\.ruff_cache"#, "\"[BASEPATH]/.ruff_cache"),
-    (r#"".*\\ruff""#, "\"[BASEPATH]\""),
-    (r#"\\+(\w\w|\s|")"#, "/$1"),
-];
-
 #[test]
 fn display_default_settings() {
-    insta::with_settings!({ filters => TEST_FILTERS.to_vec() }, {
+    // Navigate from the crate directory to the workspace root.
+    let base_path = Path::new(env!("CARGO_MANIFEST_DIR"))
+        .parent()
+        .unwrap()
+        .parent()
+        .unwrap();
+    let base_path = base_path.to_string_lossy();
+
+    // Escape the backslashes for the regex.
+    let base_path = regex::escape(&base_path);
+
+    #[cfg(not(target_os = "windows"))]
+    let test_filters = &[(base_path.as_ref(), "[BASEPATH]")];
+
+    #[cfg(target_os = "windows")]
+    let test_filters = &[
+        (base_path.as_ref(), "[BASEPATH]"),
+        (r#"\\+(\w\w|\s|\.|")"#, "/$1"),
+    ];
+
+    insta::with_settings!({ filters => test_filters.to_vec() }, {
         assert_cmd_snapshot!(Command::new(get_cargo_bin(BIN_NAME))
             .args(["check", "--show-settings", "unformatted.py"]).current_dir(Path::new("./resources/test/fixtures")));
     });

crates/ruff/tests/snapshots/integration_test__rule_f401.snap

@@ -3,7 +3,7 @@ source: crates/ruff/tests/integration_test.rs
 info:
   program: ruff
   args:
-    - "--explain"
+    - rule
     - F401
 ---
 success: true
@@ -25,6 +25,20 @@ import cycles. They also increase the cognitive load of reading the code.
 If an import statement is used to check for the availability or existence
 of a module, consider using `importlib.util.find_spec` instead.
 
+If an import statement is used to re-export a symbol as part of a module's
+public interface, consider using a "redundant" import alias, which
+instructs Ruff (and other tools) to respect the re-export, and avoid
+marking it as unused, as in:
+
+```python
+from module import member as member
+```
+
+## Fix safety
+
+When `ignore_init_module_imports` is disabled, fixes can remove for unused imports in `__init__` files.
+These fixes are considered unsafe because they can change the public interface.
+
 ## Example
 ```python
 import numpy as np  # unused import
@@ -51,11 +65,11 @@ else:
 ```
 
 ## Options
-- `lint.pyflakes.extend-generics`
+- `lint.ignore-init-module-imports`
 
 ## References
 - [Python documentation: `import`](https://docs.python.org/3/reference/simple_stmts.html#the-import-statement)
 - [Python documentation: `importlib.util.find_spec`](https://docs.python.org/3/library/importlib.html#importlib.util.find_spec)
+- [Typing documentation: interface conventions](https://typing.readthedocs.io/en/latest/source/libraries.html#library-interface-public-and-private-symbols)
 
 ----- stderr -----
-

crates/ruff/tests/snapshots/show_settings__display_default_settings.snap

@@ -44,15 +44,16 @@ file_resolver.exclude = [
 	"__pypackages__",
 	"_build",
 	"buck-out",
-	"build",
 	"dist",
 	"node_modules",
 	"site-packages",
 	"venv",
 ]
 file_resolver.extend_exclude = [
+	"crates/ruff/resources/",
 	"crates/ruff_linter/resources/",
 	"crates/ruff_python_formatter/resources/",
+	"crates/ruff_python_parser/resources/",
 ]
 file_resolver.force_exclude = false
 file_resolver.include = [
@@ -68,128 +69,128 @@ file_resolver.project_root = "[BASEPATH]"
 linter.exclude = []
 linter.project_root = "[BASEPATH]"
 linter.rules.enabled = [
-	MultipleImportsOnOneLine,
-	ModuleImportNotAtTopOfFile,
-	MultipleStatementsOnOneLineColon,
-	MultipleStatementsOnOneLineSemicolon,
-	UselessSemicolon,
-	NoneComparison,
-	TrueFalseComparison,
-	NotInTest,
-	NotIsTest,
-	TypeComparison,
-	BareExcept,
-	LambdaAssignment,
-	AmbiguousVariableName,
-	AmbiguousClassName,
-	AmbiguousFunctionName,
-	IOError,
-	SyntaxError,
-	UnusedImport,
-	ImportShadowedByLoopVar,
-	UndefinedLocalWithImportStar,
-	LateFutureImport,
-	UndefinedLocalWithImportStarUsage,
-	UndefinedLocalWithNestedImportStarUsage,
-	FutureFeatureNotDefined,
-	PercentFormatInvalidFormat,
-	PercentFormatExpectedMapping,
-	PercentFormatExpectedSequence,
-	PercentFormatExtraNamedArguments,
-	PercentFormatMissingArgument,
-	PercentFormatMixedPositionalAndNamed,
-	PercentFormatPositionalCountMismatch,
-	PercentFormatStarRequiresSequence,
-	PercentFormatUnsupportedFormatCharacter,
-	StringDotFormatInvalidFormat,
-	StringDotFormatExtraNamedArguments,
-	StringDotFormatExtraPositionalArguments,
-	StringDotFormatMissingArguments,
-	StringDotFormatMixingAutomatic,
-	FStringMissingPlaceholders,
-	MultiValueRepeatedKeyLiteral,
-	MultiValueRepeatedKeyVariable,
-	ExpressionsInStarAssignment,
-	MultipleStarredExpressions,
-	AssertTuple,
-	IsLiteral,
-	InvalidPrintSyntax,
-	IfTuple,
-	BreakOutsideLoop,
-	ContinueOutsideLoop,
-	YieldOutsideFunction,
-	ReturnOutsideFunction,
-	DefaultExceptNotLast,
-	ForwardAnnotationSyntaxError,
-	RedefinedWhileUnused,
-	UndefinedName,
-	UndefinedExport,
-	UndefinedLocal,
-	UnusedVariable,
-	UnusedAnnotation,
-	RaiseNotImplemented,
+	multiple-imports-on-one-line (E401),
+	module-import-not-at-top-of-file (E402),
+	multiple-statements-on-one-line-colon (E701),
+	multiple-statements-on-one-line-semicolon (E702),
+	useless-semicolon (E703),
+	none-comparison (E711),
+	true-false-comparison (E712),
+	not-in-test (E713),
+	not-is-test (E714),
+	type-comparison (E721),
+	bare-except (E722),
+	lambda-assignment (E731),
+	ambiguous-variable-name (E741),
+	ambiguous-class-name (E742),
+	ambiguous-function-name (E743),
+	io-error (E902),
+	syntax-error (E999),
+	unused-import (F401),
+	import-shadowed-by-loop-var (F402),
+	undefined-local-with-import-star (F403),
+	late-future-import (F404),
+	undefined-local-with-import-star-usage (F405),
+	undefined-local-with-nested-import-star-usage (F406),
+	future-feature-not-defined (F407),
+	percent-format-invalid-format (F501),
+	percent-format-expected-mapping (F502),
+	percent-format-expected-sequence (F503),
+	percent-format-extra-named-arguments (F504),
+	percent-format-missing-argument (F505),
+	percent-format-mixed-positional-and-named (F506),
+	percent-format-positional-count-mismatch (F507),
+	percent-format-star-requires-sequence (F508),
+	percent-format-unsupported-format-character (F509),
+	string-dot-format-invalid-format (F521),
+	string-dot-format-extra-named-arguments (F522),
+	string-dot-format-extra-positional-arguments (F523),
+	string-dot-format-missing-arguments (F524),
+	string-dot-format-mixing-automatic (F525),
+	f-string-missing-placeholders (F541),
+	multi-value-repeated-key-literal (F601),
+	multi-value-repeated-key-variable (F602),
+	expressions-in-star-assignment (F621),
+	multiple-starred-expressions (F622),
+	assert-tuple (F631),
+	is-literal (F632),
+	invalid-print-syntax (F633),
+	if-tuple (F634),
+	break-outside-loop (F701),
+	continue-outside-loop (F702),
+	yield-outside-function (F704),
+	return-outside-function (F706),
+	default-except-not-last (F707),
+	forward-annotation-syntax-error (F722),
+	redefined-while-unused (F811),
+	undefined-name (F821),
+	undefined-export (F822),
+	undefined-local (F823),
+	unused-variable (F841),
+	unused-annotation (F842),
+	raise-not-implemented (F901),
 ]
 linter.rules.should_fix = [
-	MultipleImportsOnOneLine,
-	ModuleImportNotAtTopOfFile,
-	MultipleStatementsOnOneLineColon,
-	MultipleStatementsOnOneLineSemicolon,
-	UselessSemicolon,
-	NoneComparison,
-	TrueFalseComparison,
-	NotInTest,
-	NotIsTest,
-	TypeComparison,
-	BareExcept,
-	LambdaAssignment,
-	AmbiguousVariableName,
-	AmbiguousClassName,
-	AmbiguousFunctionName,
-	IOError,
-	SyntaxError,
-	UnusedImport,
-	ImportShadowedByLoopVar,
-	UndefinedLocalWithImportStar,
-	LateFutureImport,
-	UndefinedLocalWithImportStarUsage,
-	UndefinedLocalWithNestedImportStarUsage,
-	FutureFeatureNotDefined,
-	PercentFormatInvalidFormat,
-	PercentFormatExpectedMapping,
-	PercentFormatExpectedSequence,
-	PercentFormatExtraNamedArguments,
-	PercentFormatMissingArgument,
-	PercentFormatMixedPositionalAndNamed,
-	PercentFormatPositionalCountMismatch,
-	PercentFormatStarRequiresSequence,
-	PercentFormatUnsupportedFormatCharacter,
-	StringDotFormatInvalidFormat,
-	StringDotFormatExtraNamedArguments,
-	StringDotFormatExtraPositionalArguments,
-	StringDotFormatMissingArguments,
-	StringDotFormatMixingAutomatic,
-	FStringMissingPlaceholders,
-	MultiValueRepeatedKeyLiteral,
-	MultiValueRepeatedKeyVariable,
-	ExpressionsInStarAssignment,
-	MultipleStarredExpressions,
-	AssertTuple,
-	IsLiteral,
-	InvalidPrintSyntax,
-	IfTuple,
-	BreakOutsideLoop,
-	ContinueOutsideLoop,
-	YieldOutsideFunction,
-	ReturnOutsideFunction,
-	DefaultExceptNotLast,
-	ForwardAnnotationSyntaxError,
-	RedefinedWhileUnused,
-	UndefinedName,
-	UndefinedExport,
-	UndefinedLocal,
-	UnusedVariable,
-	UnusedAnnotation,
-	RaiseNotImplemented,
+	multiple-imports-on-one-line (E401),
+	module-import-not-at-top-of-file (E402),
+	multiple-statements-on-one-line-colon (E701),
+	multiple-statements-on-one-line-semicolon (E702),
+	useless-semicolon (E703),
+	none-comparison (E711),
+	true-false-comparison (E712),
+	not-in-test (E713),
+	not-is-test (E714),
+	type-comparison (E721),
+	bare-except (E722),
+	lambda-assignment (E731),
+	ambiguous-variable-name (E741),
+	ambiguous-class-name (E742),
+	ambiguous-function-name (E743),
+	io-error (E902),
+	syntax-error (E999),
+	unused-import (F401),
+	import-shadowed-by-loop-var (F402),
+	undefined-local-with-import-star (F403),
+	late-future-import (F404),
+	undefined-local-with-import-star-usage (F405),
+	undefined-local-with-nested-import-star-usage (F406),
+	future-feature-not-defined (F407),
+	percent-format-invalid-format (F501),
+	percent-format-expected-mapping (F502),
+	percent-format-expected-sequence (F503),
+	percent-format-extra-named-arguments (F504),
+	percent-format-missing-argument (F505),
+	percent-format-mixed-positional-and-named (F506),
+	percent-format-positional-count-mismatch (F507),
+	percent-format-star-requires-sequence (F508),
+	percent-format-unsupported-format-character (F509),
+	string-dot-format-invalid-format (F521),
+	string-dot-format-extra-named-arguments (F522),
+	string-dot-format-extra-positional-arguments (F523),
+	string-dot-format-missing-arguments (F524),
+	string-dot-format-mixing-automatic (F525),
+	f-string-missing-placeholders (F541),
+	multi-value-repeated-key-literal (F601),
+	multi-value-repeated-key-variable (F602),
+	expressions-in-star-assignment (F621),
+	multiple-starred-expressions (F622),
+	assert-tuple (F631),
+	is-literal (F632),
+	invalid-print-syntax (F633),
+	if-tuple (F634),
+	break-outside-loop (F701),
+	continue-outside-loop (F702),
+	yield-outside-function (F704),
+	return-outside-function (F706),
+	default-except-not-last (F707),
+	forward-annotation-syntax-error (F722),
+	redefined-while-unused (F811),
+	undefined-name (F821),
+	undefined-export (F822),
+	undefined-local (F823),
+	unused-variable (F841),
+	unused-annotation (F842),
+	raise-not-implemented (F901),
 ]
 linter.per_file_ignores = {}
 linter.safety_table.forced_safe = []
@@ -202,10 +203,12 @@ linter.allowed_confusables = []
 linter.builtins = []
 linter.dummy_variable_rgx = ^(_+|(_+[a-zA-Z0-9_]*[a-zA-Z0-9]+?))$
 linter.external = []
-linter.ignore_init_module_imports = false
+linter.ignore_init_module_imports = true
 linter.logger_objects = []
 linter.namespace_packages = []
-linter.src = ["[BASEPATH]"]
+linter.src = [
+	"[BASEPATH]",
+]
 linter.tab_size = 4
 linter.line_length = 88
 linter.task_tags = [
@@ -230,7 +233,7 @@ linter.flake8_bandit.check_typed_exception = false
 linter.flake8_bugbear.extend_immutable_calls = []
 linter.flake8_builtins.builtins_ignorelist = []
 linter.flake8_comprehensions.allow_dict_calls_with_keyword_arguments = false
-linter.flake8_copyright.notice_rgx = (?i)Copyright\s+(\(C\)\s+)?\d{4}(-\d{4})*
+linter.flake8_copyright.notice_rgx = (?i)Copyright\s+((?:\(C\)|©)\s+)?\d{4}((-|,\s)\d{4})*
 linter.flake8_copyright.author = none
 linter.flake8_copyright.min_file_size = 0
 linter.flake8_errmsg.max_string_length = 0
@@ -240,7 +243,22 @@ linter.flake8_gettext.functions_names = [
 	ngettext,
 ]
 linter.flake8_implicit_str_concat.allow_multiline = true
-linter.flake8_import_conventions.aliases = {"matplotlib": "mpl", "matplotlib.pyplot": "plt", "pandas": "pd", "seaborn": "sns", "tensorflow": "tf", "networkx": "nx", "plotly.express": "px", "polars": "pl", "numpy": "np", "panel": "pn", "pyarrow": "pa", "altair": "alt", "tkinter": "tk", "holoviews": "hv"}
+linter.flake8_import_conventions.aliases = {
+	altair = alt,
+	holoviews = hv,
+	matplotlib = mpl,
+	matplotlib.pyplot = plt,
+	networkx = nx,
+	numpy = np,
+	pandas = pd,
+	panel = pn,
+	plotly.express = px,
+	polars = pl,
+	pyarrow = pa,
+	seaborn = sns,
+	tensorflow = tf,
+	tkinter = tk,
+}
 linter.flake8_import_conventions.banned_aliases = {}
 linter.flake8_import_conventions.banned_from = []
 linter.flake8_pytest_style.fixture_parentheses = true
@@ -310,6 +328,7 @@ linter.isort.section_order = [
 	known { type = first_party },
 	known { type = local_folder },
 ]
+linter.isort.default_section = known { type = third_party }
 linter.isort.no_sections = false
 linter.isort.from_first = false
 linter.isort.length_sort = false
@@ -364,4 +383,3 @@ formatter.docstring_code_format = disabled
 formatter.docstring_code_line_width = dynamic
 
 ----- stderr -----
-

crates/ruff/tests/version.rs

@@ -0,0 +1,105 @@
+//! Tests for the --version command
+use std::fs;
+use std::process::Command;
+
+use anyhow::Result;
+use insta_cmd::{assert_cmd_snapshot, get_cargo_bin};
+use tempfile::TempDir;
+
+const BIN_NAME: &str = "ruff";
+const VERSION_FILTER: [(&str, &str); 1] = [(
+    r"\d+\.\d+\.\d+(\+\d+)?( \(\w{9} \d\d\d\d-\d\d-\d\d\))?",
+    "[VERSION]",
+)];
+
+#[test]
+fn version_basics() {
+    insta::with_settings!({filters => VERSION_FILTER.to_vec()}, {
+        assert_cmd_snapshot!(
+            Command::new(get_cargo_bin(BIN_NAME)).arg("version"), @r###"
+        success: true
+        exit_code: 0
+        ----- stdout -----
+        ruff [VERSION]
+
+        ----- stderr -----
+        "###
+        );
+    });
+}
+
+/// `--config` is a global option,
+/// so it's allowed to pass --config to subcommands such as `version`
+/// -- the flag is simply ignored
+#[test]
+fn config_option_allowed_but_ignored() -> Result<()> {
+    let tempdir = TempDir::new()?;
+    let ruff_dot_toml = tempdir.path().join("ruff.toml");
+    fs::File::create(&ruff_dot_toml)?;
+    insta::with_settings!({filters => VERSION_FILTER.to_vec()}, {
+        assert_cmd_snapshot!(
+            Command::new(get_cargo_bin(BIN_NAME))
+                .arg("version")
+                .arg("--config")
+                .arg(&ruff_dot_toml)
+                .args(["--config", "lint.isort.extra-standard-library = ['foo', 'bar']"]), @r###"
+        success: true
+        exit_code: 0
+        ----- stdout -----
+        ruff [VERSION]
+
+        ----- stderr -----
+        "###
+        );
+    });
+    Ok(())
+}
+#[test]
+fn config_option_ignored_but_validated() {
+    insta::with_settings!({filters => VERSION_FILTER.to_vec()}, {
+        assert_cmd_snapshot!(
+            Command::new(get_cargo_bin(BIN_NAME))
+                .arg("version")
+                .args(["--config", "foo = bar"]), @r###"
+        success: false
+        exit_code: 2
+        ----- stdout -----
+
+        ----- stderr -----
+        error: invalid value 'foo = bar' for '--config <CONFIG_OPTION>'
+
+          tip: A `--config` flag must either be a path to a `.toml` configuration file
+               or a TOML `<KEY> = <VALUE>` pair overriding a specific configuration
+               option
+
+        The supplied argument is not valid TOML:
+
+        TOML parse error at line 1, column 7
+          |
+        1 | foo = bar
+          |       ^
+        invalid string
+        expected `"`, `'`
+
+        For more information, try '--help'.
+        "###
+        );
+    });
+}
+
+/// `--isolated` is also a global option,
+#[test]
+fn isolated_option_allowed() {
+    insta::with_settings!({filters => VERSION_FILTER.to_vec()}, {
+        assert_cmd_snapshot!(
+            Command::new(get_cargo_bin(BIN_NAME)).arg("version").arg("--isolated"), @r###"
+        success: true
+        exit_code: 0
+        ----- stdout -----
+        ruff [VERSION]
+
+        ----- stderr -----
+        "###
+        );
+    });
+}

crates/ruff_benchmark/Cargo.toml

@@ -13,6 +13,7 @@ license = { workspace = true }
 
 [lib]
 bench = false
+doctest = false
 
 [[bench]]
 name = "linter"

crates/ruff_cache/Cargo.toml

@@ -11,9 +11,9 @@ repository = { workspace = true }
 license = { workspace = true }
 
 [dependencies]
-itertools = { workspace = true }
 glob = { workspace = true }
 globset = { workspace = true }
+itertools = { workspace = true }
 regex = { workspace = true }
 filetime = { workspace = true }
 seahash = { workspace = true }

crates/ruff_cache/src/cache_key.rs

@@ -65,7 +65,7 @@ use seahash::SeaHasher;
 /// The main reason is that hashes and cache keys have different constraints:
 ///
 /// * Cache keys are less performance sensitive: Hashes must be super fast to compute for performant hashed-collections. That's
-///   why some standard types don't implement [`Hash`] where it would be safe to to implement [`CacheKey`], e.g. `HashSet`
+///   why some standard types don't implement [`Hash`] where it would be safe to implement [`CacheKey`], e.g. `HashSet`
 /// * Cache keys must be deterministic where hash keys do not have this constraint. That's why pointers don't implement [`CacheKey`] but they implement [`Hash`].
 /// * Ideally, cache keys are portable
 ///

crates/ruff_dev/Cargo.toml

@@ -22,7 +22,7 @@ ruff_python_formatter = { path = "../ruff_python_formatter" }
 ruff_python_parser = { path = "../ruff_python_parser" }
 ruff_python_stdlib = { path = "../ruff_python_stdlib" }
 ruff_python_trivia = { path = "../ruff_python_trivia" }
-ruff_workspace = { path = "../ruff_workspace", features = ["schemars"]}
+ruff_workspace = { path = "../ruff_workspace", features = ["schemars"] }
 
 anyhow = { workspace = true }
 clap = { workspace = true, features = ["wrap_help"] }
@@ -31,7 +31,6 @@ imara-diff = { workspace = true }
 indicatif = { workspace = true }
 itertools = { workspace = true }
 libcst = { workspace = true }
-once_cell = { workspace = true }
 pretty_assertions = { workspace = true }
 rayon = { workspace = true }
 regex = { workspace = true }

crates/ruff_dev/src/format_dev.rs

@@ -27,7 +27,7 @@ use tracing_subscriber::layer::SubscriberExt;
 use tracing_subscriber::util::SubscriberInitExt;
 use tracing_subscriber::EnvFilter;
 
-use ruff::args::{CliOverrides, FormatArguments, FormatCommand, LogLevelArgs};
+use ruff::args::{ConfigArguments, FormatArguments, FormatCommand, GlobalConfigArgs, LogLevelArgs};
 use ruff::resolve::resolve;
 use ruff_formatter::{FormatError, LineWidth, PrintError};
 use ruff_linter::logging::LogLevel;
@@ -38,26 +38,21 @@ use ruff_python_formatter::{
 use ruff_python_parser::ParseError;
 use ruff_workspace::resolver::{python_files_in_path, PyprojectConfig, ResolvedFile, Resolver};
 
-fn parse_cli(dirs: &[PathBuf]) -> anyhow::Result<(FormatArguments, CliOverrides)> {
+fn parse_cli(dirs: &[PathBuf]) -> anyhow::Result<(FormatArguments, ConfigArguments)> {
     let args_matches = FormatCommand::command()
         .no_binary_name(true)
         .get_matches_from(dirs);
     let arguments: FormatCommand = FormatCommand::from_arg_matches(&args_matches)?;
-    let (cli, overrides) = arguments.partition();
-    Ok((cli, overrides))
+    let (cli, config_arguments) = arguments.partition(GlobalConfigArgs::default())?;
+    Ok((cli, config_arguments))
 }
 
 /// Find the [`PyprojectConfig`] to use for formatting.
 fn find_pyproject_config(
     cli: &FormatArguments,
-    overrides: &CliOverrides,
+    config_arguments: &ConfigArguments,
 ) -> anyhow::Result<PyprojectConfig> {
-    let mut pyproject_config = resolve(
-        cli.isolated,
-        cli.config.as_deref(),
-        overrides,
-        cli.stdin_filename.as_deref(),
-    )?;
+    let mut pyproject_config = resolve(config_arguments, cli.stdin_filename.as_deref())?;
     // We don't want to format pyproject.toml
     pyproject_config.settings.file_resolver.include = FilePatternSet::try_from_iter([
         FilePattern::Builtin("*.py"),
@@ -72,9 +67,9 @@ fn find_pyproject_config(
 fn ruff_check_paths<'a>(
     pyproject_config: &'a PyprojectConfig,
     cli: &FormatArguments,
-    overrides: &CliOverrides,
+    config_arguments: &ConfigArguments,
 ) -> anyhow::Result<(Vec<Result<ResolvedFile, ignore::Error>>, Resolver<'a>)> {
-    let (paths, resolver) = python_files_in_path(&cli.files, pyproject_config, overrides)?;
+    let (paths, resolver) = python_files_in_path(&cli.files, pyproject_config, config_arguments)?;
     Ok((paths, resolver))
 }
 
@@ -139,7 +134,7 @@ impl Statistics {
         }
     }
 
-    /// We currently prefer the the similarity index, but i'd like to keep this around
+    /// We currently prefer the similarity index, but i'd like to keep this around
     #[allow(clippy::cast_precision_loss, unused)]
     pub(crate) fn jaccard_index(&self) -> f32 {
         self.intersection as f32 / (self.black_input + self.ruff_output + self.intersection) as f32

crates/ruff_dev/src/generate_docs.rs

@@ -1,6 +1,7 @@
 //! Generate Markdown documentation for applicable rules.
 #![allow(clippy::print_stdout, clippy::print_stderr)]
 
+use std::collections::HashSet;
 use std::fs;
 use std::path::PathBuf;
 
@@ -97,12 +98,13 @@ pub(crate) fn main(args: &Args) -> Result<()> {
 fn process_documentation(documentation: &str, out: &mut String, rule_name: &str) {
     let mut in_options = false;
     let mut after = String::new();
+    let mut referenced_options = HashSet::new();
 
     // HACK: This is an ugly regex hack that's necessary because mkdocs uses
     // a non-CommonMark-compliant Markdown parser, which doesn't support code
     // tags in link definitions
     // (see https://github.com/Python-Markdown/markdown/issues/280).
-    let documentation = Regex::new(r"\[`([^`]*?)`]($|[^\[])").unwrap().replace_all(
+    let documentation = Regex::new(r"\[`([^`]*?)`]($|[^\[(])").unwrap().replace_all(
         documentation,
         |caps: &Captures| {
             format!(
@@ -135,6 +137,7 @@ fn process_documentation(documentation: &str, out: &mut String, rule_name: &str)
                 let anchor = option.replace('.', "_");
                 out.push_str(&format!("- [`{option}`][{option}]\n"));
                 after.push_str(&format!("[{option}]: ../settings.md#{anchor}\n"));
+                referenced_options.insert(option);
 
                 continue;
             }
@@ -142,6 +145,20 @@ fn process_documentation(documentation: &str, out: &mut String, rule_name: &str)
 
         out.push_str(line);
     }
+
+    let re = Regex::new(r"\[`([^`]*?)`]\[(.*?)]").unwrap();
+    for (_, [option, _]) in re.captures_iter(&documentation).map(|c| c.extract()) {
+        if let Some(OptionEntry::Field(field)) = Options::metadata().find(option) {
+            if referenced_options.insert(option) {
+                let anchor = option.replace('.', "_");
+                after.push_str(&format!("[{option}]: ../settings.md#{anchor}\n"));
+            }
+            if field.deprecated.is_some() {
+                eprintln!("Rule {rule_name} references deprecated option {option}.");
+            }
+        }
+    }
+
     if !after.is_empty() {
         out.push('\n');
         out.push('\n');
@@ -159,7 +176,7 @@ mod tests {
         process_documentation(
             "
 See also [`lint.mccabe.max-complexity`] and [`lint.task-tags`].
-Something [`else`][other].
+Something [`else`][other]. Some [link](https://example.com).
 
 ## Options
 
@@ -174,7 +191,7 @@ Something [`else`][other].
             output,
             "
 See also [`lint.mccabe.max-complexity`][lint.mccabe.max-complexity] and [`lint.task-tags`][lint.task-tags].
-Something [`else`][other].
+Something [`else`][other]. Some [link](https://example.com).
 
 ## Options
 

crates/ruff_dev/src/generate_rules_table.rs

@@ -180,8 +180,22 @@ pub(crate) fn generate() -> String {
             .map(|rule| (rule.upstream_category(&linter), rule))
             .into_group_map();
 
+        let mut rules_by_upstream_category: Vec<_> = rules_by_upstream_category.iter().collect();
+
+        // Sort the upstream categories alphabetically by prefix.
+        rules_by_upstream_category.sort_by(|(a, _), (b, _)| {
+            a.as_ref()
+                .map(|category| category.prefix)
+                .unwrap_or_default()
+                .cmp(
+                    b.as_ref()
+                        .map(|category| category.prefix)
+                        .unwrap_or_default(),
+                )
+        });
+
         if rules_by_upstream_category.len() > 1 {
-            for (opt, rules) in &rules_by_upstream_category {
+            for (opt, rules) in rules_by_upstream_category {
                 if opt.is_some() {
                     let UpstreamCategoryAndPrefix { category, prefix } = opt.unwrap();
                     table_out.push_str(&format!("#### {category} ({prefix})"));

crates/ruff_dev/src/main.rs

@@ -4,8 +4,8 @@
 
 use anyhow::Result;
 use clap::{Parser, Subcommand};
-use ruff::check;
-use ruff_linter::logging::{set_up_logging, LogLevel};
+use ruff::{args::GlobalConfigArgs, check};
+use ruff_linter::logging::set_up_logging;
 use std::process::ExitCode;
 
 mod format_dev;
@@ -28,6 +28,8 @@ const ROOT_DIR: &str = concat!(env!("CARGO_MANIFEST_DIR"), "/../../");
 struct Args {
     #[command(subcommand)]
     command: Command,
+    #[clap(flatten)]
+    global_options: GlobalConfigArgs,
 }
 
 #[derive(Subcommand)]
@@ -57,8 +59,6 @@ enum Command {
     Repeat {
         #[clap(flatten)]
         args: ruff::args::CheckCommand,
-        #[clap(flatten)]
-        log_level_args: ruff::args::LogLevelArgs,
         /// Run this many times
         #[clap(long)]
         repeat: usize,
@@ -75,9 +75,12 @@ enum Command {
 }
 
 fn main() -> Result<ExitCode> {
-    let args = Args::parse();
+    let Args {
+        command,
+        global_options,
+    } = Args::parse();
     #[allow(clippy::print_stdout)]
-    match args.command {
+    match command {
         Command::GenerateAll(args) => generate_all::main(&args)?,
         Command::GenerateJSONSchema(args) => generate_json_schema::main(&args)?,
         Command::GenerateRulesTable => println!("{}", generate_rules_table::generate()),
@@ -89,14 +92,12 @@ fn main() -> Result<ExitCode> {
         Command::PrintTokens(args) => print_tokens::main(&args)?,
         Command::RoundTrip(args) => round_trip::main(&args)?,
         Command::Repeat {
-            args,
+            args: subcommand_args,
             repeat,
-            log_level_args,
         } => {
-            let log_level = LogLevel::from(&log_level_args);
-            set_up_logging(&log_level)?;
+            set_up_logging(global_options.log_level())?;
             for _ in 0..repeat {
-                check(args.clone(), log_level)?;
+                check(subcommand_args.clone(), global_options.clone())?;
             }
         }
         Command::FormatDev(args) => {

crates/ruff_diagnostics/Cargo.toml

@@ -11,6 +11,7 @@ repository = { workspace = true }
 license = { workspace = true }
 
 [lib]
+doctest = false
 
 [dependencies]
 ruff_text_size = { path = "../ruff_text_size" }

crates/ruff_diagnostics/src/diagnostic.rs

@@ -71,6 +71,14 @@ impl Diagnostic {
         }
     }
 
+    /// Consumes `self` and returns a new `Diagnostic` with the given parent node.
+    #[inline]
+    #[must_use]
+    pub fn with_parent(mut self, parent: TextSize) -> Self {
+        self.set_parent(parent);
+        self
+    }
+
     /// Set the location of the diagnostic's parent node.
     #[inline]
     pub fn set_parent(&mut self, parent: TextSize) {

crates/ruff_diagnostics/src/violation.rs

@@ -25,11 +25,6 @@ pub trait Violation: Debug + PartialEq + Eq {
     /// The message used to describe the violation.
     fn message(&self) -> String;
 
-    /// The explanation used in documentation and elsewhere.
-    fn explanation() -> Option<&'static str> {
-        None
-    }
-
     // TODO(micha): Move `fix_title` to `Fix`, add new `advice` method that is shown as an advice.
     // Change the `Diagnostic` renderer to show the advice, and render the fix message after the `Suggested fix: <here>`
 
@@ -50,11 +45,6 @@ pub trait AlwaysFixableViolation: Debug + PartialEq + Eq {
     /// The message used to describe the violation.
     fn message(&self) -> String;
 
-    /// The explanation used in documentation and elsewhere.
-    fn explanation() -> Option<&'static str> {
-        None
-    }
-
     /// The title displayed for the available fix.
     fn fix_title(&self) -> String;
 
@@ -71,10 +61,6 @@ impl<V: AlwaysFixableViolation> Violation for V {
         <Self as AlwaysFixableViolation>::message(self)
     }
 
-    fn explanation() -> Option<&'static str> {
-        <Self as AlwaysFixableViolation>::explanation()
-    }
-
     fn fix_title(&self) -> Option<String> {
         Some(<Self as AlwaysFixableViolation>::fix_title(self))
     }

crates/ruff_formatter/Cargo.toml

@@ -24,7 +24,6 @@ tracing = { workspace = true }
 unicode-width = { workspace = true }
 
 [dev-dependencies]
-insta = { workspace = true }
 
 [features]
 serde = ["dep:serde", "ruff_text_size/serde"]

crates/ruff_formatter/src/buffer.rs

@@ -37,7 +37,7 @@ pub trait Buffer {
     #[doc(hidden)]
     fn elements(&self) -> &[FormatElement];
 
-    /// Glue for usage of the [`write!`] macro with implementors of this trait.
+    /// Glue for usage of the [`write!`] macro with implementers of this trait.
     ///
     /// This method should generally not be invoked manually, but rather through the [`write!`] macro itself.
     ///

crates/ruff_formatter/src/builders.rs

@@ -138,7 +138,7 @@ pub const fn empty_line() -> Line {
 ///
 /// # Examples
 ///
-/// The line breaks are emitted as spaces if the enclosing `Group` fits on a a single line:
+/// The line breaks are emitted as spaces if the enclosing `Group` fits on a single line:
 /// ```
 /// use ruff_formatter::{format, format_args};
 /// use ruff_formatter::prelude::*;

crates/ruff_formatter/src/lib.rs

@@ -545,6 +545,10 @@ impl PrintedRange {
         &self.code
     }
 
+    pub fn into_code(self) -> String {
+        self.code
+    }
+
     /// The range the formatted code corresponds to in the source document.
     pub fn source_range(&self) -> TextRange {
         self.source_range

crates/ruff_formatter/src/printer/line_suffixes.rs

@@ -21,8 +21,7 @@ impl<'a> LineSuffixes<'a> {
     /// Takes all the pending line suffixes.
     pub(super) fn take_pending<'l>(
         &'l mut self,
-    ) -> impl Iterator<Item = LineSuffixEntry<'a>> + DoubleEndedIterator + 'l + ExactSizeIterator
-    {
+    ) -> impl DoubleEndedIterator<Item = LineSuffixEntry<'a>> + 'l + ExactSizeIterator {
         self.suffixes.drain(..)
     }
 

crates/ruff_formatter/src/printer/queue.rs

@@ -78,27 +78,28 @@ impl<'a> PrintQueue<'a> {
 impl<'a> Queue<'a> for PrintQueue<'a> {
     fn pop(&mut self) -> Option<&'a FormatElement> {
         let elements = self.element_slices.last_mut()?;
-        elements.next().or_else(|| {
-            self.element_slices.pop();
-            let elements = self.element_slices.last_mut()?;
-            elements.next()
-        })
+        elements.next().or_else(
+            #[cold]
+            || {
+                self.element_slices.pop();
+                let elements = self.element_slices.last_mut()?;
+                elements.next()
+            },
+        )
     }
 
     fn top_with_interned(&self) -> Option<&'a FormatElement> {
         let mut slices = self.element_slices.iter().rev();
         let slice = slices.next()?;
 
-        match slice.as_slice().first() {
-            Some(element) => Some(element),
-            None => {
-                if let Some(next_elements) = slices.next() {
-                    next_elements.as_slice().first()
-                } else {
-                    None
-                }
-            }
-        }
+        slice.as_slice().first().or_else(
+            #[cold]
+            || {
+                slices
+                    .next()
+                    .and_then(|next_elements| next_elements.as_slice().first())
+            },
+        )
     }
 
     fn extend_back(&mut self, elements: &'a [FormatElement]) {
@@ -146,24 +147,30 @@ impl<'a, 'print> FitsQueue<'a, 'print> {
 
 impl<'a, 'print> Queue<'a> for FitsQueue<'a, 'print> {
     fn pop(&mut self) -> Option<&'a FormatElement> {
-        self.queue.pop().or_else(|| {
-            if let Some(next_slice) = self.rest_elements.next_back() {
-                self.queue.extend_back(next_slice.as_slice());
-                self.queue.pop()
-            } else {
-                None
-            }
-        })
+        self.queue.pop().or_else(
+            #[cold]
+            || {
+                if let Some(next_slice) = self.rest_elements.next_back() {
+                    self.queue.extend_back(next_slice.as_slice());
+                    self.queue.pop()
+                } else {
+                    None
+                }
+            },
+        )
     }
 
     fn top_with_interned(&self) -> Option<&'a FormatElement> {
-        self.queue.top_with_interned().or_else(|| {
-            if let Some(next_elements) = self.rest_elements.as_slice().last() {
-                next_elements.as_slice().first()
-            } else {
-                None
-            }
-        })
+        self.queue.top_with_interned().or_else(
+            #[cold]
+            || {
+                if let Some(next_elements) = self.rest_elements.as_slice().last() {
+                    next_elements.as_slice().first()
+                } else {
+                    None
+                }
+            },
+        )
     }
 
     fn extend_back(&mut self, elements: &'a [FormatElement]) {

crates/ruff_index/Cargo.toml

@@ -11,6 +11,7 @@ repository = { workspace = true }
 license = { workspace = true }
 
 [lib]
+doctest = false
 
 [dependencies]
 ruff_macros = { path = "../ruff_macros" }

crates/ruff_linter/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ruff_linter"
-version = "0.2.1"
+version = "0.4.2"
 publish = false
 authors = { workspace = true }
 edition = { workspace = true }
@@ -15,7 +15,6 @@ license = { workspace = true }
 [dependencies]
 ruff_cache = { path = "../ruff_cache" }
 ruff_diagnostics = { path = "../ruff_diagnostics", features = ["serde"] }
-ruff_index = { path = "../ruff_index" }
 ruff_notebook = { path = "../ruff_notebook" }
 ruff_macros = { path = "../ruff_macros" }
 ruff_python_ast = { path = "../ruff_python_ast", features = ["serde"] }
@@ -60,7 +59,6 @@ regex = { workspace = true }
 result-like = { workspace = true }
 rustc-hash = { workspace = true }
 schemars = { workspace = true, optional = true }
-semver = { workspace = true }
 serde = { workspace = true }
 serde_json = { workspace = true }
 similar = { workspace = true }
@@ -76,11 +74,9 @@ url = { workspace = true }
 
 [dev-dependencies]
 insta = { workspace = true }
-pretty_assertions = { workspace = true }
 test-case = { workspace = true }
 # Disable colored output in tests
 colored = { workspace = true, features = ["no-color"] }
-tempfile = { workspace = true }
 
 [features]
 default = []

crates/ruff_linter/resources/test/fixtures/eradicate/ERA001.py

@@ -28,3 +28,40 @@ dictionary = {
 }
 
 #import os  # noqa
+
+# case 1:
+# try:
+# try:  # with comment
+# try: print()
+# except:
+# except Foo:
+# except Exception as e: print(e)
+
+
+# Script tag without an opening tag (Error)
+
+# requires-python = ">=3.11"
+# dependencies = [
+#   "requests<3",
+#   "rich",
+# ]
+# ///
+
+# Script tag (OK)
+
+# /// script
+# requires-python = ">=3.11"
+# dependencies = [
+#   "requests<3",
+#   "rich",
+# ]
+# ///
+
+# Script tag without a closing tag (OK)
+
+# /// script
+# requires-python = ">=3.11"
+# dependencies = [
+#   "requests<3",
+#   "rich",
+# ]

crates/ruff_linter/resources/test/fixtures/flake8_bandit/S104.py

@@ -18,3 +18,7 @@ func("0.0.0.0")
 def my_func():
     x = "0.0.0.0"
     print(x)
+
+
+# Implicit string concatenation
+"0.0.0.0" f"0.0.0.0{expr}0.0.0.0"

crates/ruff_linter/resources/test/fixtures/flake8_bandit/S108.py

@@ -18,6 +18,13 @@ with open("/dev/shm/unit/test", "w") as f:
 with open("/foo/bar", "w") as f:
     f.write("def")
 
+# Implicit string concatenation
+with open("/tmp/" "abc", "w") as f:
+    f.write("def")
+
+with open("/tmp/abc" f"/tmp/abc", "w") as f:
+    f.write("def")
+
 # Using `tempfile` module should be ok
 import tempfile
 from tempfile import TemporaryDirectory

crates/ruff_linter/resources/test/fixtures/flake8_bandit/S308.py

@@ -0,0 +1,22 @@
+from django.utils.safestring import mark_safe
+
+
+def some_func():
+    return mark_safe('<script>alert("evil!")</script>')
+
+
+@mark_safe
+def some_func():
+    return '<script>alert("evil!")</script>'
+
+
+from django.utils.html import mark_safe
+
+
+def some_func():
+    return mark_safe('<script>alert("evil!")</script>')
+
+
+@mark_safe
+def some_func():
+    return '<script>alert("evil!")</script>'

crates/ruff_linter/resources/test/fixtures/flake8_bandit/S310.py

@@ -17,3 +17,9 @@ urllib.request.URLopener().open(fullurl='http://www.google.com')
 urllib.request.URLopener().open('http://www.google.com')
 urllib.request.URLopener().open('file:///foo/bar/baz')
 urllib.request.URLopener().open(url)
+
+urllib.request.urlopen(url=urllib.request.Request('http://www.google.com'))
+urllib.request.urlopen(url=urllib.request.Request('http://www.google.com'), **kwargs)
+urllib.request.urlopen(urllib.request.Request('http://www.google.com'))
+urllib.request.urlopen(urllib.request.Request('file:///foo/bar/baz'))
+urllib.request.urlopen(urllib.request.Request(url))

crates/ruff_linter/resources/test/fixtures/flake8_bandit/S311.py

@@ -0,0 +1,22 @@
+import os
+import random
+
+import a_lib
+
+# OK
+random.SystemRandom()
+
+# Errors
+random.Random()
+random.random()
+random.randrange()
+random.randint()
+random.choice()
+random.choices()
+random.uniform()
+random.triangular()
+random.randbytes()
+
+# Unrelated
+os.urandom()
+a_lib.random()

crates/ruff_linter/resources/test/fixtures/flake8_bandit/S324.py

@@ -1,52 +1,47 @@
+import crypt
 import hashlib
 from hashlib import new as hashlib_new
 from hashlib import sha1 as hashlib_sha1
 
-# Invalid
-
+# Errors
 hashlib.new('md5')
-
 hashlib.new('md4', b'test')
-
 hashlib.new(name='md5', data=b'test')
-
 hashlib.new('MD4', data=b'test')
-
 hashlib.new('sha1')
-
 hashlib.new('sha1', data=b'test')
-
 hashlib.new('sha', data=b'test')
-
 hashlib.new(name='SHA', data=b'test')
-
 hashlib.sha(data=b'test')
-
 hashlib.md5()
-
 hashlib_new('sha1')
-
 hashlib_sha1('sha1')
-
 # usedforsecurity arg only available in Python 3.9+
 hashlib.new('sha1', usedforsecurity=True)
 
-# Valid
+crypt.crypt("test", salt=crypt.METHOD_CRYPT)
+crypt.crypt("test", salt=crypt.METHOD_MD5)
+crypt.crypt("test", salt=crypt.METHOD_BLOWFISH)
+crypt.crypt("test", crypt.METHOD_BLOWFISH)
 
+crypt.mksalt(crypt.METHOD_CRYPT)
+crypt.mksalt(crypt.METHOD_MD5)
+crypt.mksalt(crypt.METHOD_BLOWFISH)
+
+# OK
 hashlib.new('sha256')
-
 hashlib.new('SHA512')
-
 hashlib.sha256(data=b'test')
-
 # usedforsecurity arg only available in Python 3.9+
 hashlib_new(name='sha1', usedforsecurity=False)
-
-# usedforsecurity arg only available in Python 3.9+
 hashlib_sha1(name='sha1', usedforsecurity=False)
-
-# usedforsecurity arg only available in Python 3.9+
 hashlib.md4(usedforsecurity=False)
-
-# usedforsecurity arg only available in Python 3.9+
 hashlib.new(name='sha256', usedforsecurity=False)
+
+crypt.crypt("test")
+crypt.crypt("test", salt=crypt.METHOD_SHA256)
+crypt.crypt("test", salt=crypt.METHOD_SHA512)
+
+crypt.mksalt()
+crypt.mksalt(crypt.METHOD_SHA256)
+crypt.mksalt(crypt.METHOD_SHA512)

crates/ruff_linter/resources/test/fixtures/flake8_bandit/S605.py

@@ -1,4 +1,5 @@
 import os
+import subprocess
 
 import commands
 import popen2
@@ -16,6 +17,8 @@ popen2.Popen3("true")
 popen2.Popen4("true")
 commands.getoutput("true")
 commands.getstatusoutput("true")
+subprocess.getoutput("true")
+subprocess.getstatusoutput("true")
 
 
 # Check command argument looks unsafe.