Minor tweaks

## Summary

Another bug found using [property
testing](https://github.com/astral-sh/ruff/pull/14178).

## Test Plan

New unit test

.github/workflows/ci.yaml

@@ -16,7 +16,7 @@ env:
   CARGO_TERM_COLOR: always
   RUSTUP_MAX_RETRIES: 10
   PACKAGE_NAME: ruff
-  PYTHON_VERSION: "3.11"
+  PYTHON_VERSION: "3.12"
 
 jobs:
   determine_changes:

.github/workflows/publish-playground.yml

@@ -47,7 +47,7 @@ jobs:
         working-directory: playground
       - name: "Deploy to Cloudflare Pages"
         if: ${{ env.CF_API_TOKEN_EXISTS == 'true' }}
-        uses: cloudflare/wrangler-action@v3.9.0
+        uses: cloudflare/wrangler-action@v3.11.0
         with:
           apiToken: ${{ secrets.CF_API_TOKEN }}
           accountId: ${{ secrets.CF_ACCOUNT_ID }}

.pre-commit-config.yaml

@@ -17,7 +17,7 @@ exclude: |
 
 repos:
   - repo: https://github.com/abravalheri/validate-pyproject
-    rev: v0.21
+    rev: v0.22
     hooks:
       - id: validate-pyproject
 
@@ -51,11 +51,15 @@ repos:
       - id: blacken-docs
         args: ["--pyi", "--line-length", "130"]
         files: '^crates/.*/resources/mdtest/.*\.md'
+        exclude: |
+          (?x)^(
+            .*?invalid(_.+)_syntax.md
+          )$
         additional_dependencies:
           - black==24.10.0
 
   - repo: https://github.com/crate-ci/typos
-    rev: v1.26.0
+    rev: v1.27.0
     hooks:
       - id: typos
 
@@ -69,7 +73,7 @@ repos:
         pass_filenames: false # This makes it a lot faster
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.7.0
+    rev: v0.7.2
     hooks:
       - id: ruff-format
       - id: ruff

CHANGELOG.md

@@ -1,5 +1,42 @@
 # Changelog
 
+## 0.7.3
+
+### Preview features
+
+- Formatter: Disallow single-line implicit concatenated strings ([#13928](https://github.com/astral-sh/ruff/pull/13928))
+- \[`flake8-pyi`\] Include all Python file types for `PYI006` and `PYI066` ([#14059](https://github.com/astral-sh/ruff/pull/14059))
+- \[`flake8-simplify`\] Implement `split-of-static-string` (`SIM905`) ([#14008](https://github.com/astral-sh/ruff/pull/14008))
+- \[`refurb`\] Implement `subclass-builtin` (`FURB189`) ([#14105](https://github.com/astral-sh/ruff/pull/14105))
+- \[`ruff`\] Improve diagnostic messages and docs (`RUF031`, `RUF032`, `RUF034`) ([#14068](https://github.com/astral-sh/ruff/pull/14068))
+
+### Rule changes
+
+- Detect items that hash to same value in duplicate sets (`B033`, `PLC0208`) ([#14064](https://github.com/astral-sh/ruff/pull/14064))
+- \[`eradicate`\] Better detection of IntelliJ language injection comments (`ERA001`) ([#14094](https://github.com/astral-sh/ruff/pull/14094))
+- \[`flake8-pyi`\] Add autofix for `docstring-in-stub` (`PYI021`) ([#14150](https://github.com/astral-sh/ruff/pull/14150))
+- \[`flake8-pyi`\] Update `duplicate-literal-member` (`PYI062`) to alawys provide an autofix ([#14188](https://github.com/astral-sh/ruff/pull/14188))
+- \[`pyflakes`\] Detect items that hash to same value in duplicate dictionaries (`F601`) ([#14065](https://github.com/astral-sh/ruff/pull/14065))
+- \[`ruff`\] Fix false positive for decorators (`RUF028`) ([#14061](https://github.com/astral-sh/ruff/pull/14061))
+
+### Bug fixes
+
+- Avoid parsing joint rule codes as distinct codes in `# noqa` ([#12809](https://github.com/astral-sh/ruff/pull/12809))
+- \[`eradicate`\] ignore `# language=` in commented-out-code rule (ERA001) ([#14069](https://github.com/astral-sh/ruff/pull/14069))
+- \[`flake8-bugbear`\] - do not run `mutable-argument-default` on stubs (`B006`) ([#14058](https://github.com/astral-sh/ruff/pull/14058))
+- \[`flake8-builtins`\] Skip lambda expressions in `builtin-argument-shadowing (A002)` ([#14144](https://github.com/astral-sh/ruff/pull/14144))
+- \[`flake8-comprehension`\] Also remove trailing comma while fixing `C409` and `C419` ([#14097](https://github.com/astral-sh/ruff/pull/14097))
+- \[`flake8-simplify`\] Allow `open` without context manager in `return` statement (`SIM115`) ([#14066](https://github.com/astral-sh/ruff/pull/14066))
+- \[`pylint`\] Respect hash-equivalent literals in `iteration-over-set` (`PLC0208`) ([#14063](https://github.com/astral-sh/ruff/pull/14063))
+- \[`pylint`\] Update known dunder methods for Python 3.13 (`PLW3201`) ([#14146](https://github.com/astral-sh/ruff/pull/14146))
+- \[`pyupgrade`\] - ignore kwarg unpacking for `UP044` ([#14053](https://github.com/astral-sh/ruff/pull/14053))
+- \[`refurb`\] Parse more exotic decimal strings in `verbose-decimal-constructor` (`FURB157`) ([#14098](https://github.com/astral-sh/ruff/pull/14098))
+
+### Documentation
+
+- Add links to missing related options within rule documentations ([#13971](https://github.com/astral-sh/ruff/pull/13971))
+- Add rule short code to mkdocs tags to allow searching via rule codes ([#14040](https://github.com/astral-sh/ruff/pull/14040))
+
 ## 0.7.2
 
 ### Preview features

Cargo.lock

@@ -123,9 +123,9 @@ dependencies = [
 
 [[package]]
 name = "anyhow"
-version = "1.0.91"
+version = "1.0.92"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c042108f3ed77fd83760a5fd79b53be043192bb3b9dba91d8c574c0ada7850c8"
+checksum = "74f37166d7d48a0284b99dd824694c26119c700b53bf0d1540cdb147dbdaaf13"
 
 [[package]]
 name = "append-only-vec"
@@ -407,7 +407,7 @@ dependencies = [
  "heck",
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -687,7 +687,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "strsim 0.10.0",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -698,7 +698,7 @@ checksum = "a668eda54683121533a393014d8692171709ff57a7d61f187b6e782719f8933f"
 dependencies = [
  "darling_core",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1162,12 +1162,12 @@ dependencies = [
 
 [[package]]
 name = "indexmap"
-version = "2.5.0"
+version = "2.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "68b900aa2f7301e21c36462b170ee99994de34dff39a4a6a528e80e7376d07e5"
+checksum = "707907fe3c25f5424cce2cb7e1cbcafee6bdbe735ca90ef77c29e84591e5b9da"
 dependencies = [
  "equivalent",
- "hashbrown 0.14.5",
+ "hashbrown 0.15.0",
  "serde",
 ]
 
@@ -1193,9 +1193,9 @@ checksum = "b248f5224d1d606005e02c97f5aa4e88eeb230488bcc03bc9ca4d7991399f2b5"
 
 [[package]]
 name = "inotify"
-version = "0.9.6"
+version = "0.10.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f8069d3ec154eb856955c1c0fbffefbf5f3c40a104ec912d4797314c1801abff"
+checksum = "fdd168d97690d0b8c412d6b6c10360277f4d7ee495c5d0d5d5fe0854923255cc"
 dependencies = [
  "bitflags 1.3.2",
  "inotify-sys",
@@ -1213,9 +1213,9 @@ dependencies = [
 
 [[package]]
 name = "insta"
-version = "1.41.0"
+version = "1.41.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a1f72d3e19488cf7d8ea52d2fc0f8754fc933398b337cd3cbdb28aaeb35159ef"
+checksum = "7e9ffc4d4892617c50a928c52b2961cb5174b6fc6ebf252b2fac9d21955c48b8"
 dependencies = [
  "console",
  "globset",
@@ -1267,7 +1267,7 @@ dependencies = [
  "Inflector",
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1393,7 +1393,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a2ae40017ac09cd2c6a53504cb3c871c7f2b41466eac5bc66ba63f39073b467b"
 dependencies = [
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -1532,14 +1532,15 @@ dependencies = [
 
 [[package]]
 name = "mio"
-version = "0.8.11"
+version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a4a650543ca06a924e8b371db273b2756685faae30f8487da1b56505a8f78b0c"
+checksum = "80e04d1dcff3aae0704555fe5fee3bcfaf3d1fdf8a7e521d5b9d2b42acb52cec"
 dependencies = [
+ "hermit-abi",
  "libc",
  "log",
  "wasi",
- "windows-sys 0.48.0",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -1593,12 +1594,11 @@ dependencies = [
 
 [[package]]
 name = "notify"
-version = "6.1.1"
+version = "7.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6205bd8bb1e454ad2e27422015fb5e4f2bcc7e08fa8f27058670d208324a4d2d"
+checksum = "c533b4c39709f9ba5005d8002048266593c1cfaf3c5f0739d5b8ab0c6c504009"
 dependencies = [
  "bitflags 2.6.0",
- "crossbeam-channel",
  "filetime",
  "fsevent-sys",
  "inotify",
@@ -1606,8 +1606,18 @@ dependencies = [
  "libc",
  "log",
  "mio",
+ "notify-types",
  "walkdir",
- "windows-sys 0.48.0",
+ "windows-sys 0.52.0",
+]
+
+[[package]]
+name = "notify-types"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7393c226621f817964ffb3dc5704f9509e107a8b024b489cc2c1b217378785df"
+dependencies = [
+ "instant",
 ]
 
 [[package]]
@@ -1786,9 +1796,9 @@ dependencies = [
 
 [[package]]
 name = "pep440_rs"
-version = "0.7.1"
+version = "0.7.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7c8ee724d21f351f9d47276614ac9710975db827ba9fe2ca5a517ba648193307"
+checksum = "0922a442c78611fa8c5ed6065d2d898a820cf12fa90604217fdb2d01675efec7"
 dependencies = [
  "serde",
  "unicode-width 0.2.0",
@@ -1848,7 +1858,7 @@ dependencies = [
  "pest_meta",
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -2102,6 +2112,7 @@ dependencies = [
  "countme",
  "dir-test",
  "hashbrown 0.15.0",
+ "indexmap",
  "insta",
  "itertools 0.13.0",
  "memchr",
@@ -2306,7 +2317,7 @@ dependencies = [
 
 [[package]]
 name = "ruff"
-version = "0.7.2"
+version = "0.7.3"
 dependencies = [
  "anyhow",
  "argfile",
@@ -2523,7 +2534,7 @@ dependencies = [
 
 [[package]]
 name = "ruff_linter"
-version = "0.7.2"
+version = "0.7.3"
 dependencies = [
  "aho-corasick",
  "annotate-snippets 0.9.2",
@@ -2546,7 +2557,7 @@ dependencies = [
  "natord",
  "path-absolutize",
  "pathdiff",
- "pep440_rs 0.7.1",
+ "pep440_rs 0.7.2",
  "pyproject-toml",
  "quick-junit",
  "regex",
@@ -2590,7 +2601,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "ruff_python_trivia",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -2627,6 +2638,7 @@ dependencies = [
  "ruff_source_file",
  "ruff_text_size",
  "rustc-hash 2.0.0",
+ "salsa",
  "schemars",
  "serde",
 ]
@@ -2837,7 +2849,7 @@ dependencies = [
 
 [[package]]
 name = "ruff_wasm"
-version = "0.7.2"
+version = "0.7.3"
 dependencies = [
  "console_error_panic_hook",
  "console_log",
@@ -2876,7 +2888,7 @@ dependencies = [
  "matchit",
  "path-absolutize",
  "path-slash",
- "pep440_rs 0.7.1",
+ "pep440_rs 0.7.2",
  "regex",
  "ruff_cache",
  "ruff_formatter",
@@ -3008,7 +3020,7 @@ dependencies = [
  "heck",
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
  "synstructure",
 ]
 
@@ -3042,7 +3054,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "serde_derive_internals",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3065,9 +3077,9 @@ checksum = "1c107b6f4780854c8b126e228ea8869f4d7b71260f962fefb57b996b8959ba6b"
 
 [[package]]
 name = "serde"
-version = "1.0.213"
+version = "1.0.214"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3ea7893ff5e2466df8d720bb615088341b295f849602c6956047f8f80f0e9bc1"
+checksum = "f55c3193aca71c12ad7890f1785d2b73e1b9f63a0bbc353c08ef26fe03fc56b5"
 dependencies = [
  "serde_derive",
 ]
@@ -3085,13 +3097,13 @@ dependencies = [
 
 [[package]]
 name = "serde_derive"
-version = "1.0.213"
+version = "1.0.214"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7e85ad2009c50b58e87caa8cd6dac16bdf511bbfb7af6c33df902396aa480fa5"
+checksum = "de523f781f095e28fa605cdce0f8307e451cc0fd14e2eb4cd2e98a355b147766"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3102,7 +3114,7 @@ checksum = "330f01ce65a3a5fe59a60c82f3c9a024b573b8a6e875bd233fe5f934e71d54e3"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3125,7 +3137,7 @@ checksum = "6c64451ba24fc7a6a2d60fc75dd9c83c90903b19028d4eff35e88fc1e86564e9"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3166,7 +3178,7 @@ dependencies = [
  "darling",
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3268,7 +3280,7 @@ dependencies = [
  "proc-macro2",
  "quote",
  "rustversion",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3290,9 +3302,9 @@ dependencies = [
 
 [[package]]
 name = "syn"
-version = "2.0.85"
+version = "2.0.87"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5023162dfcd14ef8f32034d8bcd4cc5ddc61ef7a247c024a33e24e1f24d21b56"
+checksum = "25aa4ce346d03a6dcd68dd8b4010bcb74e54e62c90c573f394c46eae99aba32d"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -3307,7 +3319,7 @@ checksum = "c8af7666ab7b6390ab78131fb5b0fce11d6b7a6951602017c35fa82800708971"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3370,7 +3382,7 @@ dependencies = [
  "cfg-if",
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3381,28 +3393,28 @@ checksum = "5c89e72a01ed4c579669add59014b9a524d609c0c88c6a585ce37485879f6ffb"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
  "test-case-core",
 ]
 
 [[package]]
 name = "thiserror"
-version = "1.0.65"
+version = "1.0.67"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5d11abd9594d9b38965ef50805c5e469ca9cc6f197f883f717e0269a3057b3d5"
+checksum = "3b3c6efbfc763e64eb85c11c25320f0737cb7364c4b6336db90aa9ebe27a0bbd"
 dependencies = [
  "thiserror-impl",
 ]
 
 [[package]]
 name = "thiserror-impl"
-version = "1.0.65"
+version = "1.0.67"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ae71770322cbd277e69d762a16c444af02aa0575ac0d174f0b9562d3b37f8602"
+checksum = "b607164372e89797d78b8e23a6d67d5d1038c1c65efd52e1389ef8b77caba2a6"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3514,7 +3526,7 @@ checksum = "34704c8d6ebcbc939824180af020566b01a7c01f80641264eba0999f6c2b6be7"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3772,7 +3784,7 @@ checksum = "6b91f57fe13a38d0ce9e28a03463d8d3c2468ed03d75375110ec71d93b449a08"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -3858,7 +3870,7 @@ dependencies = [
  "once_cell",
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
  "wasm-bindgen-shared",
 ]
 
@@ -3892,7 +3904,7 @@ checksum = "26c6ab57572f7a24a4985830b120de1594465e5d500f24afe89e16b4e833ef68"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
  "wasm-bindgen-backend",
  "wasm-bindgen-shared",
 ]
@@ -3926,7 +3938,7 @@ checksum = "c97b2ef2c8d627381e51c071c2ab328eac606d3f69dd82bcbca20a9e389d95f0"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]
@@ -4214,7 +4226,7 @@ checksum = "9ce1b18ccd8e73a9321186f97e46f9f04b778851177567b1975109d26a08d2a6"
 dependencies = [
  "proc-macro2",
  "quote",
- "syn 2.0.85",
+ "syn 2.0.87",
 ]
 
 [[package]]

Cargo.toml

@@ -81,6 +81,7 @@ hashbrown = { version = "0.15.0", default-features = false, features = [
 ignore = { version = "0.4.22" }
 imara-diff = { version = "0.1.5" }
 imperative = { version = "1.0.4" }
+indexmap = {version = "2.6.0" }
 indicatif = { version = "0.17.8" }
 indoc = { version = "2.0.4" }
 insta = { version = "1.35.1" }
@@ -101,7 +102,7 @@ matchit = { version = "0.8.1" }
 memchr = { version = "2.7.1" }
 mimalloc = { version = "0.1.39" }
 natord = { version = "1.0.9" }
-notify = { version = "6.1.1" }
+notify = { version = "7.0.0" }
 ordermap = { version = "0.5.0" }
 path-absolutize = { version = "3.1.1" }
 path-slash = { version = "0.2.1" }

README.md

@@ -136,8 +136,8 @@ curl -LsSf https://astral.sh/ruff/install.sh | sh
 powershell -c "irm https://astral.sh/ruff/install.ps1 | iex"
 
 # For a specific version.
-curl -LsSf https://astral.sh/ruff/0.7.2/install.sh | sh
-powershell -c "irm https://astral.sh/ruff/0.7.2/install.ps1 | iex"
+curl -LsSf https://astral.sh/ruff/0.7.3/install.sh | sh
+powershell -c "irm https://astral.sh/ruff/0.7.3/install.ps1 | iex"
 ```
 
 You can also install Ruff via [Homebrew](https://formulae.brew.sh/formula/ruff), [Conda](https://anaconda.org/conda-forge/ruff),
@@ -170,7 +170,7 @@ 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: v0.7.2
+  rev: v0.7.3
   hooks:
     # Run the linter.
     - id: ruff

crates/red_knot/src/main.rs

@@ -5,8 +5,6 @@ use anyhow::{anyhow, Context};
 use clap::Parser;
 use colored::Colorize;
 use crossbeam::channel as crossbeam_channel;
-use salsa::plumbing::ZalsaDatabase;
-
 use red_knot_python_semantic::SitePackages;
 use red_knot_server::run_server;
 use red_knot_workspace::db::RootDatabase;
@@ -14,7 +12,9 @@ use red_knot_workspace::watch;
 use red_knot_workspace::watch::WorkspaceWatcher;
 use red_knot_workspace::workspace::settings::Configuration;
 use red_knot_workspace::workspace::WorkspaceMetadata;
+use ruff_db::diagnostic::Diagnostic;
 use ruff_db::system::{OsSystem, System, SystemPath, SystemPathBuf};
+use salsa::plumbing::ZalsaDatabase;
 use target_version::TargetVersion;
 
 use crate::logging::{setup_tracing, Verbosity};
@@ -318,8 +318,9 @@ impl MainLoop {
                 } => {
                     let has_diagnostics = !result.is_empty();
                     if check_revision == revision {
+                        #[allow(clippy::print_stdout)]
                         for diagnostic in result {
-                            tracing::error!("{}", diagnostic);
+                            println!("{}", diagnostic.display(db));
                         }
                     } else {
                         tracing::debug!(
@@ -378,7 +379,10 @@ impl MainLoopCancellationToken {
 #[derive(Debug)]
 enum MainLoopMessage {
     CheckWorkspace,
-    CheckCompleted { result: Vec<String>, revision: u64 },
+    CheckCompleted {
+        result: Vec<Box<dyn Diagnostic>>,
+        revision: u64,
+    },
     ApplyChanges(Vec<watch::ChangeEvent>),
     Exit,
 }

crates/red_knot_python_semantic/Cargo.toml

@@ -13,7 +13,7 @@ license = { workspace = true }
 [dependencies]
 ruff_db = { workspace = true }
 ruff_index = { workspace = true }
-ruff_python_ast = { workspace = true }
+ruff_python_ast = { workspace = true, features = ["salsa"] }
 ruff_python_stdlib = { workspace = true }
 ruff_source_file = { workspace = true }
 ruff_text_size = { workspace = true }
@@ -24,7 +24,8 @@ bitflags = { workspace = true }
 camino = { workspace = true }
 compact_str = { workspace = true }
 countme = { workspace = true }
-itertools = { workspace = true}
+indexmap = { workspace = true }
+itertools = { workspace = true }
 ordermap = { workspace = true }
 salsa = { workspace = true }
 thiserror = { workspace = true }
@@ -43,10 +44,9 @@ red_knot_test = { workspace = true }
 red_knot_vendored = { workspace = true }
 
 anyhow = { workspace = true }
-dir-test = {workspace = true}
+dir-test = { workspace = true }
 insta = { workspace = true }
 tempfile = { workspace = true }
 
 [lints]
 workspace = true
-

crates/red_knot_python_semantic/resources/mdtest/annotations/starred_expressions.md

@@ -0,0 +1,18 @@
+# Starred expression annotations
+
+Type annotations for `*args` can be starred expressions themselves:
+
+```py
+from typing_extensions import TypeVarTuple
+
+Ts = TypeVarTuple("Ts")
+
+def append_int(*args: *Ts) -> tuple[*Ts, int]:
+    # TODO: should show some representation of the variadic generic type
+    reveal_type(args)  # revealed: @Todo
+
+    return (*args, 1)
+
+# TODO should be tuple[Literal[True], Literal["a"], int]
+reveal_type(append_int(True, "a"))  # revealed: @Todo
+```

crates/red_knot_python_semantic/resources/mdtest/annotations/string_annotations.md

@@ -0,0 +1,9 @@
+# String annotations
+
+```py
+def f() -> "int":
+    return 1
+
+# TODO: We do not support string annotations, but we should not panic if we encounter them
+reveal_type(f())  # revealed: @Todo
+```

crates/red_knot_python_semantic/resources/mdtest/assignment/augmented.md

@@ -85,7 +85,7 @@ f = Foo()
 # that `Foo.__iadd__` may be unbound as additional context.
 f += "Hello, world!"
 
-reveal_type(f)  # revealed: int
+reveal_type(f)  # revealed: int | Unknown
 ```
 
 ## Partially bound with `__add__`
@@ -104,8 +104,7 @@ class Foo:
 f = Foo()
 f += "Hello, world!"
 
-# TODO(charlie): This should be `int | str`, since `__iadd__` may be unbound.
-reveal_type(f)  # revealed: int
+reveal_type(f)  # revealed: int | str
 ```
 
 ## Partially bound target union
@@ -127,8 +126,7 @@ else:
     f = 42.0
 f += 12
 
-# TODO(charlie): This should be `str | int | float`
-reveal_type(f)  # revealed: @Todo
+reveal_type(f)  # revealed: int | str | float
 ```
 
 ## Target union
@@ -149,6 +147,36 @@ else:
     f = 42.0
 f += 12
 
-# TODO(charlie): This should be `str | float`.
-reveal_type(f)  # revealed: @Todo
+reveal_type(f)  # revealed: str | float
+```
+
+## Partially bound target union with `__add__`
+
+```py
+def bool_instance() -> bool:
+    return True
+
+flag = bool_instance()
+
+class Foo:
+    def __add__(self, other: int) -> str:
+        return "Hello, world!"
+    if bool_instance():
+        def __iadd__(self, other: int) -> int:
+            return 42
+
+class Bar:
+    def __add__(self, other: int) -> bytes:
+        return b"Hello, world!"
+
+    def __iadd__(self, other: int) -> float:
+        return 42.0
+
+if flag:
+    f = Foo()
+else:
+    f = Bar()
+f += 12
+
+reveal_type(f)  # revealed: int | str | float
 ```

crates/red_knot_python_semantic/resources/mdtest/attributes.md

@@ -18,3 +18,38 @@ else:
 
 reveal_type(C.x)  # revealed: Literal[1, 2]
 ```
+
+## Inherited attributes
+
+```py
+class A:
+    X = "foo"
+
+class B(A): ...
+class C(B): ...
+
+reveal_type(C.X)  # revealed: Literal["foo"]
+```
+
+## Inherited attributes (multiple inheritance)
+
+```py
+class O: ...
+
+class F(O):
+    X = 56
+
+class E(O):
+    X = 42
+
+class D(O): ...
+class C(D, F): ...
+class B(E, D): ...
+class A(B, C): ...
+
+# revealed: tuple[Literal[A], Literal[B], Literal[E], Literal[C], Literal[D], Literal[F], Literal[O], Literal[object]]
+reveal_type(A.__mro__)
+
+# `E` is earlier in the MRO than `F`, so we should use the type of `E.X`
+reveal_type(A.X)  # revealed: Literal[42]
+```

crates/red_knot_python_semantic/resources/mdtest/binary/instances.md

@@ -202,11 +202,7 @@ reveal_type(A() + B())  # revealed: MyString
 # N.B. Still a subtype of `A`, even though `A` does not appear directly in the class's `__bases__`
 class C(B): ...
 
-# TODO: we currently only understand direct subclasses as subtypes of the superclass.
-# We need to iterate through the full MRO rather than just the class's bases;
-# if we do, we'll understand `C` as a subtype of `A`, and correctly understand this as being
-# `MyString` rather than `str`
-reveal_type(A() + C())  # revealed: str
+reveal_type(A() + C())  # revealed: MyString
 ```
 
 ## Reflected precedence 2

crates/red_knot_python_semantic/resources/mdtest/call/callable_instance.md

@@ -18,3 +18,58 @@ class Unit: ...
 b = Unit()(3.0)  # error: "Object of type `Unit` is not callable"
 reveal_type(b)  # revealed: Unknown
 ```
+
+## Possibly unbound `__call__` method
+
+```py
+def flag() -> bool: ...
+
+class PossiblyNotCallable:
+    if flag():
+        def __call__(self) -> int: ...
+
+a = PossiblyNotCallable()
+result = a()  # error: "Object of type `PossiblyNotCallable` is not callable (possibly unbound `__call__` method)"
+reveal_type(result)  # revealed: int
+```
+
+## Possibly unbound callable
+
+```py
+def flag() -> bool: ...
+
+if flag():
+    class PossiblyUnbound:
+        def __call__(self) -> int: ...
+
+# error: [possibly-unresolved-reference]
+a = PossiblyUnbound()
+reveal_type(a())  # revealed: int
+```
+
+## Non-callable `__call__`
+
+```py
+class NonCallable:
+    __call__ = 1
+
+a = NonCallable()
+# error: "Object of type `NonCallable` is not callable"
+reveal_type(a())  # revealed: Unknown
+```
+
+## Possibly non-callable `__call__`
+
+```py
+def flag() -> bool: ...
+
+class NonCallable:
+    if flag():
+        __call__ = 1
+    else:
+        def __call__(self) -> int: ...
+
+a = NonCallable()
+# error: "Object of type `Literal[1] | Literal[__call__]` is not callable (due to union element `Literal[1]`)"
+reveal_type(a())  # revealed: Unknown | int
+```

crates/red_knot_python_semantic/resources/mdtest/call/function.md

@@ -44,3 +44,16 @@ reveal_type(bar())  # revealed: @Todo
 nonsense = 123
 x = nonsense()  # error: "Object of type `Literal[123]` is not callable"
 ```
+
+## Potentially unbound function
+
+```py
+def flag() -> bool: ...
+
+if flag():
+    def foo() -> int:
+        return 42
+
+# error: [possibly-unresolved-reference]
+reveal_type(foo())  # revealed: int
+```

crates/red_knot_python_semantic/resources/mdtest/comparison/identity_tests.md

@@ -0,0 +1,40 @@
+# Identity tests
+
+```py
+class A: ...
+
+def get_a() -> A: ...
+def get_object() -> object: ...
+
+a1 = get_a()
+a2 = get_a()
+
+n1 = None
+n2 = None
+
+o = get_object()
+
+reveal_type(a1 is a1)  # revealed: bool
+reveal_type(a1 is a2)  # revealed: bool
+
+reveal_type(n1 is n1)  # revealed: Literal[True]
+reveal_type(n1 is n2)  # revealed: Literal[True]
+
+reveal_type(a1 is n1)  # revealed: Literal[False]
+reveal_type(n1 is a1)  # revealed: Literal[False]
+
+reveal_type(a1 is o)  # revealed: bool
+reveal_type(n1 is o)  # revealed: bool
+
+reveal_type(a1 is not a1)  # revealed: bool
+reveal_type(a1 is not a2)  # revealed: bool
+
+reveal_type(n1 is not n1)  # revealed: Literal[False]
+reveal_type(n1 is not n2)  # revealed: Literal[False]
+
+reveal_type(a1 is not n1)  # revealed: Literal[True]
+reveal_type(n1 is not a1)  # revealed: Literal[True]
+
+reveal_type(a1 is not o)  # revealed: bool
+reveal_type(n1 is not o)  # revealed: bool
+```

crates/red_knot_python_semantic/resources/mdtest/comparison/instances/membership_test.md

@@ -93,13 +93,11 @@ class AlwaysFalse:
     def __contains__(self, item: int) -> Literal[""]:
         return ""
 
-# TODO: it should be Literal[True] and Literal[False]
-reveal_type(42 in AlwaysTrue())  # revealed: @Todo
-reveal_type(42 not in AlwaysTrue())  # revealed: @Todo
+reveal_type(42 in AlwaysTrue())  # revealed: Literal[True]
+reveal_type(42 not in AlwaysTrue())  # revealed: Literal[False]
 
-# TODO: it should be Literal[False] and Literal[True]
-reveal_type(42 in AlwaysFalse())  # revealed: @Todo
-reveal_type(42 not in AlwaysFalse())  # revealed: @Todo
+reveal_type(42 in AlwaysFalse())  # revealed: Literal[False]
+reveal_type(42 not in AlwaysFalse())  # revealed: Literal[True]
 ```
 
 ## No Fallback for `__contains__`

crates/red_knot_python_semantic/resources/mdtest/comparison/intersections.md

@@ -0,0 +1,155 @@
+# Comparison: Intersections
+
+## Positive contributions
+
+If we have an intersection type `A & B` and we get a definitive true/false answer for one of the
+types, we can infer that the result for the intersection type is also true/false:
+
+```py
+class Base: ...
+
+class Child1(Base):
+    def __eq__(self, other) -> Literal[True]:
+        return True
+
+class Child2(Base): ...
+
+def get_base() -> Base: ...
+
+x = get_base()
+c1 = Child1()
+
+# Create an intersection type through narrowing:
+if isinstance(x, Child1):
+    if isinstance(x, Child2):
+        reveal_type(x)  # revealed: Child1 & Child2
+
+        reveal_type(x == 1)  # revealed: Literal[True]
+
+        # Other comparison operators fall back to the base type:
+        reveal_type(x > 1)  # revealed: bool
+        reveal_type(x is c1)  # revealed: bool
+```
+
+## Negative contributions
+
+Negative contributions to the intersection type only allow simplifications in a few special cases
+(equality and identity comparisons).
+
+### Equality comparisons
+
+#### Literal strings
+
+```py
+x = "x" * 1_000_000_000
+y = "y" * 1_000_000_000
+reveal_type(x)  # revealed: LiteralString
+
+if x != "abc":
+    reveal_type(x)  # revealed: LiteralString & ~Literal["abc"]
+
+    reveal_type(x == "abc")  # revealed: Literal[False]
+    reveal_type("abc" == x)  # revealed: Literal[False]
+    reveal_type(x == "something else")  # revealed: bool
+    reveal_type("something else" == x)  # revealed: bool
+
+    reveal_type(x != "abc")  # revealed: Literal[True]
+    reveal_type("abc" != x)  # revealed: Literal[True]
+    reveal_type(x != "something else")  # revealed: bool
+    reveal_type("something else" != x)  # revealed: bool
+
+    reveal_type(x == y)  # revealed: bool
+    reveal_type(y == x)  # revealed: bool
+    reveal_type(x != y)  # revealed: bool
+    reveal_type(y != x)  # revealed: bool
+
+    reveal_type(x >= "abc")  # revealed: bool
+    reveal_type("abc" >= x)  # revealed: bool
+
+    reveal_type(x in "abc")  # revealed: bool
+    reveal_type("abc" in x)  # revealed: bool
+```
+
+#### Integers
+
+```py
+def get_int() -> int: ...
+
+x = get_int()
+
+if x != 1:
+    reveal_type(x)  # revealed: int & ~Literal[1]
+
+    reveal_type(x != 1)  # revealed: Literal[True]
+    reveal_type(x != 2)  # revealed: bool
+
+    reveal_type(x == 1)  # revealed: Literal[False]
+    reveal_type(x == 2)  # revealed: bool
+```
+
+### Identity comparisons
+
+```py
+class A: ...
+
+def get_object() -> object: ...
+
+o = object()
+
+a = A()
+n = None
+
+if o is not None:
+    reveal_type(o)  # revealed: object & ~None
+
+    reveal_type(o is n)  # revealed: Literal[False]
+    reveal_type(o is not n)  # revealed: Literal[True]
+```
+
+## Diagnostics
+
+### Unsupported operators for positive contributions
+
+Raise an error if any of the positive contributions to the intersection type are unsupported for the
+given operator:
+
+```py
+class Container:
+    def __contains__(self, x) -> bool: ...
+
+class NonContainer: ...
+
+def get_object() -> object: ...
+
+x = get_object()
+
+if isinstance(x, Container):
+    if isinstance(x, NonContainer):
+        reveal_type(x)  # revealed: Container & NonContainer
+
+        # error: [unsupported-operator] "Operator `in` is not supported for types `int` and `NonContainer`"
+        reveal_type(2 in x)  # revealed: bool
+```
+
+### Unsupported operators for negative contributions
+
+Do *not* raise an error if any of the negative contributions to the intersection type are
+unsupported for the given operator:
+
+```py
+class Container:
+    def __contains__(self, x) -> bool: ...
+
+class NonContainer: ...
+
+def get_object() -> object: ...
+
+x = get_object()
+
+if isinstance(x, Container):
+    if not isinstance(x, NonContainer):
+        reveal_type(x)  # revealed: Container & ~NonContainer
+
+        # No error here!
+        reveal_type(2 in x)  # revealed: bool
+```

crates/red_knot_python_semantic/resources/mdtest/exception/invalid_exception_syntax.md

@@ -0,0 +1,13 @@
+# Exception Handling
+
+## Invalid syntax
+
+```py
+from typing_extensions import reveal_type
+
+try:
+    print
+except as e:  # error: [invalid-syntax]
+    reveal_type(e)  # revealed: Unknown
+
+```

crates/red_knot_python_semantic/resources/mdtest/expression/if.md

@@ -0,0 +1,24 @@
+# If expression
+
+## Union
+
+```py
+def bool_instance() -> bool:
+    return True
+
+reveal_type(1 if bool_instance() else 2)  # revealed: Literal[1, 2]
+```
+
+## Statically known branches
+
+```py
+reveal_type(1 if True else 2)  # revealed: Literal[1]
+reveal_type(1 if "not empty" else 2)  # revealed: Literal[1]
+reveal_type(1 if (1,) else 2)  # revealed: Literal[1]
+reveal_type(1 if 1 else 2)  # revealed: Literal[1]
+
+reveal_type(1 if False else 2)  # revealed: Literal[2]
+reveal_type(1 if None else 2)  # revealed: Literal[2]
+reveal_type(1 if "" else 2)  # revealed: Literal[2]
+reveal_type(1 if 0 else 2)  # revealed: Literal[2]
+```

crates/red_knot_python_semantic/resources/mdtest/literal/literal.md

@@ -0,0 +1,91 @@
+# Literal
+
+<https://typing.readthedocs.io/en/latest/spec/literal.html#literals>
+
+## Parameterization
+
+```py
+from typing import Literal
+from enum import Enum
+
+mode: Literal["w", "r"]
+mode2: Literal["w"] | Literal["r"]
+union_var: Literal[Literal[Literal[1, 2, 3], "foo"], 5, None]
+a1: Literal[26]
+a2: Literal[0x1A]
+a3: Literal[-4]
+a4: Literal["hello world"]
+a5: Literal[b"hello world"]
+a6: Literal[True]
+a7: Literal[None]
+a8: Literal[Literal[1]]
+a9: Literal[Literal["w"], Literal["r"], Literal[Literal["w+"]]]
+
+class Color(Enum):
+    RED = 0
+    GREEN = 1
+    BLUE = 2
+
+b1: Literal[Color.RED]
+
+def f():
+    reveal_type(mode)  # revealed: Literal["w", "r"]
+    reveal_type(mode2)  # revealed: Literal["w", "r"]
+    # TODO: should be revealed: Literal[1, 2, 3, "foo", 5] | None
+    reveal_type(union_var)  # revealed: Literal[1, 2, 3, 5] | Literal["foo"] | None
+    reveal_type(a1)  # revealed: Literal[26]
+    reveal_type(a2)  # revealed: Literal[26]
+    reveal_type(a3)  # revealed: Literal[-4]
+    reveal_type(a4)  # revealed: Literal["hello world"]
+    reveal_type(a5)  # revealed: Literal[b"hello world"]
+    reveal_type(a6)  # revealed: Literal[True]
+    reveal_type(a7)  # revealed: None
+    reveal_type(a8)  # revealed: Literal[1]
+    reveal_type(a9)  # revealed: Literal["w", "r", "w+"]
+    # TODO: This should be Color.RED
+    reveal_type(b1)  # revealed: Literal[0]
+
+# error: [invalid-literal-parameter]
+invalid1: Literal[3 + 4]
+# error: [invalid-literal-parameter]
+invalid2: Literal[4 + 3j]
+# error: [invalid-literal-parameter]
+invalid3: Literal[(3, 4)]
+invalid4: Literal[
+    1 + 2,  # error: [invalid-literal-parameter]
+    "foo",
+    hello,  # error: [invalid-literal-parameter]
+    (1, 2, 3),  # error: [invalid-literal-parameter]
+]
+```
+
+## Detecting Literal outside typing and typing_extensions
+
+Only Literal that is defined in typing and typing_extension modules is detected as the special
+Literal.
+
+```pyi path=other.pyi
+from typing import _SpecialForm
+
+Literal: _SpecialForm
+```
+
+```py
+from other import Literal
+
+a1: Literal[26]
+
+def f():
+    reveal_type(a1)  # revealed: @Todo
+```
+
+## Detecting typing_extensions.Literal
+
+```py
+from typing_extensions import Literal
+
+a1: Literal[26]
+
+def f():
+    reveal_type(a1)  # revealed: Literal[26]
+```

crates/red_knot_python_semantic/resources/mdtest/loops/for_loop.md

@@ -238,7 +238,7 @@ class Test:
 def coinflip() -> bool:
     return True
 
-# TODO: we should emit a diagnostic here (it might not be iterable)
+# error: [not-iterable] "Object of type `Test | Literal[42]` is not iterable because its `__iter__` method is possibly unbound"
 for x in Test() if coinflip() else 42:
     reveal_type(x)  # revealed: int
 ```

crates/red_knot_python_semantic/resources/mdtest/metaclass.md

@@ -0,0 +1,196 @@
+## Default
+
+```py
+class M(type): ...
+
+reveal_type(M.__class__)  # revealed: Literal[type]
+```
+
+## `object`
+
+```py
+reveal_type(object.__class__)  # revealed: Literal[type]
+```
+
+## `type`
+
+```py
+reveal_type(type.__class__)  # revealed: Literal[type]
+```
+
+## Basic
+
+```py
+class M(type): ...
+class B(metaclass=M): ...
+
+reveal_type(B.__class__)  # revealed: Literal[M]
+```
+
+## Invalid metaclass
+
+A class which doesn't inherit `type` (and/or doesn't implement a custom `__new__` accepting the same
+arguments as `type.__new__`) isn't a valid metaclass.
+
+```py
+class M: ...
+class A(metaclass=M): ...
+
+# TODO: emit a diagnostic for the invalid metaclass
+reveal_type(A.__class__)  # revealed: Literal[M]
+```
+
+## Linear inheritance
+
+If a class is a subclass of a class with a custom metaclass, then the subclass will also have that
+metaclass.
+
+```py
+class M(type): ...
+class A(metaclass=M): ...
+class B(A): ...
+
+reveal_type(B.__class__)  # revealed: Literal[M]
+```
+
+## Conflict (1)
+
+The metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its
+bases. ("Strict subclass" is a synonym for "proper subclass"; a non-strict subclass can be a
+subclass or the class itself.)
+
+```py
+class M1(type): ...
+class M2(type): ...
+class A(metaclass=M1): ...
+class B(metaclass=M2): ...
+
+# error: [conflicting-metaclass] "The metaclass of a derived class (`C`) must be a subclass of the metaclasses of all its bases, but `M1` (metaclass of base class `A`) and `M2` (metaclass of base class `B`) have no subclass relationship"
+class C(A, B): ...
+
+reveal_type(C.__class__)  # revealed: Unknown
+```
+
+## Conflict (2)
+
+The metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its
+bases. ("Strict subclass" is a synonym for "proper subclass"; a non-strict subclass can be a
+subclass or the class itself.)
+
+```py
+class M1(type): ...
+class M2(type): ...
+class A(metaclass=M1): ...
+
+# error: [conflicting-metaclass] "The metaclass of a derived class (`B`) must be a subclass of the metaclasses of all its bases, but `M2` (metaclass of `B`) and `M1` (metaclass of base class `A`) have no subclass relationship"
+class B(A, metaclass=M2): ...
+
+reveal_type(B.__class__)  # revealed: Unknown
+```
+
+## Common metaclass
+
+A class has two explicit bases, both of which have the same metaclass.
+
+```py
+class M(type): ...
+class A(metaclass=M): ...
+class B(metaclass=M): ...
+class C(A, B): ...
+
+reveal_type(C.__class__)  # revealed: Literal[M]
+```
+
+## Metaclass metaclass
+
+A class has an explicit base with a custom metaclass. That metaclass itself has a custom metaclass.
+
+```py
+class M1(type): ...
+class M2(type, metaclass=M1): ...
+class M3(M2): ...
+class A(metaclass=M3): ...
+class B(A): ...
+
+reveal_type(A.__class__)  # revealed: Literal[M3]
+```
+
+## Diamond inheritance
+
+```py
+class M(type): ...
+class M1(M): ...
+class M2(M): ...
+class M12(M1, M2): ...
+class A(metaclass=M1): ...
+class B(metaclass=M2): ...
+class C(metaclass=M12): ...
+
+# error: [conflicting-metaclass] "The metaclass of a derived class (`D`) must be a subclass of the metaclasses of all its bases, but `M1` (metaclass of base class `A`) and `M2` (metaclass of base class `B`) have no subclass relationship"
+class D(A, B, C): ...
+
+reveal_type(D.__class__)  # revealed: Unknown
+```
+
+## Unknown
+
+```py
+from nonexistent_module import UnknownClass  # error: [unresolved-import]
+
+class C(UnknownClass): ...
+
+# TODO: should be `type[type] & Unknown`
+reveal_type(C.__class__)  # revealed: Literal[type]
+
+class M(type): ...
+class A(metaclass=M): ...
+class B(A, UnknownClass): ...
+
+# TODO: should be `type[M] & Unknown`
+reveal_type(B.__class__)  # revealed: Literal[M]
+```
+
+## Duplicate
+
+```py
+class M(type): ...
+class A(metaclass=M): ...
+class B(A, A): ...  # error: [duplicate-base] "Duplicate base class `A`"
+
+reveal_type(B.__class__)  # revealed: Literal[M]
+```
+
+## Non-class
+
+When a class has an explicit `metaclass` that is not a class, but is a callable that accepts
+`type.__new__` arguments, we should return the meta type of its return type.
+
+```py
+def f(*args, **kwargs) -> int: ...
+
+class A(metaclass=f): ...
+
+# TODO should be `type[int]`
+reveal_type(A.__class__)  # revealed: @Todo
+```
+
+## Cyclic
+
+Retrieving the metaclass of a cyclically defined class should not cause an infinite loop.
+
+```py path=a.pyi
+class A(B): ...  # error: [cyclic-class-def]
+class B(C): ...  # error: [cyclic-class-def]
+class C(A): ...  # error: [cyclic-class-def]
+
+reveal_type(A.__class__)  # revealed: Unknown
+```
+
+## PEP 695 generic
+
+```py
+class M(type): ...
+class A[T: str](metaclass=M): ...
+
+reveal_type(A.__class__)  # revealed: Literal[M]
+```

crates/red_knot_python_semantic/resources/mdtest/mro.md

@@ -0,0 +1,409 @@
+# Method Resolution Order tests
+
+Tests that assert that we can infer the correct type for a class's `__mro__` attribute.
+
+This attribute is rarely accessed directly at runtime. However, it's extremely important for *us* to
+know the precise possible values of a class's Method Resolution Order, or we won't be able to infer
+the correct type of attributes accessed from instances.
+
+For documentation on method resolution orders, see:
+
+- <https://docs.python.org/3/glossary.html#term-method-resolution-order>
+- <https://docs.python.org/3/howto/mro.html#python-2-3-mro>
+
+## No bases
+
+```py
+class C: ...
+
+reveal_type(C.__mro__)  # revealed: tuple[Literal[C], Literal[object]]
+```
+
+## The special case: `object` itself
+
+```py
+reveal_type(object.__mro__)  # revealed: tuple[Literal[object]]
+```
+
+## Explicit inheritance from `object`
+
+```py
+class C(object): ...
+
+reveal_type(C.__mro__)  # revealed: tuple[Literal[C], Literal[object]]
+```
+
+## Explicit inheritance from non-`object` single base
+
+```py
+class A: ...
+class B(A): ...
+
+reveal_type(B.__mro__)  # revealed: tuple[Literal[B], Literal[A], Literal[object]]
+```
+
+## Linearization of multiple bases
+
+```py
+class A: ...
+class B: ...
+class C(A, B): ...
+
+reveal_type(C.__mro__)  # revealed: tuple[Literal[C], Literal[A], Literal[B], Literal[object]]
+```
+
+## Complex diamond inheritance (1)
+
+This is "ex_2" from <https://docs.python.org/3/howto/mro.html#the-end>
+
+```py
+class O: ...
+class X(O): ...
+class Y(O): ...
+class A(X, Y): ...
+class B(Y, X): ...
+
+reveal_type(A.__mro__)  # revealed: tuple[Literal[A], Literal[X], Literal[Y], Literal[O], Literal[object]]
+reveal_type(B.__mro__)  # revealed: tuple[Literal[B], Literal[Y], Literal[X], Literal[O], Literal[object]]
+```
+
+## Complex diamond inheritance (2)
+
+This is "ex_5" from <https://docs.python.org/3/howto/mro.html#the-end>
+
+```py
+class O: ...
+class F(O): ...
+class E(O): ...
+class D(O): ...
+class C(D, F): ...
+class B(D, E): ...
+class A(B, C): ...
+
+# revealed: tuple[Literal[C], Literal[D], Literal[F], Literal[O], Literal[object]]
+reveal_type(C.__mro__)
+# revealed: tuple[Literal[B], Literal[D], Literal[E], Literal[O], Literal[object]]
+reveal_type(B.__mro__)
+# revealed: tuple[Literal[A], Literal[B], Literal[C], Literal[D], Literal[E], Literal[F], Literal[O], Literal[object]]
+reveal_type(A.__mro__)
+```
+
+## Complex diamond inheritance (3)
+
+This is "ex_6" from <https://docs.python.org/3/howto/mro.html#the-end>
+
+```py
+class O: ...
+class F(O): ...
+class E(O): ...
+class D(O): ...
+class C(D, F): ...
+class B(E, D): ...
+class A(B, C): ...
+
+# revealed: tuple[Literal[C], Literal[D], Literal[F], Literal[O], Literal[object]]
+reveal_type(C.__mro__)
+# revealed: tuple[Literal[B], Literal[E], Literal[D], Literal[O], Literal[object]]
+reveal_type(B.__mro__)
+# revealed: tuple[Literal[A], Literal[B], Literal[E], Literal[C], Literal[D], Literal[F], Literal[O], Literal[object]]
+reveal_type(A.__mro__)
+```
+
+## Complex diamond inheritance (4)
+
+This is "ex_9" from <https://docs.python.org/3/howto/mro.html#the-end>
+
+```py
+class O: ...
+class A(O): ...
+class B(O): ...
+class C(O): ...
+class D(O): ...
+class E(O): ...
+class K1(A, B, C): ...
+class K2(D, B, E): ...
+class K3(D, A): ...
+class Z(K1, K2, K3): ...
+
+# revealed: tuple[Literal[K1], Literal[A], Literal[B], Literal[C], Literal[O], Literal[object]]
+reveal_type(K1.__mro__)
+# revealed: tuple[Literal[K2], Literal[D], Literal[B], Literal[E], Literal[O], Literal[object]]
+reveal_type(K2.__mro__)
+# revealed: tuple[Literal[K3], Literal[D], Literal[A], Literal[O], Literal[object]]
+reveal_type(K3.__mro__)
+# revealed: tuple[Literal[Z], Literal[K1], Literal[K2], Literal[K3], Literal[D], Literal[A], Literal[B], Literal[C], Literal[E], Literal[O], Literal[object]]
+reveal_type(Z.__mro__)
+```
+
+## Inheritance from `Unknown`
+
+```py
+from does_not_exist import DoesNotExist  # error: [unresolved-import]
+
+class A(DoesNotExist): ...
+class B: ...
+class C: ...
+class D(A, B, C): ...
+class E(B, C): ...
+class F(E, A): ...
+
+reveal_type(A.__mro__)  # revealed: tuple[Literal[A], Unknown, Literal[object]]
+reveal_type(D.__mro__)  # revealed: tuple[Literal[D], Literal[A], Unknown, Literal[B], Literal[C], Literal[object]]
+reveal_type(E.__mro__)  # revealed: tuple[Literal[E], Literal[B], Literal[C], Literal[object]]
+reveal_type(F.__mro__)  # revealed: tuple[Literal[F], Literal[E], Literal[B], Literal[C], Literal[A], Unknown, Literal[object]]
+```
+
+## `__bases__` lists that cause errors at runtime
+
+If the class's `__bases__` cause an exception to be raised at runtime and therefore the class
+creation to fail, we infer the class's `__mro__` as being `[<class>, Unknown, object]`:
+
+```py
+# error: [inconsistent-mro] "Cannot create a consistent method resolution order (MRO) for class `Foo` with bases list `[<class 'object'>, <class 'int'>]`"
+class Foo(object, int): ...
+
+reveal_type(Foo.__mro__)  # revealed: tuple[Literal[Foo], Unknown, Literal[object]]
+
+class Bar(Foo): ...
+
+reveal_type(Bar.__mro__)  # revealed: tuple[Literal[Bar], Literal[Foo], Unknown, Literal[object]]
+
+# This is the `TypeError` at the bottom of "ex_2"
+# in the examples at <https://docs.python.org/3/howto/mro.html#the-end>
+class O: ...
+class X(O): ...
+class Y(O): ...
+class A(X, Y): ...
+class B(Y, X): ...
+
+reveal_type(A.__mro__)  # revealed: tuple[Literal[A], Literal[X], Literal[Y], Literal[O], Literal[object]]
+reveal_type(B.__mro__)  # revealed: tuple[Literal[B], Literal[Y], Literal[X], Literal[O], Literal[object]]
+
+# error: [inconsistent-mro] "Cannot create a consistent method resolution order (MRO) for class `Z` with bases list `[<class 'A'>, <class 'B'>]`"
+class Z(A, B): ...
+
+reveal_type(Z.__mro__)  # revealed: tuple[Literal[Z], Unknown, Literal[object]]
+
+class AA(Z): ...
+
+reveal_type(AA.__mro__)  # revealed: tuple[Literal[AA], Literal[Z], Unknown, Literal[object]]
+```
+
+## `__bases__` includes a `Union`
+
+We don't support union types in a class's bases; a base must resolve to a single `ClassLiteralType`.
+If we find a union type in a class's bases, we infer the class's `__mro__` as being
+`[<class>, Unknown, object]`, the same as for MROs that cause errors at runtime.
+
+```py
+def returns_bool() -> bool:
+    return True
+
+class A: ...
+class B: ...
+
+if returns_bool():
+    x = A
+else:
+    x = B
+
+reveal_type(x)  # revealed: Literal[A, B]
+
+# error: 11 [invalid-base] "Invalid class base with type `Literal[A, B]` (all bases must be a class, `Any`, `Unknown` or `Todo`)"
+class Foo(x): ...
+
+reveal_type(Foo.__mro__)  # revealed: tuple[Literal[Foo], Unknown, Literal[object]]
+```
+
+## `__bases__` includes multiple `Union`s
+
+```py
+def returns_bool() -> bool:
+    return True
+
+class A: ...
+class B: ...
+class C: ...
+class D: ...
+
+if returns_bool():
+    x = A
+else:
+    x = B
+
+if returns_bool():
+    y = C
+else:
+    y = D
+
+reveal_type(x)  # revealed: Literal[A, B]
+reveal_type(y)  # revealed: Literal[C, D]
+
+# error: 11 [invalid-base] "Invalid class base with type `Literal[A, B]` (all bases must be a class, `Any`, `Unknown` or `Todo`)"
+# error: 14 [invalid-base] "Invalid class base with type `Literal[C, D]` (all bases must be a class, `Any`, `Unknown` or `Todo`)"
+class Foo(x, y): ...
+
+reveal_type(Foo.__mro__)  # revealed: tuple[Literal[Foo], Unknown, Literal[object]]
+```
+
+## `__bases__` lists that cause errors... now with `Union`s
+
+```py
+def returns_bool() -> bool:
+    return True
+
+class O: ...
+class X(O): ...
+class Y(O): ...
+
+if bool():
+    foo = Y
+else:
+    foo = object
+
+# error: 21 [invalid-base] "Invalid class base with type `Literal[Y, object]` (all bases must be a class, `Any`, `Unknown` or `Todo`)"
+class PossibleError(foo, X): ...
+
+reveal_type(PossibleError.__mro__)  # revealed: tuple[Literal[PossibleError], Unknown, Literal[object]]
+
+class A(X, Y): ...
+
+reveal_type(A.__mro__)  # revealed: tuple[Literal[A], Literal[X], Literal[Y], Literal[O], Literal[object]]
+
+if returns_bool():
+    class B(X, Y): ...
+
+else:
+    class B(Y, X): ...
+
+# revealed: tuple[Literal[B], Literal[X], Literal[Y], Literal[O], Literal[object]] | tuple[Literal[B], Literal[Y], Literal[X], Literal[O], Literal[object]]
+reveal_type(B.__mro__)
+
+# error: 12 [invalid-base] "Invalid class base with type `Literal[B, B]` (all bases must be a class, `Any`, `Unknown` or `Todo`)"
+class Z(A, B): ...
+
+reveal_type(Z.__mro__)  # revealed: tuple[Literal[Z], Unknown, Literal[object]]
+```
+
+## `__bases__` lists with duplicate bases
+
+```py
+class Foo(str, str): ...  # error: 16 [duplicate-base] "Duplicate base class `str`"
+
+reveal_type(Foo.__mro__)  # revealed: tuple[Literal[Foo], Unknown, Literal[object]]
+
+class Spam: ...
+class Eggs: ...
+class Ham(
+    Spam,
+    Eggs,
+    Spam,  # error: [duplicate-base] "Duplicate base class `Spam`"
+    Eggs,  # error: [duplicate-base] "Duplicate base class `Eggs`"
+): ...
+
+reveal_type(Ham.__mro__)  # revealed: tuple[Literal[Ham], Unknown, Literal[object]]
+
+class Mushrooms: ...
+class Omelette(Spam, Eggs, Mushrooms, Mushrooms): ...  # error: [duplicate-base]
+
+reveal_type(Omelette.__mro__)  # revealed: tuple[Literal[Omelette], Unknown, Literal[object]]
+```
+
+## `__bases__` lists with duplicate `Unknown` bases
+
+```py
+# error: [unresolved-import]
+# error: [unresolved-import]
+from does_not_exist import unknown_object_1, unknown_object_2
+
+reveal_type(unknown_object_1)  # revealed: Unknown
+reveal_type(unknown_object_2)  # revealed: Unknown
+
+# We *should* emit an error here to warn the user that we have no idea
+# what the MRO of this class should really be.
+# However, we don't complain about "duplicate base classes" here,
+# even though two classes are both inferred as being `Unknown`.
+#
+# (TODO: should we revisit this? Does it violate the gradual guarantee?
+# Should we just silently infer `[Foo, Unknown, object]` as the MRO here
+# without emitting any error at all? Not sure...)
+#
+# error: [inconsistent-mro] "Cannot create a consistent method resolution order (MRO) for class `Foo` with bases list `[Unknown, Unknown]`"
+class Foo(unknown_object_1, unknown_object_2): ...
+
+reveal_type(Foo.__mro__)  # revealed: tuple[Literal[Foo], Unknown, Literal[object]]
+```
+
+## Unrelated objects inferred as `Any`/`Unknown` do not have special `__mro__` attributes
+
+```py
+from does_not_exist import unknown_object  # error: [unresolved-import]
+
+reveal_type(unknown_object)  # revealed: Unknown
+reveal_type(unknown_object.__mro__)  # revealed: Unknown
+```
+
+## Classes that inherit from themselves
+
+These are invalid, but we need to be able to handle them gracefully without panicking.
+
+```py path=a.pyi
+class Foo(Foo): ...  # error: [cyclic-class-def]
+
+reveal_type(Foo)  # revealed: Literal[Foo]
+reveal_type(Foo.__mro__)  # revealed: tuple[Literal[Foo], Unknown, Literal[object]]
+
+class Bar: ...
+class Baz: ...
+class Boz(Bar, Baz, Boz): ...  # error: [cyclic-class-def]
+
+reveal_type(Boz)  # revealed: Literal[Boz]
+reveal_type(Boz.__mro__)  # revealed: tuple[Literal[Boz], Unknown, Literal[object]]
+```
+
+## Classes with indirect cycles in their MROs
+
+These are similarly unlikely, but we still shouldn't crash:
+
+```py path=a.pyi
+class Foo(Bar): ...  # error: [cyclic-class-def]
+class Bar(Baz): ...  # error: [cyclic-class-def]
+class Baz(Foo): ...  # error: [cyclic-class-def]
+
+reveal_type(Foo.__mro__)  # revealed: tuple[Literal[Foo], Unknown, Literal[object]]
+reveal_type(Bar.__mro__)  # revealed: tuple[Literal[Bar], Unknown, Literal[object]]
+reveal_type(Baz.__mro__)  # revealed: tuple[Literal[Baz], Unknown, Literal[object]]
+```
+
+## Classes with cycles in their MROs, and multiple inheritance
+
+```py path=a.pyi
+class Spam: ...
+class Foo(Bar): ...  # error: [cyclic-class-def]
+class Bar(Baz): ...  # error: [cyclic-class-def]
+class Baz(Foo, Spam): ...  # error: [cyclic-class-def]
+
+reveal_type(Foo.__mro__)  # revealed: tuple[Literal[Foo], Unknown, Literal[object]]
+reveal_type(Bar.__mro__)  # revealed: tuple[Literal[Bar], Unknown, Literal[object]]
+reveal_type(Baz.__mro__)  # revealed: tuple[Literal[Baz], Unknown, Literal[object]]
+```
+
+## Classes with cycles in their MRO, and a sub-graph
+
+```py path=a.pyi
+class FooCycle(BarCycle): ...  # error: [cyclic-class-def]
+class Foo: ...
+class BarCycle(FooCycle): ...  # error: [cyclic-class-def]
+class Bar(Foo): ...
+
+# TODO: can we avoid emitting the errors for these?
+# The classes have cyclic superclasses,
+# but are not themselves cyclic...
+class Baz(Bar, BarCycle): ...  # error: [cyclic-class-def]
+class Spam(Baz): ...  # error: [cyclic-class-def]
+
+reveal_type(FooCycle.__mro__)  # revealed: tuple[Literal[FooCycle], Unknown, Literal[object]]
+reveal_type(BarCycle.__mro__)  # revealed: tuple[Literal[BarCycle], Unknown, Literal[object]]
+reveal_type(Baz.__mro__)  # revealed: tuple[Literal[Baz], Unknown, Literal[object]]
+reveal_type(Spam.__mro__)  # revealed: tuple[Literal[Spam], Unknown, Literal[object]]
+```

crates/red_knot_python_semantic/resources/mdtest/narrow/conditionals_boolean.md

@@ -0,0 +1,282 @@
+# Narrowing for conditionals with boolean expressions
+
+## Narrowing in `and` conditional
+
+```py
+class A: ...
+class B: ...
+
+def instance() -> A | B:
+    return A()
+
+x = instance()
+
+if isinstance(x, A) and isinstance(x, B):
+    reveal_type(x)  # revealed:  A & B
+else:
+    reveal_type(x)  # revealed:  B & ~A | A & ~B
+```
+
+## Arms might not add narrowing constraints
+
+```py
+class A: ...
+class B: ...
+
+def bool_instance() -> bool:
+    return True
+
+def instance() -> A | B:
+    return A()
+
+x = instance()
+
+if isinstance(x, A) and bool_instance():
+    reveal_type(x)  # revealed: A
+else:
+    reveal_type(x)  # revealed: A | B
+
+if bool_instance() and isinstance(x, A):
+    reveal_type(x)  # revealed: A
+else:
+    reveal_type(x)  # revealed: A | B
+
+reveal_type(x)  # revealed: A | B
+```
+
+## Statically known arms
+
+```py
+class A: ...
+class B: ...
+
+def instance() -> A | B:
+    return A()
+
+x = instance()
+
+if isinstance(x, A) and True:
+    reveal_type(x)  # revealed: A
+else:
+    reveal_type(x)  # revealed: B & ~A
+
+if True and isinstance(x, A):
+    reveal_type(x)  # revealed: A
+else:
+    reveal_type(x)  # revealed: B & ~A
+
+if False and isinstance(x, A):
+    # TODO: should emit an `unreachable code` diagnostic
+    reveal_type(x)  # revealed: A
+else:
+    reveal_type(x)  # revealed: A | B
+
+if False or isinstance(x, A):
+    reveal_type(x)  # revealed: A
+else:
+    reveal_type(x)  # revealed: B & ~A
+
+if True or isinstance(x, A):
+    reveal_type(x)  # revealed: A | B
+else:
+    # TODO: should emit an `unreachable code` diagnostic
+    reveal_type(x)  # revealed: B & ~A
+
+reveal_type(x)  # revealed: A | B
+```
+
+## The type of multiple symbols can be narrowed down
+
+```py
+class A: ...
+class B: ...
+
+def instance() -> A | B:
+    return A()
+
+x = instance()
+y = instance()
+
+if isinstance(x, A) and isinstance(y, B):
+    reveal_type(x)  # revealed: A
+    reveal_type(y)  # revealed: B
+else:
+    # No narrowing: Only-one or both checks might have failed
+    reveal_type(x)  # revealed: A | B
+    reveal_type(y)  # revealed: A | B
+
+reveal_type(x)  # revealed: A | B
+reveal_type(y)  # revealed: A | B
+```
+
+## Narrowing in `or` conditional
+
+```py
+class A: ...
+class B: ...
+class C: ...
+
+def instance() -> A | B | C:
+    return A()
+
+x = instance()
+
+if isinstance(x, A) or isinstance(x, B):
+    reveal_type(x)  # revealed:  A | B
+else:
+    reveal_type(x)  # revealed:  C & ~A & ~B
+```
+
+## In `or`, all arms should add constraint in order to narrow
+
+```py
+class A: ...
+class B: ...
+class C: ...
+
+def instance() -> A | B | C:
+    return A()
+
+def bool_instance() -> bool:
+    return True
+
+x = instance()
+
+if isinstance(x, A) or isinstance(x, B) or bool_instance():
+    reveal_type(x)  # revealed:  A | B | C
+else:
+    reveal_type(x)  # revealed:  C & ~A & ~B
+```
+
+## in `or`, all arms should narrow the same set of symbols
+
+```py
+class A: ...
+class B: ...
+class C: ...
+
+def instance() -> A | B | C:
+    return A()
+
+x = instance()
+y = instance()
+
+if isinstance(x, A) or isinstance(y, A):
+    # The predicate might be satisfied by the right side, so the type of `x` can’t be narrowed down here.
+    reveal_type(x)  # revealed:  A | B | C
+    # The same for `y`
+    reveal_type(y)  # revealed:  A | B | C
+else:
+    reveal_type(x)  # revealed:  B & ~A | C & ~A
+    reveal_type(y)  # revealed:  B & ~A | C & ~A
+
+if (isinstance(x, A) and isinstance(y, A)) or (isinstance(x, B) and isinstance(y, B)):
+    # Here, types of `x` and `y` can be narrowd since all `or` arms constraint them.
+    reveal_type(x)  # revealed:  A | B
+    reveal_type(y)  # revealed:  A | B
+else:
+    reveal_type(x)  # revealed:  A | B | C
+    reveal_type(y)  # revealed:  A | B | C
+```
+
+## mixing `and` and `not`
+
+```py
+class A: ...
+class B: ...
+class C: ...
+
+def instance() -> A | B | C:
+    return A()
+
+x = instance()
+
+if isinstance(x, B) and not isinstance(x, C):
+    reveal_type(x)  # revealed:  B & ~C
+else:
+    # ~(B & ~C) -> ~B | C -> (A & ~B) | (C & ~B) | C -> (A & ~B) | C
+    reveal_type(x)  # revealed: A & ~B | C
+```
+
+## mixing `or` and `not`
+
+```py
+class A: ...
+class B: ...
+class C: ...
+
+def instance() -> A | B | C:
+    return A()
+
+x = instance()
+
+if isinstance(x, B) or not isinstance(x, C):
+    reveal_type(x)  # revealed: B | A & ~C
+else:
+    reveal_type(x)  # revealed: C & ~B
+```
+
+## `or` with nested `and`
+
+```py
+class A: ...
+class B: ...
+class C: ...
+
+def instance() -> A | B | C:
+    return A()
+
+x = instance()
+
+if isinstance(x, A) or (isinstance(x, B) and not isinstance(x, C)):
+    reveal_type(x)  # revealed:  A | B & ~C
+else:
+    # ~(A | (B & ~C)) -> ~A & ~(B & ~C) -> ~A & (~B | C) -> (~A & C) | (~A ~ B)
+    reveal_type(x)  # revealed:  C & ~A
+```
+
+## `and` with nested `or`
+
+```py
+class A: ...
+class B: ...
+class C: ...
+
+def instance() -> A | B | C:
+    return A()
+
+x = instance()
+
+if isinstance(x, A) and (isinstance(x, B) or not isinstance(x, C)):
+    # A & (B | ~C) -> (A & B) | (A & ~C)
+    reveal_type(x)  # revealed:  A & B | A & ~C
+else:
+    # ~((A & B) | (A & ~C)) ->
+    # ~(A & B) & ~(A & ~C) ->
+    # (~A | ~B) & (~A | C) ->
+    # [(~A | ~B) & ~A] | [(~A | ~B) & C] ->
+    # ~A | (~A & C) | (~B & C) ->
+    # ~A | (C & ~B) ->
+    # ~A | (C & ~B)  The positive side of ~A is  A | B | C ->
+    reveal_type(x)  # revealed:  B & ~A | C & ~A | C & ~B
+```
+
+## Boolean expression internal narrowing
+
+```py
+def optional_string() -> str | None:
+    return None
+
+x = optional_string()
+y = optional_string()
+
+if x is None and y is not x:
+    reveal_type(y)  # revealed: str
+
+# Neither of the conditions alone is sufficient for narrowing y's type:
+if x is None:
+    reveal_type(y)  # revealed: str | None
+
+if y is not x:
+    reveal_type(y)  # revealed: str | None
+```

crates/red_knot_python_semantic/resources/mdtest/narrow/issubclass.md

@@ -0,0 +1,244 @@
+# Narrowing for `issubclass` checks
+
+Narrowing for `issubclass(class, classinfo)` expressions.
+
+## `classinfo` is a single type
+
+### Basic example
+
+```py
+def flag() -> bool: ...
+
+t = int if flag() else str
+
+if issubclass(t, bytes):
+    reveal_type(t)  # revealed: Never
+
+if issubclass(t, object):
+    reveal_type(t)  # revealed: Literal[int, str]
+
+if issubclass(t, int):
+    reveal_type(t)  # revealed: Literal[int]
+else:
+    reveal_type(t)  # revealed: Literal[str]
+
+if issubclass(t, str):
+    reveal_type(t)  # revealed: Literal[str]
+    if issubclass(t, int):
+        reveal_type(t)  # revealed: Never
+```
+
+### Proper narrowing in `elif` and `else` branches
+
+```py
+def flag() -> bool: ...
+
+t = int if flag() else str if flag() else bytes
+
+if issubclass(t, int):
+    reveal_type(t)  # revealed: Literal[int]
+else:
+    reveal_type(t)  # revealed: Literal[str, bytes]
+
+if issubclass(t, int):
+    reveal_type(t)  # revealed: Literal[int]
+elif issubclass(t, str):
+    reveal_type(t)  # revealed: Literal[str]
+else:
+    reveal_type(t)  # revealed: Literal[bytes]
+```
+
+### Multiple derived classes
+
+```py
+class Base: ...
+class Derived1(Base): ...
+class Derived2(Base): ...
+class Unrelated: ...
+
+def flag() -> bool: ...
+
+t1 = Derived1 if flag() else Derived2
+
+if issubclass(t1, Base):
+    reveal_type(t1)  # revealed: Literal[Derived1, Derived2]
+
+if issubclass(t1, Derived1):
+    reveal_type(t1)  # revealed: Literal[Derived1]
+else:
+    reveal_type(t1)  # revealed: Literal[Derived2]
+
+t2 = Derived1 if flag() else Base
+
+if issubclass(t2, Base):
+    reveal_type(t2)  # revealed: Literal[Derived1, Base]
+
+t3 = Derived1 if flag() else Unrelated
+
+if issubclass(t3, Base):
+    reveal_type(t3)  # revealed: Literal[Derived1]
+else:
+    reveal_type(t3)  # revealed: Literal[Unrelated]
+```
+
+### Narrowing for non-literals
+
+```py
+class A: ...
+class B: ...
+
+def get_class() -> type[object]: ...
+
+t = get_class()
+
+if issubclass(t, A):
+    reveal_type(t)  # revealed: type[A]
+    if issubclass(t, B):
+        reveal_type(t)  # revealed: type[A] & type[B]
+else:
+    reveal_type(t)  # revealed: type[object] & ~type[A]
+```
+
+### Handling of `None`
+
+```py
+from types import NoneType
+
+def flag() -> bool: ...
+
+t = int if flag() else NoneType
+
+if issubclass(t, NoneType):
+    reveal_type(t)  # revealed: Literal[NoneType]
+
+if issubclass(t, type(None)):
+    # TODO: this should be just `Literal[NoneType]`
+    reveal_type(t)  # revealed: Literal[int, NoneType]
+```
+
+## `classinfo` contains multiple types
+
+### (Nested) tuples of types
+
+```py
+class Unrelated: ...
+
+def flag() -> bool: ...
+
+t = int if flag() else str if flag() else bytes
+
+if issubclass(t, (int, (Unrelated, (bytes,)))):
+    reveal_type(t)  # revealed: Literal[int, bytes]
+else:
+    reveal_type(t)  # revealed: Literal[str]
+```
+
+## Special cases
+
+### Emit a diagnostic if the first argument is of wrong type
+
+#### Too wide
+
+`type[object]` is a subtype of `object`, but not every `object` can be passed as the first argument
+to `issubclass`:
+
+```py
+class A: ...
+
+def get_object() -> object: ...
+
+t = get_object()
+
+# TODO: we should emit a diagnostic here
+if issubclass(t, A):
+    reveal_type(t)  # revealed: type[A]
+```
+
+#### Wrong
+
+`Literal[1]` and `type` are entirely disjoint, so the inferred type of `Literal[1] & type[int]` is
+eagerly simplified to `Never` as a result of the type narrowing in the `if issubclass(t, int)`
+branch:
+
+```py
+t = 1
+
+# TODO: we should emit a diagnostic here
+if issubclass(t, int):
+    reveal_type(t)  # revealed: Never
+```
+
+### Do not use custom `issubclass` for narrowing
+
+```py
+def issubclass(c, ci):
+    return True
+
+def flag() -> bool: ...
+
+t = int if flag() else str
+if issubclass(t, int):
+    reveal_type(t)  # revealed: Literal[int, str]
+```
+
+### Do support narrowing if `issubclass` is aliased
+
+```py
+issubclass_alias = issubclass
+
+def flag() -> bool: ...
+
+t = int if flag() else str
+if issubclass_alias(t, int):
+    reveal_type(t)  # revealed: Literal[int]
+```
+
+### Do support narrowing if `issubclass` is imported
+
+```py
+from builtins import issubclass as imported_issubclass
+
+def flag() -> bool: ...
+
+t = int if flag() else str
+if imported_issubclass(t, int):
+    reveal_type(t)  # revealed: Literal[int]
+```
+
+### Do not narrow if second argument is not a proper `classinfo` argument
+
+```py
+from typing import Any
+
+def flag() -> bool: ...
+
+t = int if flag() else str
+
+# TODO: this should cause us to emit a diagnostic during
+# type checking
+if issubclass(t, "str"):
+    reveal_type(t)  # revealed: Literal[int, str]
+
+# TODO: this should cause us to emit a diagnostic during
+# type checking
+if issubclass(t, (bytes, "str")):
+    reveal_type(t)  # revealed: Literal[int, str]
+
+# TODO: this should cause us to emit a diagnostic during
+# type checking
+if issubclass(t, Any):
+    reveal_type(t)  # revealed: Literal[int, str]
+```
+
+### Do not narrow if there are keyword arguments
+
+```py
+def flag() -> bool: ...
+
+t = int if flag() else str
+
+# TODO: this should cause us to emit a diagnostic
+# (`issubclass` has no `foo` parameter)
+if issubclass(t, int, foo="bar"):
+    reveal_type(t)  # revealed: Literal[int, str]
+```

crates/red_knot_python_semantic/resources/mdtest/scopes/moduletype_attrs.md

@@ -58,11 +58,9 @@ reveal_type(typing.__name__)  # revealed: str
 reveal_type(typing.__init__)  # revealed: Literal[__init__]
 
 # These come from `builtins.object`, not `types.ModuleType`:
-# TODO: we don't currently understand `types.ModuleType` as inheriting from `object`;
-# these should not reveal `Unknown`:
-reveal_type(typing.__eq__)  # revealed: Unknown
-reveal_type(typing.__class__)  # revealed: Unknown
-reveal_type(typing.__module__)  # revealed: Unknown
+reveal_type(typing.__eq__)  # revealed: Literal[__eq__]
+
+reveal_type(typing.__class__)  # revealed: Literal[type]
 
 # TODO: needs support for attribute access on instances, properties and generics;
 # should be `dict[str, Any]`

crates/red_knot_python_semantic/resources/mdtest/stubs/class.md

@@ -6,7 +6,12 @@ In type stubs, classes can reference themselves in their base class definitions.
 `typeshed`, we have `class str(Sequence[str]): ...`.
 
 ```py path=a.pyi
-class C(C): ...
+class Foo[T]: ...
 
-reveal_type(C)  # revealed: Literal[C]
+# TODO: actually is subscriptable
+# error: [non-subscriptable]
+class Bar(Foo[Bar]): ...
+
+reveal_type(Bar)  # revealed: Literal[Bar]
+reveal_type(Bar.__mro__)  # revealed: tuple[Literal[Bar], Unknown, Literal[object]]
 ```

crates/red_knot_python_semantic/resources/mdtest/subscript/class.md

@@ -39,7 +39,8 @@ reveal_type(UnionClassGetItem[0])  # revealed: str | int
 ## Class getitem with class union
 
 ```py
-flag = True
+def bool_instance() -> bool:
+    return True
 
 class A:
     def __class_getitem__(cls, item: int) -> str:
@@ -49,7 +50,7 @@ class B:
     def __class_getitem__(cls, item: int) -> int:
         return item
 
-x = A if flag else B
+x = A if bool_instance() else B
 
 reveal_type(x)  # revealed: Literal[A, B]
 reveal_type(x[0])  # revealed: str | int

crates/red_knot_python_semantic/resources/mdtest/unary/instance.md

@@ -1,6 +1,8 @@
 # Unary Operations
 
 ```py
+from typing import Literal
+
 class Number:
     def __init__(self, value: int):
         self.value = 1
@@ -18,7 +20,7 @@ a = Number()
 
 reveal_type(+a)  # revealed: int
 reveal_type(-a)  # revealed: int
-reveal_type(~a)  # revealed: @Todo
+reveal_type(~a)  # revealed: Literal[True]
 
 class NoDunder: ...
 

crates/red_knot_python_semantic/resources/mdtest/unpacking.md

@@ -145,13 +145,8 @@ reveal_type(f)  # revealed: Unknown
 
 ### Non-iterable unpacking
 
-TODO: Remove duplicate diagnostics. This is happening because for a sequence-like assignment target,
-multiple definitions are created and the inference engine runs on each of them which results in
-duplicate diagnostics.
-
 ```py
 # error: "Object of type `Literal[1]` is not iterable"
-# error: "Object of type `Literal[1]` is not iterable"
 a, b = 1
 reveal_type(a)  # revealed: Unknown
 reveal_type(b)  # revealed: Unknown

crates/red_knot_python_semantic/src/lib.rs

@@ -22,6 +22,7 @@ pub(crate) mod site_packages;
 mod stdlib;
 pub(crate) mod symbol;
 pub mod types;
+mod unpack;
 mod util;
 
 type FxOrderSet<V> = ordermap::set::OrderSet<V, BuildHasherDefault<FxHasher>>;

crates/red_knot_python_semantic/src/semantic_index.rs

@@ -125,6 +125,7 @@ impl<'db> SemanticIndex<'db> {
     ///
     /// Use the Salsa cached [`symbol_table()`] query if you only need the
     /// symbol table for a single scope.
+    #[track_caller]
     pub(super) fn symbol_table(&self, scope_id: FileScopeId) -> Arc<SymbolTable> {
         self.symbol_tables[scope_id].clone()
     }
@@ -133,15 +134,18 @@ impl<'db> SemanticIndex<'db> {
     ///
     /// Use the Salsa cached [`use_def_map()`] query if you only need the
     /// use-def map for a single scope.
+    #[track_caller]
     pub(super) fn use_def_map(&self, scope_id: FileScopeId) -> Arc<UseDefMap> {
         self.use_def_maps[scope_id].clone()
     }
 
+    #[track_caller]
     pub(crate) fn ast_ids(&self, scope_id: FileScopeId) -> &AstIds {
         &self.ast_ids[scope_id]
     }
 
     /// Returns the ID of the `expression`'s enclosing scope.
+    #[track_caller]
     pub(crate) fn expression_scope_id(
         &self,
         expression: impl Into<ExpressionNodeKey>,
@@ -151,11 +155,13 @@ impl<'db> SemanticIndex<'db> {
 
     /// Returns the [`Scope`] of the `expression`'s enclosing scope.
     #[allow(unused)]
+    #[track_caller]
     pub(crate) fn expression_scope(&self, expression: impl Into<ExpressionNodeKey>) -> &Scope {
         &self.scopes[self.expression_scope_id(expression)]
     }
 
     /// Returns the [`Scope`] with the given id.
+    #[track_caller]
     pub(crate) fn scope(&self, id: FileScopeId) -> &Scope {
         &self.scopes[id]
     }
@@ -172,6 +178,7 @@ impl<'db> SemanticIndex<'db> {
 
     /// Returns the parent scope of `scope_id`.
     #[allow(unused)]
+    #[track_caller]
     pub(crate) fn parent_scope(&self, scope_id: FileScopeId) -> Option<&Scope> {
         Some(&self.scopes[self.parent_scope_id(scope_id)?])
     }
@@ -195,6 +202,7 @@ impl<'db> SemanticIndex<'db> {
     }
 
     /// Returns the [`Definition`] salsa ingredient for `definition_key`.
+    #[track_caller]
     pub(crate) fn definition(
         &self,
         definition_key: impl Into<DefinitionNodeKey>,
@@ -206,6 +214,7 @@ impl<'db> SemanticIndex<'db> {
     /// Panics if we have no expression ingredient for that node. We can only call this method for
     /// standalone-inferable expressions, which we call `add_standalone_expression` for in
     /// [`SemanticIndexBuilder`].
+    #[track_caller]
     pub(crate) fn expression(
         &self,
         expression_key: impl Into<ExpressionNodeKey>,
@@ -213,8 +222,18 @@ impl<'db> SemanticIndex<'db> {
         self.expressions_by_node[&expression_key.into()]
     }
 
+    pub(crate) fn try_expression(
+        &self,
+        expression_key: impl Into<ExpressionNodeKey>,
+    ) -> Option<Expression<'db>> {
+        self.expressions_by_node
+            .get(&expression_key.into())
+            .copied()
+    }
+
     /// Returns the id of the scope that `node` creates. This is different from [`Definition::scope`] which
     /// returns the scope in which that definition is defined in.
+    #[track_caller]
     pub(crate) fn node_scope(&self, node: NodeWithScopeRef) -> FileScopeId {
         self.scopes_by_node[&node.node_key()]
     }

crates/red_knot_python_semantic/src/semantic_index/ast_ids.rs

@@ -87,6 +87,14 @@ pub trait HasScopedAstId {
     fn scoped_ast_id(&self, db: &dyn Db, scope: ScopeId) -> Self::Id;
 }
 
+impl<T: HasScopedAstId> HasScopedAstId for Box<T> {
+    type Id = <T as HasScopedAstId>::Id;
+
+    fn scoped_ast_id(&self, db: &dyn Db, scope: ScopeId) -> Self::Id {
+        self.as_ref().scoped_ast_id(db, scope)
+    }
+}
+
 /// Uniquely identifies an [`ast::Expr`] in a [`crate::semantic_index::symbol::FileScopeId`].
 #[newtype_index]
 pub struct ScopedExpressionId;

crates/red_knot_python_semantic/src/semantic_index/builder.rs

@@ -25,12 +25,13 @@ use crate::semantic_index::symbol::{
 };
 use crate::semantic_index::use_def::{FlowSnapshot, UseDefMapBuilder};
 use crate::semantic_index::SemanticIndex;
+use crate::unpack::Unpack;
 use crate::Db;
 
 use super::constraint::{Constraint, ConstraintNode, PatternConstraint};
 use super::definition::{
-    AssignmentKind, DefinitionCategory, ExceptHandlerDefinitionNodeRef,
-    MatchPatternDefinitionNodeRef, WithItemDefinitionNodeRef,
+    DefinitionCategory, ExceptHandlerDefinitionNodeRef, MatchPatternDefinitionNodeRef,
+    WithItemDefinitionNodeRef,
 };
 
 mod except_handlers;
@@ -46,6 +47,7 @@ pub(super) struct SemanticIndexBuilder<'db> {
     current_assignments: Vec<CurrentAssignment<'db>>,
     /// The match case we're currently visiting.
     current_match_case: Option<CurrentMatchCase<'db>>,
+
     /// Flow states at each `break` in the current loop.
     loop_break_states: Vec<FlowSnapshot>,
     /// Per-scope contexts regarding nested `try`/`except` statements
@@ -112,9 +114,11 @@ impl<'db> SemanticIndexBuilder<'db> {
     fn push_scope_with_parent(&mut self, node: NodeWithScopeRef, parent: Option<FileScopeId>) {
         let children_start = self.scopes.next_index() + 1;
 
+        #[allow(unsafe_code)]
         let scope = Scope {
             parent,
-            kind: node.scope_kind(),
+            // SAFETY: `node` is guaranteed to be a child of `self.module`
+            node: unsafe { node.to_kind(self.module.clone()) },
             descendents: children_start..children_start,
         };
         self.try_node_context_stack_manager.enter_nested_scope();
@@ -124,15 +128,7 @@ impl<'db> SemanticIndexBuilder<'db> {
         self.use_def_maps.push(UseDefMapBuilder::new());
         let ast_id_scope = self.ast_ids.push(AstIdsBuilder::new());
 
-        #[allow(unsafe_code)]
-        // SAFETY: `node` is guaranteed to be a child of `self.module`
-        let scope_id = ScopeId::new(
-            self.db,
-            self.file,
-            file_scope_id,
-            unsafe { node.to_kind(self.module.clone()) },
-            countme::Count::default(),
-        );
+        let scope_id = ScopeId::new(self.db, self.file, file_scope_id, countme::Count::default());
 
         self.scope_ids_by_scope.push(scope_id);
         self.scopes_by_node.insert(node.node_key(), file_scope_id);
@@ -203,10 +199,10 @@ impl<'db> SemanticIndexBuilder<'db> {
         self.current_symbol_table().mark_symbol_used(id);
     }
 
-    fn add_definition<'a>(
+    fn add_definition(
         &mut self,
         symbol: ScopedSymbolId,
-        definition_node: impl Into<DefinitionNodeRef<'a>>,
+        definition_node: impl Into<DefinitionNodeRef<'db>>,
     ) -> Definition<'db> {
         let definition_node: DefinitionNodeRef<'_> = definition_node.into();
         #[allow(unsafe_code)]
@@ -285,8 +281,12 @@ impl<'db> SemanticIndexBuilder<'db> {
         debug_assert!(popped_assignment.is_some());
     }
 
-    fn current_assignment(&self) -> Option<&CurrentAssignment<'db>> {
-        self.current_assignments.last()
+    fn current_assignment(&self) -> Option<CurrentAssignment<'db>> {
+        self.current_assignments.last().copied()
+    }
+
+    fn current_assignment_mut(&mut self) -> Option<&mut CurrentAssignment<'db>> {
+        self.current_assignments.last_mut()
     }
 
     fn add_pattern_constraint(
@@ -445,7 +445,7 @@ impl<'db> SemanticIndexBuilder<'db> {
         self.pop_scope();
     }
 
-    fn declare_parameter(&mut self, parameter: AnyParameterRef) {
+    fn declare_parameter(&mut self, parameter: AnyParameterRef<'db>) {
         let symbol = self.add_symbol(parameter.name().id().clone());
 
         let definition = self.add_definition(symbol, parameter);
@@ -619,24 +619,48 @@ where
             }
             ast::Stmt::Assign(node) => {
                 debug_assert_eq!(&self.current_assignments, &[]);
+
                 self.visit_expr(&node.value);
-                self.add_standalone_expression(&node.value);
-                for (target_index, target) in node.targets.iter().enumerate() {
-                    let kind = match target {
-                        ast::Expr::List(_) | ast::Expr::Tuple(_) => Some(AssignmentKind::Sequence),
-                        ast::Expr::Name(_) => Some(AssignmentKind::Name),
+                let value = self.add_standalone_expression(&node.value);
+
+                for target in &node.targets {
+                    // We only handle assignments to names and unpackings here, other targets like
+                    // attribute and subscript are handled separately as they don't create a new
+                    // definition.
+                    let current_assignment = match target {
+                        ast::Expr::List(_) | ast::Expr::Tuple(_) => {
+                            Some(CurrentAssignment::Assign {
+                                node,
+                                first: true,
+                                unpack: Some(Unpack::new(
+                                    self.db,
+                                    self.file,
+                                    self.current_scope(),
+                                    #[allow(unsafe_code)]
+                                    unsafe {
+                                        AstNodeRef::new(self.module.clone(), target)
+                                    },
+                                    value,
+                                    countme::Count::default(),
+                                )),
+                            })
+                        }
+                        ast::Expr::Name(_) => Some(CurrentAssignment::Assign {
+                            node,
+                            unpack: None,
+                            first: false,
+                        }),
                         _ => None,
                     };
-                    if let Some(kind) = kind {
-                        self.push_assignment(CurrentAssignment::Assign {
-                            assignment: node,
-                            target_index,
-                            kind,
-                        });
+
+                    if let Some(current_assignment) = current_assignment {
+                        self.push_assignment(current_assignment);
                     }
+
                     self.visit_expr(target);
-                    if kind.is_some() {
-                        // only need to pop in the case where we pushed something
+
+                    if current_assignment.is_some() {
+                        // Only need to pop in the case where we pushed something
                         self.pop_assignment();
                     }
                 }
@@ -970,19 +994,19 @@ where
                 }
 
                 if is_definition {
-                    match self.current_assignment().copied() {
+                    match self.current_assignment() {
                         Some(CurrentAssignment::Assign {
-                            assignment,
-                            target_index,
-                            kind,
+                            node,
+                            first,
+                            unpack,
                         }) => {
                             self.add_definition(
                                 symbol,
                                 AssignmentDefinitionNodeRef {
-                                    assignment,
-                                    target_index,
+                                    unpack,
+                                    value: &node.value,
                                     name: name_node,
-                                    kind,
+                                    first,
                                 },
                             );
                         }
@@ -1033,6 +1057,11 @@ where
                     }
                 }
 
+                if let Some(CurrentAssignment::Assign { first, .. }) = self.current_assignment_mut()
+                {
+                    *first = false;
+                }
+
                 walk_expr(self, expr);
             }
             ast::Expr::Named(node) => {
@@ -1073,10 +1102,13 @@ where
                 // AST inspection, so we can't simplify here, need to record test expression for
                 // later checking)
                 self.visit_expr(test);
+                let constraint = self.record_expression_constraint(test);
                 let pre_if = self.flow_snapshot();
                 self.visit_expr(body);
                 let post_body = self.flow_snapshot();
                 self.flow_restore(pre_if);
+
+                self.record_negated_constraint(constraint);
                 self.visit_expr(orelse);
                 self.flow_merge(post_body);
             }
@@ -1229,9 +1261,9 @@ where
 #[derive(Copy, Clone, Debug, PartialEq)]
 enum CurrentAssignment<'a> {
     Assign {
-        assignment: &'a ast::StmtAssign,
-        target_index: usize,
-        kind: AssignmentKind,
+        node: &'a ast::StmtAssign,
+        first: bool,
+        unpack: Option<Unpack<'a>>,
     },
     AnnAssign(&'a ast::StmtAnnAssign),
     AugAssign(&'a ast::StmtAugAssign),

crates/red_knot_python_semantic/src/semantic_index/definition.rs

@@ -6,8 +6,22 @@ use crate::ast_node_ref::AstNodeRef;
 use crate::module_resolver::file_to_module;
 use crate::node_key::NodeKey;
 use crate::semantic_index::symbol::{FileScopeId, ScopeId, ScopedSymbolId};
+use crate::unpack::Unpack;
 use crate::Db;
 
+/// A definition of a symbol.
+///
+/// ## Module-local type
+/// This type should not be used as part of any cross-module API because
+/// it holds a reference to the AST node. Range-offset changes
+/// then propagate through all usages, and deserialization requires
+/// reparsing the entire module.
+///
+/// E.g. don't use this type in:
+///
+/// * a return type of a cross-module query
+/// * a field of a type that is a return type of a cross-module query
+/// * an argument of a cross-module query
 #[salsa::tracked]
 pub struct Definition<'db> {
     /// The file in which the definition occurs.
@@ -24,7 +38,7 @@ pub struct Definition<'db> {
 
     #[no_eq]
     #[return_ref]
-    pub(crate) kind: DefinitionKind,
+    pub(crate) kind: DefinitionKind<'db>,
 
     #[no_eq]
     count: countme::Count<Definition<'static>>,
@@ -166,10 +180,10 @@ pub(crate) struct ImportFromDefinitionNodeRef<'a> {
 
 #[derive(Copy, Clone, Debug)]
 pub(crate) struct AssignmentDefinitionNodeRef<'a> {
-    pub(crate) assignment: &'a ast::StmtAssign,
-    pub(crate) target_index: usize,
+    pub(crate) unpack: Option<Unpack<'a>>,
+    pub(crate) value: &'a ast::Expr,
     pub(crate) name: &'a ast::ExprName,
-    pub(crate) kind: AssignmentKind,
+    pub(crate) first: bool,
 }
 
 #[derive(Copy, Clone, Debug)]
@@ -211,9 +225,9 @@ pub(crate) struct MatchPatternDefinitionNodeRef<'a> {
     pub(crate) index: u32,
 }
 
-impl DefinitionNodeRef<'_> {
+impl<'db> DefinitionNodeRef<'db> {
     #[allow(unsafe_code)]
-    pub(super) unsafe fn into_owned(self, parsed: ParsedModule) -> DefinitionKind {
+    pub(super) unsafe fn into_owned(self, parsed: ParsedModule) -> DefinitionKind<'db> {
         match self {
             DefinitionNodeRef::Import(alias) => {
                 DefinitionKind::Import(AstNodeRef::new(parsed, alias))
@@ -234,15 +248,15 @@ impl DefinitionNodeRef<'_> {
                 DefinitionKind::NamedExpression(AstNodeRef::new(parsed, named))
             }
             DefinitionNodeRef::Assignment(AssignmentDefinitionNodeRef {
-                assignment,
-                target_index,
+                unpack,
+                value,
                 name,
-                kind,
+                first,
             }) => DefinitionKind::Assignment(AssignmentDefinitionKind {
-                assignment: AstNodeRef::new(parsed.clone(), assignment),
-                target_index,
+                target: TargetKind::from(unpack),
+                value: AstNodeRef::new(parsed.clone(), value),
                 name: AstNodeRef::new(parsed, name),
-                kind,
+                first,
             }),
             DefinitionNodeRef::AnnotatedAssignment(assign) => {
                 DefinitionKind::AnnotatedAssignment(AstNodeRef::new(parsed, assign))
@@ -316,10 +330,10 @@ impl DefinitionNodeRef<'_> {
             Self::Class(node) => node.into(),
             Self::NamedExpression(node) => node.into(),
             Self::Assignment(AssignmentDefinitionNodeRef {
-                assignment: _,
-                target_index: _,
+                value: _,
+                unpack: _,
                 name,
-                kind: _,
+                first: _,
             }) => name.into(),
             Self::AnnotatedAssignment(node) => node.into(),
             Self::AugmentedAssignment(node) => node.into(),
@@ -382,13 +396,13 @@ impl DefinitionCategory {
 }
 
 #[derive(Clone, Debug)]
-pub enum DefinitionKind {
+pub enum DefinitionKind<'db> {
     Import(AstNodeRef<ast::Alias>),
     ImportFrom(ImportFromDefinitionKind),
     Function(AstNodeRef<ast::StmtFunctionDef>),
     Class(AstNodeRef<ast::StmtClassDef>),
     NamedExpression(AstNodeRef<ast::ExprNamed>),
-    Assignment(AssignmentDefinitionKind),
+    Assignment(AssignmentDefinitionKind<'db>),
     AnnotatedAssignment(AstNodeRef<ast::StmtAnnAssign>),
     AugmentedAssignment(AstNodeRef<ast::StmtAugAssign>),
     For(ForStmtDefinitionKind),
@@ -400,7 +414,7 @@ pub enum DefinitionKind {
     ExceptHandler(ExceptHandlerDefinitionKind),
 }
 
-impl DefinitionKind {
+impl DefinitionKind<'_> {
     pub(crate) fn category(&self) -> DefinitionCategory {
         match self {
             // functions, classes, and imports always bind, and we consider them declarations
@@ -445,6 +459,21 @@ impl DefinitionKind {
     }
 }
 
+#[derive(Copy, Clone, Debug, PartialEq)]
+pub(crate) enum TargetKind<'db> {
+    Sequence(Unpack<'db>),
+    Name,
+}
+
+impl<'db> From<Option<Unpack<'db>>> for TargetKind<'db> {
+    fn from(value: Option<Unpack<'db>>) -> Self {
+        match value {
+            Some(unpack) => TargetKind::Sequence(unpack),
+            None => TargetKind::Name,
+        }
+    }
+}
+
 #[derive(Clone, Debug)]
 #[allow(dead_code)]
 pub struct MatchPatternDefinitionKind {
@@ -506,38 +535,31 @@ impl ImportFromDefinitionKind {
 }
 
 #[derive(Clone, Debug)]
-pub struct AssignmentDefinitionKind {
-    assignment: AstNodeRef<ast::StmtAssign>,
-    target_index: usize,
+pub struct AssignmentDefinitionKind<'db> {
+    target: TargetKind<'db>,
+    value: AstNodeRef<ast::Expr>,
     name: AstNodeRef<ast::ExprName>,
-    kind: AssignmentKind,
+    first: bool,
 }
 
-impl AssignmentDefinitionKind {
-    pub(crate) fn value(&self) -> &ast::Expr {
-        &self.assignment.node().value
+impl<'db> AssignmentDefinitionKind<'db> {
+    pub(crate) fn target(&self) -> TargetKind<'db> {
+        self.target
     }
 
-    pub(crate) fn target(&self) -> &ast::Expr {
-        &self.assignment.node().targets[self.target_index]
+    pub(crate) fn value(&self) -> &ast::Expr {
+        self.value.node()
     }
 
     pub(crate) fn name(&self) -> &ast::ExprName {
         self.name.node()
     }
 
-    pub(crate) fn kind(&self) -> AssignmentKind {
-        self.kind
+    pub(crate) fn is_first(&self) -> bool {
+        self.first
     }
 }
 
-/// The kind of assignment target expression.
-#[derive(Clone, Copy, Debug, PartialEq, Eq)]
-pub enum AssignmentKind {
-    Sequence,
-    Name,
-}
-
 #[derive(Clone, Debug)]
 pub struct WithItemDefinitionKind {
     node: AstNodeRef<ast::WithItem>,

crates/red_knot_python_semantic/src/semantic_index/expression.rs

@@ -8,6 +8,18 @@ use salsa;
 /// An independently type-inferable expression.
 ///
 /// Includes constraint expressions (e.g. if tests) and the RHS of an unpacking assignment.
+///
+/// ## Module-local type
+/// This type should not be used as part of any cross-module API because
+/// it holds a reference to the AST node. Range-offset changes
+/// then propagate through all usages, and deserialization requires
+/// reparsing the entire module.
+///
+/// E.g. don't use this type in:
+///
+/// * a return type of a cross-module query
+/// * a field of a type that is a return type of a cross-module query
+/// * an argument of a cross-module query
 #[salsa::tracked]
 pub(crate) struct Expression<'db> {
     /// The file in which the expression occurs.

crates/red_knot_python_semantic/src/semantic_index/symbol.rs

@@ -103,14 +103,10 @@ pub struct ScopedSymbolId;
 pub struct ScopeId<'db> {
     #[id]
     pub file: File,
+
     #[id]
     pub file_scope_id: FileScopeId,
 
-    /// The node that introduces this scope.
-    #[no_eq]
-    #[return_ref]
-    pub node: NodeWithScopeKind,
-
     #[no_eq]
     count: countme::Count<ScopeId<'static>>,
 }
@@ -131,6 +127,14 @@ impl<'db> ScopeId<'db> {
         )
     }
 
+    pub(crate) fn node(self, db: &dyn Db) -> &NodeWithScopeKind {
+        self.scope(db).node()
+    }
+
+    pub(crate) fn scope(self, db: &dyn Db) -> &Scope {
+        semantic_index(db, self.file(db)).scope(self.file_scope_id(db))
+    }
+
     #[cfg(test)]
     pub(crate) fn name(self, db: &'db dyn Db) -> &'db str {
         match self.node(db) {
@@ -169,10 +173,10 @@ impl FileScopeId {
     }
 }
 
-#[derive(Debug, Eq, PartialEq)]
+#[derive(Debug)]
 pub struct Scope {
     pub(super) parent: Option<FileScopeId>,
-    pub(super) kind: ScopeKind,
+    pub(super) node: NodeWithScopeKind,
     pub(super) descendents: Range<FileScopeId>,
 }
 
@@ -181,8 +185,12 @@ impl Scope {
         self.parent
     }
 
+    pub fn node(&self) -> &NodeWithScopeKind {
+        &self.node
+    }
+
     pub fn kind(&self) -> ScopeKind {
-        self.kind
+        self.node().scope_kind()
     }
 }
 
@@ -376,21 +384,6 @@ impl NodeWithScopeRef<'_> {
         }
     }
 
-    pub(super) fn scope_kind(self) -> ScopeKind {
-        match self {
-            NodeWithScopeRef::Module => ScopeKind::Module,
-            NodeWithScopeRef::Class(_) => ScopeKind::Class,
-            NodeWithScopeRef::Function(_) => ScopeKind::Function,
-            NodeWithScopeRef::Lambda(_) => ScopeKind::Function,
-            NodeWithScopeRef::FunctionTypeParameters(_)
-            | NodeWithScopeRef::ClassTypeParameters(_) => ScopeKind::Annotation,
-            NodeWithScopeRef::ListComprehension(_)
-            | NodeWithScopeRef::SetComprehension(_)
-            | NodeWithScopeRef::DictComprehension(_)
-            | NodeWithScopeRef::GeneratorExpression(_) => ScopeKind::Comprehension,
-        }
-    }
-
     pub(crate) fn node_key(self) -> NodeWithScopeKey {
         match self {
             NodeWithScopeRef::Module => NodeWithScopeKey::Module,
@@ -438,6 +431,36 @@ pub enum NodeWithScopeKind {
     GeneratorExpression(AstNodeRef<ast::ExprGenerator>),
 }
 
+impl NodeWithScopeKind {
+    pub(super) const fn scope_kind(&self) -> ScopeKind {
+        match self {
+            Self::Module => ScopeKind::Module,
+            Self::Class(_) => ScopeKind::Class,
+            Self::Function(_) => ScopeKind::Function,
+            Self::Lambda(_) => ScopeKind::Function,
+            Self::FunctionTypeParameters(_) | Self::ClassTypeParameters(_) => ScopeKind::Annotation,
+            Self::ListComprehension(_)
+            | Self::SetComprehension(_)
+            | Self::DictComprehension(_)
+            | Self::GeneratorExpression(_) => ScopeKind::Comprehension,
+        }
+    }
+
+    pub fn expect_class(&self) -> &ast::StmtClassDef {
+        match self {
+            Self::Class(class) => class.node(),
+            _ => panic!("expected class"),
+        }
+    }
+
+    pub fn expect_function(&self) -> &ast::StmtFunctionDef {
+        match self {
+            Self::Function(function) => function.node(),
+            _ => panic!("expected function"),
+        }
+    }
+}
+
 #[derive(Copy, Clone, Debug, Eq, PartialEq, Hash)]
 pub(crate) enum NodeWithScopeKey {
     Module,

crates/red_knot_python_semantic/src/stdlib.rs

@@ -11,11 +11,9 @@ use crate::Db;
 enum CoreStdlibModule {
     Builtins,
     Types,
-    // the Typing enum is currently only used in tests
-    #[allow(dead_code)]
-    Typing,
     Typeshed,
     TypingExtensions,
+    Typing,
 }
 
 impl CoreStdlibModule {

crates/red_knot_python_semantic/src/symbol.rs

@@ -3,7 +3,7 @@ use crate::{
     Db,
 };
 
-#[derive(Debug, Clone, Copy, PartialEq)]
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
 pub(crate) enum Boundness {
     Bound,
     MayBeUnbound,
@@ -44,17 +44,13 @@ impl<'db> Symbol<'db> {
         }
     }
 
-    pub(crate) fn unwrap_or(&self, other: Type<'db>) -> Type<'db> {
+    pub(crate) fn unwrap_or_unknown(&self) -> Type<'db> {
         match self {
             Symbol::Type(ty, _) => *ty,
-            Symbol::Unbound => other,
+            Symbol::Unbound => Type::Unknown,
         }
     }
 
-    pub(crate) fn unwrap_or_unknown(&self) -> Type<'db> {
-        self.unwrap_or(Type::Unknown)
-    }
-
     pub(crate) fn as_type(&self) -> Option<Type<'db>> {
         match self {
             Symbol::Type(ty, _) => Some(*ty),

crates/red_knot_python_semantic/src/types/builder.rs

@@ -25,12 +25,11 @@
 //!   * No type in an intersection can be a supertype of any other type in the intersection (just
 //!     eliminate the supertype from the intersection).
 //!   * An intersection containing two non-overlapping types should simplify to [`Type::Never`].
-use crate::types::{IntersectionType, Type, UnionType};
+
+use crate::types::{InstanceType, IntersectionType, KnownClass, Type, UnionType};
 use crate::{Db, FxOrderSet};
 use smallvec::SmallVec;
 
-use super::KnownClass;
-
 pub(crate) struct UnionBuilder<'db> {
     elements: Vec<Type<'db>>,
     db: &'db dyn Db,
@@ -80,7 +79,6 @@ impl<'db> UnionBuilder<'db> {
                         to_remove.push(index);
                     }
                 }
-
                 match to_remove[..] {
                     [] => self.elements.push(to_add),
                     [index] => self.elements[index] = to_add,
@@ -103,7 +101,6 @@ impl<'db> UnionBuilder<'db> {
                 }
             }
         }
-
         self
     }
 
@@ -249,8 +246,8 @@ impl<'db> InnerIntersectionBuilder<'db> {
             }
         } else {
             // ~Literal[True] & bool = Literal[False]
-            if let Type::Instance(class_type) = new_positive {
-                if class_type.is_known(db, KnownClass::Bool) {
+            if let Type::Instance(InstanceType { class }) = new_positive {
+                if class.is_known(db, KnownClass::Bool) {
                     if let Some(&Type::BooleanLiteral(value)) = self
                         .negative
                         .iter()
@@ -320,7 +317,7 @@ impl<'db> InnerIntersectionBuilder<'db> {
                 // Adding any of these types to the negative side of an intersection
                 // is equivalent to adding it to the positive side. We do this to
                 // simplify the representation.
-                self.positive.insert(ty);
+                self.add_positive(db, ty);
             }
             // ~Literal[True] & bool = Literal[False]
             Type::BooleanLiteral(bool)
@@ -386,8 +383,9 @@ mod tests {
     use crate::program::{Program, SearchPathSettings};
     use crate::python_version::PythonVersion;
     use crate::stdlib::typing_symbol;
-    use crate::types::{KnownClass, StringLiteralType, UnionBuilder};
+    use crate::types::{global_symbol, KnownClass, StringLiteralType, UnionBuilder};
     use crate::ProgramSettings;
+    use ruff_db::files::system_path_to_file;
     use ruff_db::system::{DbWithTestSystem, SystemPathBuf};
     use test_case::test_case;
 
@@ -594,6 +592,22 @@ mod tests {
         assert_eq!(ta_not_i0.display(&db).to_string(), "int & Any | Literal[1]");
     }
 
+    #[test]
+    fn build_intersection_simplify_negative_any() {
+        let db = setup_db();
+
+        let ty = IntersectionBuilder::new(&db)
+            .add_negative(Type::Any)
+            .build();
+        assert_eq!(ty, Type::Any);
+
+        let ty = IntersectionBuilder::new(&db)
+            .add_positive(Type::Never)
+            .add_negative(Type::Any)
+            .build();
+        assert_eq!(ty, Type::Never);
+    }
+
     #[test]
     fn intersection_distributes_over_union() {
         let db = setup_db();
@@ -675,8 +689,8 @@ mod tests {
     fn build_intersection_self_negation() {
         let db = setup_db();
         let ty = IntersectionBuilder::new(&db)
-            .add_positive(Type::None)
-            .add_negative(Type::None)
+            .add_positive(Type::none(&db))
+            .add_negative(Type::none(&db))
             .build();
 
         assert_eq!(ty, Type::Never);
@@ -686,18 +700,18 @@ mod tests {
     fn build_intersection_simplify_negative_never() {
         let db = setup_db();
         let ty = IntersectionBuilder::new(&db)
-            .add_positive(Type::None)
+            .add_positive(Type::none(&db))
             .add_negative(Type::Never)
             .build();
 
-        assert_eq!(ty, Type::None);
+        assert_eq!(ty, Type::none(&db));
     }
 
     #[test]
     fn build_intersection_simplify_positive_never() {
         let db = setup_db();
         let ty = IntersectionBuilder::new(&db)
-            .add_positive(Type::None)
+            .add_positive(Type::none(&db))
             .add_positive(Type::Never)
             .build();
 
@@ -709,14 +723,14 @@ mod tests {
         let db = setup_db();
 
         let ty = IntersectionBuilder::new(&db)
-            .add_negative(Type::None)
+            .add_negative(Type::none(&db))
             .add_positive(Type::IntLiteral(1))
             .build();
         assert_eq!(ty, Type::IntLiteral(1));
 
         let ty = IntersectionBuilder::new(&db)
             .add_positive(Type::IntLiteral(1))
-            .add_negative(Type::None)
+            .add_negative(Type::none(&db))
             .build();
         assert_eq!(ty, Type::IntLiteral(1));
     }
@@ -875,7 +889,7 @@ mod tests {
         let db = setup_db();
 
         let t1 = Type::IntLiteral(1);
-        let t2 = Type::None;
+        let t2 = Type::none(&db);
 
         let ty = IntersectionBuilder::new(&db)
             .add_positive(t1)
@@ -993,4 +1007,66 @@ mod tests {
             .build();
         assert_eq!(result, ty);
     }
+
+    #[test]
+    fn build_intersection_of_two_unions_simplify() {
+        let mut db = setup_db();
+        db.write_dedented(
+            "/src/module.py",
+            "
+            class A: ...
+            class B: ...
+            a = A()
+            b = B()
+        ",
+        )
+        .unwrap();
+
+        let file = system_path_to_file(&db, "src/module.py").expect("file to exist");
+
+        let a = global_symbol(&db, file, "a").expect_type();
+        let b = global_symbol(&db, file, "b").expect_type();
+        let union = UnionBuilder::new(&db).add(a).add(b).build();
+        assert_eq!(union.display(&db).to_string(), "A | B");
+        let reversed_union = UnionBuilder::new(&db).add(b).add(a).build();
+        assert_eq!(reversed_union.display(&db).to_string(), "B | A");
+        let intersection = IntersectionBuilder::new(&db)
+            .add_positive(union)
+            .add_positive(reversed_union)
+            .build();
+        assert_eq!(intersection.display(&db).to_string(), "B | A");
+    }
+
+    #[test]
+    fn build_union_of_two_intersections_simplify() {
+        let mut db = setup_db();
+        db.write_dedented(
+            "/src/module.py",
+            "
+            class A: ...
+            class B: ...
+            a = A()
+            b = B()
+        ",
+        )
+        .unwrap();
+
+        let file = system_path_to_file(&db, "src/module.py").expect("file to exist");
+
+        let a = global_symbol(&db, file, "a").expect_type();
+        let b = global_symbol(&db, file, "b").expect_type();
+        let intersection = IntersectionBuilder::new(&db)
+            .add_positive(a)
+            .add_positive(b)
+            .build();
+        let reversed_intersection = IntersectionBuilder::new(&db)
+            .add_positive(b)
+            .add_positive(a)
+            .build();
+        let union = UnionBuilder::new(&db)
+            .add(intersection)
+            .add(reversed_intersection)
+            .build();
+        assert_eq!(union.display(&db).to_string(), "A & B");
+    }
 }

crates/red_knot_python_semantic/src/types/diagnostic.rs

@@ -1,14 +1,15 @@
+use crate::types::{ClassLiteralType, Type};
+use crate::Db;
+use ruff_db::diagnostic::{Diagnostic, Severity};
 use ruff_db::files::File;
 use ruff_python_ast::{self as ast, AnyNodeRef};
 use ruff_text_size::{Ranged, TextRange};
+use std::borrow::Cow;
 use std::fmt::Formatter;
 use std::ops::Deref;
 use std::sync::Arc;
 
-use crate::types::Type;
-use crate::Db;
-
-#[derive(Debug, Eq, PartialEq)]
+#[derive(Debug, Eq, PartialEq, Clone)]
 pub struct TypeCheckDiagnostic {
     // TODO: Don't use string keys for rules
     pub(super) rule: String,
@@ -31,6 +32,28 @@ impl TypeCheckDiagnostic {
     }
 }
 
+impl Diagnostic for TypeCheckDiagnostic {
+    fn rule(&self) -> &str {
+        TypeCheckDiagnostic::rule(self)
+    }
+
+    fn message(&self) -> Cow<str> {
+        TypeCheckDiagnostic::message(self).into()
+    }
+
+    fn file(&self) -> File {
+        TypeCheckDiagnostic::file(self)
+    }
+
+    fn range(&self) -> Option<TextRange> {
+        Some(Ranged::range(self))
+    }
+
+    fn severity(&self) -> Severity {
+        Severity::Error
+    }
+}
+
 impl Ranged for TypeCheckDiagnostic {
     fn range(&self) -> TextRange {
         self.range
@@ -141,6 +164,23 @@ impl<'db> TypeCheckDiagnosticsBuilder<'db> {
         );
     }
 
+    /// Emit a diagnostic declaring that the object represented by `node` is not iterable
+    /// because its `__iter__` method is possibly unbound.
+    pub(super) fn add_not_iterable_possibly_unbound(
+        &mut self,
+        node: AnyNodeRef,
+        element_ty: Type<'db>,
+    ) {
+        self.add(
+            node,
+            "not-iterable",
+            format_args!(
+                "Object of type `{}` is not iterable because its `__iter__` method is possibly unbound",
+                element_ty.display(self.db)
+            ),
+        );
+    }
+
     /// Emit a diagnostic declaring that an index is out of bounds for a tuple.
     pub(super) fn add_index_out_of_bounds(
         &mut self,
@@ -209,7 +249,7 @@ impl<'db> TypeCheckDiagnosticsBuilder<'db> {
         assigned_ty: Type<'db>,
     ) {
         match declared_ty {
-            Type::ClassLiteral(class) => {
+            Type::ClassLiteral(ClassLiteralType { class }) => {
                 self.add(node, "invalid-assignment", format_args!(
                         "Implicit shadowing of class `{}`; annotate to make it explicit if this is intentional",
                         class.name(self.db)));

crates/red_knot_python_semantic/src/types/display.rs

@@ -6,7 +6,9 @@ use ruff_db::display::FormatterJoinExtension;
 use ruff_python_ast::str::Quote;
 use ruff_python_literal::escape::AsciiEscape;
 
-use crate::types::{IntersectionType, Type, UnionType};
+use crate::types::{
+    ClassLiteralType, InstanceType, IntersectionType, KnownClass, SubclassOfType, Type, UnionType,
+};
 use crate::Db;
 use rustc_hash::FxHashMap;
 
@@ -64,7 +66,11 @@ impl Display for DisplayRepresentation<'_> {
             Type::Any => f.write_str("Any"),
             Type::Never => f.write_str("Never"),
             Type::Unknown => f.write_str("Unknown"),
-            Type::None => f.write_str("None"),
+            Type::Instance(InstanceType { class })
+                if class.is_known(self.db, KnownClass::NoneType) =>
+            {
+                f.write_str("None")
+            }
             // `[Type::Todo]`'s display should be explicit that is not a valid display of
             // any other type
             Type::Todo => f.write_str("@Todo"),
@@ -72,8 +78,12 @@ impl Display for DisplayRepresentation<'_> {
                 write!(f, "<module '{:?}'>", file.path(self.db))
             }
             // TODO functions and classes should display using a fully qualified name
-            Type::ClassLiteral(class) => f.write_str(class.name(self.db)),
-            Type::Instance(class) => f.write_str(class.name(self.db)),
+            Type::ClassLiteral(ClassLiteralType { class }) => f.write_str(class.name(self.db)),
+            Type::SubclassOf(SubclassOfType { class }) => {
+                write!(f, "type[{}]", class.name(self.db))
+            }
+            Type::Instance(InstanceType { class }) => f.write_str(class.name(self.db)),
+            Type::KnownInstance(known_instance) => f.write_str(known_instance.as_str()),
             Type::FunctionLiteral(function) => f.write_str(function.name(self.db)),
             Type::Union(union) => union.display(self.db).fmt(f),
             Type::Intersection(intersection) => intersection.display(self.db).fmt(f),
@@ -380,7 +390,7 @@ mod tests {
             global_symbol(&db, mod_file, "bar").expect_type(),
             global_symbol(&db, mod_file, "B").expect_type(),
             Type::BooleanLiteral(true),
-            Type::None,
+            Type::none(&db),
         ];
         let union = UnionType::from_elements(&db, union_elements).expect_union();
         let display = format!("{}", union.display(&db));

crates/red_knot_python_semantic/src/types/mro.rs

@@ -0,0 +1,505 @@
+use std::collections::VecDeque;
+use std::ops::Deref;
+
+use indexmap::IndexSet;
+use itertools::Either;
+use rustc_hash::FxHashSet;
+
+use super::{Class, ClassLiteralType, KnownClass, KnownInstanceType, Type};
+use crate::Db;
+
+/// The inferred method resolution order of a given class.
+///
+/// See [`Class::iter_mro`] for more details.
+#[derive(PartialEq, Eq, Clone, Debug)]
+pub(super) struct Mro<'db>(Box<[ClassBase<'db>]>);
+
+impl<'db> Mro<'db> {
+    /// Attempt to resolve the MRO of a given class
+    ///
+    /// In the event that a possible list of bases would (or could) lead to a
+    /// `TypeError` being raised at runtime due to an unresolvable MRO, we infer
+    /// the MRO of the class as being `[<the class in question>, Unknown, object]`.
+    /// This seems most likely to reduce the possibility of cascading errors
+    /// elsewhere.
+    ///
+    /// (We emit a diagnostic warning about the runtime `TypeError` in
+    /// [`super::infer::TypeInferenceBuilder::infer_region_scope`].)
+    pub(super) fn of_class(db: &'db dyn Db, class: Class<'db>) -> Result<Self, MroError<'db>> {
+        Self::of_class_impl(db, class).map_err(|error_kind| {
+            let fallback_mro = Self::from([
+                ClassBase::Class(class),
+                ClassBase::Unknown,
+                ClassBase::object(db),
+            ]);
+            MroError {
+                kind: error_kind,
+                fallback_mro,
+            }
+        })
+    }
+
+    fn of_class_impl(db: &'db dyn Db, class: Class<'db>) -> Result<Self, MroErrorKind<'db>> {
+        let class_bases = class.explicit_bases(db);
+
+        match class_bases {
+            // `builtins.object` is the special case:
+            // the only class in Python that has an MRO with length <2
+            [] if class.is_known(db, KnownClass::Object) => {
+                Ok(Self::from([ClassBase::Class(class)]))
+            }
+
+            // All other classes in Python have an MRO with length >=2.
+            // Even if a class has no explicit base classes,
+            // it will implicitly inherit from `object` at runtime;
+            // `object` will appear in the class's `__bases__` list and `__mro__`:
+            //
+            // ```pycon
+            // >>> class Foo: ...
+            // ...
+            // >>> Foo.__bases__
+            // (<class 'object'>,)
+            // >>> Foo.__mro__
+            // (<class '__main__.Foo'>, <class 'object'>)
+            // ```
+            [] => Ok(Self::from([ClassBase::Class(class), ClassBase::object(db)])),
+
+            // Fast path for a class that has only a single explicit base.
+            //
+            // This *could* theoretically be handled by the final branch below,
+            // but it's a common case (i.e., worth optimizing for),
+            // and the `c3_merge` function requires lots of allocations.
+            [single_base] => {
+                let single_base = ClassBase::try_from_ty(*single_base).ok_or(*single_base);
+                single_base.map_or_else(
+                    |invalid_base_ty| {
+                        let bases_info = Box::from([(0, invalid_base_ty)]);
+                        Err(MroErrorKind::InvalidBases(bases_info))
+                    },
+                    |single_base| {
+                        if let ClassBase::Class(class_base) = single_base {
+                            if class_is_cyclically_defined(db, class_base) {
+                                return Err(MroErrorKind::CyclicClassDefinition);
+                            }
+                        }
+                        let mro = std::iter::once(ClassBase::Class(class))
+                            .chain(single_base.mro(db))
+                            .collect();
+                        Ok(mro)
+                    },
+                )
+            }
+
+            // The class has multiple explicit bases.
+            //
+            // We'll fallback to a full implementation of the C3-merge algorithm to determine
+            // what MRO Python will give this class at runtime
+            // (if an MRO is indeed resolvable at all!)
+            multiple_bases => {
+                if class_is_cyclically_defined(db, class) {
+                    return Err(MroErrorKind::CyclicClassDefinition);
+                }
+
+                let mut valid_bases = vec![];
+                let mut invalid_bases = vec![];
+
+                for (i, base) in multiple_bases.iter().enumerate() {
+                    match ClassBase::try_from_ty(*base).ok_or(*base) {
+                        Ok(valid_base) => valid_bases.push(valid_base),
+                        Err(invalid_base) => invalid_bases.push((i, invalid_base)),
+                    }
+                }
+
+                if !invalid_bases.is_empty() {
+                    return Err(MroErrorKind::InvalidBases(invalid_bases.into_boxed_slice()));
+                }
+
+                let mut seqs = vec![VecDeque::from([ClassBase::Class(class)])];
+                for base in &valid_bases {
+                    seqs.push(base.mro(db).collect());
+                }
+                seqs.push(valid_bases.iter().copied().collect());
+
+                c3_merge(seqs).ok_or_else(|| {
+                    let mut seen_bases = FxHashSet::default();
+                    let mut duplicate_bases = vec![];
+                    for (index, base) in valid_bases
+                        .iter()
+                        .enumerate()
+                        .filter_map(|(index, base)| Some((index, base.into_class_literal_type()?)))
+                    {
+                        if !seen_bases.insert(base) {
+                            duplicate_bases.push((index, base));
+                        }
+                    }
+
+                    if duplicate_bases.is_empty() {
+                        MroErrorKind::UnresolvableMro {
+                            bases_list: valid_bases.into_boxed_slice(),
+                        }
+                    } else {
+                        MroErrorKind::DuplicateBases(duplicate_bases.into_boxed_slice())
+                    }
+                })
+            }
+        }
+    }
+}
+
+impl<'db, const N: usize> From<[ClassBase<'db>; N]> for Mro<'db> {
+    fn from(value: [ClassBase<'db>; N]) -> Self {
+        Self(Box::from(value))
+    }
+}
+
+impl<'db> From<Vec<ClassBase<'db>>> for Mro<'db> {
+    fn from(value: Vec<ClassBase<'db>>) -> Self {
+        Self(value.into_boxed_slice())
+    }
+}
+
+impl<'db> Deref for Mro<'db> {
+    type Target = [ClassBase<'db>];
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+impl<'db> FromIterator<ClassBase<'db>> for Mro<'db> {
+    fn from_iter<T: IntoIterator<Item = ClassBase<'db>>>(iter: T) -> Self {
+        Self(iter.into_iter().collect())
+    }
+}
+
+/// Iterator that yields elements of a class's MRO.
+///
+/// We avoid materialising the *full* MRO unless it is actually necessary:
+/// - Materialising the full MRO is expensive
+/// - We need to do it for every class in the code that we're checking, as we need to make sure
+///   that there are no class definitions in the code we're checking that would cause an
+///   exception to be raised at runtime. But the same does *not* necessarily apply for every class
+///   in third-party and stdlib dependencies: we never emit diagnostics about non-first-party code.
+/// - However, we *do* need to resolve attribute accesses on classes/instances from
+///   third-party and stdlib dependencies. That requires iterating over the MRO of third-party/stdlib
+///   classes, but not necessarily the *whole* MRO: often just the first element is enough.
+///   Luckily we know that for any class `X`, the first element of `X`'s MRO will always be `X` itself.
+///   We can therefore avoid resolving the full MRO for many third-party/stdlib classes while still
+///   being faithful to the runtime semantics.
+///
+/// Even for first-party code, where we will have to resolve the MRO for every class we encounter,
+/// loading the cached MRO comes with a certain amount of overhead, so it's best to avoid calling the
+/// Salsa-tracked [`Class::try_mro`] method unless it's absolutely necessary.
+pub(super) struct MroIterator<'db> {
+    db: &'db dyn Db,
+
+    /// The class whose MRO we're iterating over
+    class: Class<'db>,
+
+    /// Whether or not we've already yielded the first element of the MRO
+    first_element_yielded: bool,
+
+    /// Iterator over all elements of the MRO except the first.
+    ///
+    /// The full MRO is expensive to materialize, so this field is `None`
+    /// unless we actually *need* to iterate past the first element of the MRO,
+    /// at which point it is lazily materialized.
+    subsequent_elements: Option<std::slice::Iter<'db, ClassBase<'db>>>,
+}
+
+impl<'db> MroIterator<'db> {
+    pub(super) fn new(db: &'db dyn Db, class: Class<'db>) -> Self {
+        Self {
+            db,
+            class,
+            first_element_yielded: false,
+            subsequent_elements: None,
+        }
+    }
+
+    /// Materialize the full MRO of the class.
+    /// Return an iterator over that MRO which skips the first element of the MRO.
+    fn full_mro_except_first_element(&mut self) -> impl Iterator<Item = ClassBase<'db>> + '_ {
+        self.subsequent_elements
+            .get_or_insert_with(|| {
+                let mut full_mro_iter = match self.class.try_mro(self.db) {
+                    Ok(mro) => mro.iter(),
+                    Err(error) => error.fallback_mro().iter(),
+                };
+                full_mro_iter.next();
+                full_mro_iter
+            })
+            .copied()
+    }
+}
+
+impl<'db> Iterator for MroIterator<'db> {
+    type Item = ClassBase<'db>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        if !self.first_element_yielded {
+            self.first_element_yielded = true;
+            return Some(ClassBase::Class(self.class));
+        }
+        self.full_mro_except_first_element().next()
+    }
+}
+
+impl std::iter::FusedIterator for MroIterator<'_> {}
+
+#[derive(Debug, PartialEq, Eq)]
+pub(super) struct MroError<'db> {
+    kind: MroErrorKind<'db>,
+    fallback_mro: Mro<'db>,
+}
+
+impl<'db> MroError<'db> {
+    /// Return an [`MroErrorKind`] variant describing why we could not resolve the MRO for this class.
+    pub(super) fn reason(&self) -> &MroErrorKind<'db> {
+        &self.kind
+    }
+
+    /// Return the fallback MRO we should infer for this class during type inference
+    /// (since accurate resolution of its "true" MRO was impossible)
+    pub(super) fn fallback_mro(&self) -> &Mro<'db> {
+        &self.fallback_mro
+    }
+}
+
+/// Possible ways in which attempting to resolve the MRO of a class might fail.
+#[derive(Debug, PartialEq, Eq)]
+pub(super) enum MroErrorKind<'db> {
+    /// The class inherits from one or more invalid bases.
+    ///
+    /// To avoid excessive complexity in our implementation,
+    /// we only permit classes to inherit from class-literal types,
+    /// `Todo`, `Unknown` or `Any`. Anything else results in us
+    /// emitting a diagnostic.
+    ///
+    /// This variant records the indices and types of class bases
+    /// that we deem to be invalid. The indices are the indices of nodes
+    /// in the bases list of the class's [`StmtClassDef`](ruff_python_ast::StmtClassDef) node.
+    /// Each index is the index of a node representing an invalid base.
+    InvalidBases(Box<[(usize, Type<'db>)]>),
+
+    /// The class inherits from itself!
+    ///
+    /// This is very unlikely to happen in working real-world code,
+    /// but it's important to explicitly account for it.
+    /// If we don't, there's a possibility of an infinite loop and a panic.
+    CyclicClassDefinition,
+
+    /// The class has one or more duplicate bases.
+    ///
+    /// This variant records the indices and [`Class`]es
+    /// of the duplicate bases. The indices are the indices of nodes
+    /// in the bases list of the class's [`StmtClassDef`](ruff_python_ast::StmtClassDef) node.
+    /// Each index is the index of a node representing a duplicate base.
+    DuplicateBases(Box<[(usize, Class<'db>)]>),
+
+    /// The MRO is otherwise unresolvable through the C3-merge algorithm.
+    ///
+    /// See [`c3_merge`] for more details.
+    UnresolvableMro { bases_list: Box<[ClassBase<'db>]> },
+}
+
+/// Enumeration of the possible kinds of types we allow in class bases.
+///
+/// This is much more limited than the [`Type`] enum:
+/// all types that would be invalid to have as a class base are
+/// transformed into [`ClassBase::Unknown`]
+#[derive(Debug, Copy, Clone, PartialEq, Eq)]
+pub(super) enum ClassBase<'db> {
+    Any,
+    Unknown,
+    Todo,
+    Class(Class<'db>),
+}
+
+impl<'db> ClassBase<'db> {
+    pub(super) fn display(self, db: &'db dyn Db) -> impl std::fmt::Display + 'db {
+        struct Display<'db> {
+            base: ClassBase<'db>,
+            db: &'db dyn Db,
+        }
+
+        impl std::fmt::Display for Display<'_> {
+            fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+                match self.base {
+                    ClassBase::Any => f.write_str("Any"),
+                    ClassBase::Todo => f.write_str("Todo"),
+                    ClassBase::Unknown => f.write_str("Unknown"),
+                    ClassBase::Class(class) => write!(f, "<class '{}'>", class.name(self.db)),
+                }
+            }
+        }
+
+        Display { base: self, db }
+    }
+
+    #[cfg(test)]
+    #[track_caller]
+    pub(super) fn expect_class_base(self) -> Class<'db> {
+        match self {
+            ClassBase::Class(class) => class,
+            _ => panic!("Expected a `ClassBase::Class()` variant"),
+        }
+    }
+
+    /// Return a `ClassBase` representing the class `builtins.object`
+    fn object(db: &'db dyn Db) -> Self {
+        KnownClass::Object
+            .to_class(db)
+            .into_class_literal()
+            .map_or(Self::Unknown, |ClassLiteralType { class }| {
+                Self::Class(class)
+            })
+    }
+
+    /// Attempt to resolve `ty` into a `ClassBase`.
+    ///
+    /// Return `None` if `ty` is not an acceptable type for a class base.
+    fn try_from_ty(ty: Type<'db>) -> Option<Self> {
+        match ty {
+            Type::Any => Some(Self::Any),
+            Type::Unknown => Some(Self::Unknown),
+            Type::Todo => Some(Self::Todo),
+            Type::ClassLiteral(ClassLiteralType { class }) => Some(Self::Class(class)),
+            Type::Union(_) => None, // TODO -- forces consideration of multiple possible MROs?
+            Type::Intersection(_) => None, // TODO -- probably incorrect?
+            Type::Instance(_) => None, // TODO -- handle `__mro_entries__`?
+            Type::Never
+            | Type::BooleanLiteral(_)
+            | Type::FunctionLiteral(_)
+            | Type::BytesLiteral(_)
+            | Type::IntLiteral(_)
+            | Type::StringLiteral(_)
+            | Type::LiteralString
+            | Type::Tuple(_)
+            | Type::SliceLiteral(_)
+            | Type::ModuleLiteral(_)
+            | Type::SubclassOf(_) => None,
+            Type::KnownInstance(known_instance) => match known_instance {
+                KnownInstanceType::Literal => None,
+            },
+        }
+    }
+
+    fn into_class_literal_type(self) -> Option<Class<'db>> {
+        match self {
+            Self::Class(class) => Some(class),
+            _ => None,
+        }
+    }
+
+    /// Iterate over the MRO of this base
+    fn mro(
+        self,
+        db: &'db dyn Db,
+    ) -> Either<impl Iterator<Item = ClassBase<'db>>, impl Iterator<Item = ClassBase<'db>>> {
+        match self {
+            ClassBase::Any => Either::Left([ClassBase::Any, ClassBase::object(db)].into_iter()),
+            ClassBase::Unknown => {
+                Either::Left([ClassBase::Unknown, ClassBase::object(db)].into_iter())
+            }
+            ClassBase::Todo => Either::Left([ClassBase::Todo, ClassBase::object(db)].into_iter()),
+            ClassBase::Class(class) => Either::Right(class.iter_mro(db)),
+        }
+    }
+}
+
+impl<'db> From<ClassBase<'db>> for Type<'db> {
+    fn from(value: ClassBase<'db>) -> Self {
+        match value {
+            ClassBase::Any => Type::Any,
+            ClassBase::Todo => Type::Todo,
+            ClassBase::Unknown => Type::Unknown,
+            ClassBase::Class(class) => Type::ClassLiteral(ClassLiteralType { class }),
+        }
+    }
+}
+
+/// Implementation of the [C3-merge algorithm] for calculating a Python class's
+/// [method resolution order].
+///
+/// [C3-merge algorithm]: https://docs.python.org/3/howto/mro.html#python-2-3-mro
+/// [method resolution order]: https://docs.python.org/3/glossary.html#term-method-resolution-order
+fn c3_merge(mut sequences: Vec<VecDeque<ClassBase>>) -> Option<Mro> {
+    // Most MROs aren't that long...
+    let mut mro = Vec::with_capacity(8);
+
+    loop {
+        sequences.retain(|sequence| !sequence.is_empty());
+
+        if sequences.is_empty() {
+            return Some(Mro::from(mro));
+        }
+
+        // If the candidate exists "deeper down" in the inheritance hierarchy,
+        // we should refrain from adding it to the MRO for now. Add the first candidate
+        // for which this does not hold true. If this holds true for all candidates,
+        // return `None`; it will be impossible to find a consistent MRO for the class
+        // with the given bases.
+        let mro_entry = sequences.iter().find_map(|outer_sequence| {
+            let candidate = outer_sequence[0];
+
+            let not_head = sequences
+                .iter()
+                .all(|sequence| sequence.iter().skip(1).all(|base| base != &candidate));
+
+            not_head.then_some(candidate)
+        })?;
+
+        mro.push(mro_entry);
+
+        // Make sure we don't try to add the candidate to the MRO twice:
+        for sequence in &mut sequences {
+            if sequence[0] == mro_entry {
+                sequence.pop_front();
+            }
+        }
+    }
+}
+
+/// Return `true` if this class appears to be a cyclic definition,
+/// i.e., it inherits either directly or indirectly from itself.
+///
+/// A class definition like this will fail at runtime,
+/// but we must be resilient to it or we could panic.
+fn class_is_cyclically_defined(db: &dyn Db, class: Class) -> bool {
+    fn is_cyclically_defined_recursive<'db>(
+        db: &'db dyn Db,
+        class: Class<'db>,
+        classes_to_watch: &mut IndexSet<Class<'db>>,
+    ) -> bool {
+        if !classes_to_watch.insert(class) {
+            return true;
+        }
+        for explicit_base_class in class
+            .explicit_bases(db)
+            .iter()
+            .copied()
+            .filter_map(Type::into_class_literal)
+            .map(|ClassLiteralType { class }| class)
+        {
+            // Each base must be considered in isolation.
+            // This is due to the fact that if a class uses multiple inheritance,
+            // there could easily be a situation where two bases have the same class in their MROs;
+            // that isn't enough to constitute the class being cyclically defined.
+            let classes_to_watch_len = classes_to_watch.len();
+            if is_cyclically_defined_recursive(db, explicit_base_class, classes_to_watch) {
+                return true;
+            }
+            classes_to_watch.truncate(classes_to_watch_len);
+        }
+        false
+    }
+
+    class
+        .explicit_bases(db)
+        .iter()
+        .copied()
+        .filter_map(Type::into_class_literal)
+        .map(|ClassLiteralType { class }| class)
+        .any(|base_class| is_cyclically_defined_recursive(db, base_class, &mut IndexSet::default()))
+}

crates/red_knot_python_semantic/src/types/narrow.rs

@@ -5,12 +5,15 @@ use crate::semantic_index::expression::Expression;
 use crate::semantic_index::symbol::{ScopeId, ScopedSymbolId, SymbolTable};
 use crate::semantic_index::symbol_table;
 use crate::types::{
-    infer_expression_types, IntersectionBuilder, KnownFunction, Type, UnionBuilder,
+    infer_expression_types, ClassLiteralType, InstanceType, IntersectionBuilder, KnownClass,
+    KnownConstraintFunction, KnownFunction, Truthiness, Type, UnionBuilder,
 };
 use crate::Db;
 use itertools::Itertools;
 use ruff_python_ast as ast;
+use ruff_python_ast::{BoolOp, ExprBoolOp};
 use rustc_hash::FxHashMap;
+use std::collections::hash_map::Entry;
 use std::sync::Arc;
 
 /// Return the type constraint that `test` (if true) would place on `definition`, if any.
@@ -34,21 +37,20 @@ pub(crate) fn narrowing_constraint<'db>(
     constraint: Constraint<'db>,
     definition: Definition<'db>,
 ) -> Option<Type<'db>> {
-    match constraint.node {
+    let constraints = match constraint.node {
         ConstraintNode::Expression(expression) => {
             if constraint.is_positive {
                 all_narrowing_constraints_for_expression(db, expression)
-                    .get(&definition.symbol(db))
-                    .copied()
             } else {
                 all_negative_narrowing_constraints_for_expression(db, expression)
-                    .get(&definition.symbol(db))
-                    .copied()
             }
         }
-        ConstraintNode::Pattern(pattern) => all_narrowing_constraints_for_pattern(db, pattern)
-            .get(&definition.symbol(db))
-            .copied(),
+        ConstraintNode::Pattern(pattern) => all_narrowing_constraints_for_pattern(db, pattern),
+    };
+    if let Some(constraints) = constraints {
+        constraints.get(&definition.symbol(db)).copied()
+    } else {
+        None
     }
 }
 
@@ -56,7 +58,7 @@ pub(crate) fn narrowing_constraint<'db>(
 fn all_narrowing_constraints_for_pattern<'db>(
     db: &'db dyn Db,
     pattern: PatternConstraint<'db>,
-) -> NarrowingConstraints<'db> {
+) -> Option<NarrowingConstraints<'db>> {
     NarrowingConstraintsBuilder::new(db, ConstraintNode::Pattern(pattern), true).finish()
 }
 
@@ -64,7 +66,7 @@ fn all_narrowing_constraints_for_pattern<'db>(
 fn all_narrowing_constraints_for_expression<'db>(
     db: &'db dyn Db,
     expression: Expression<'db>,
-) -> NarrowingConstraints<'db> {
+) -> Option<NarrowingConstraints<'db>> {
     NarrowingConstraintsBuilder::new(db, ConstraintNode::Expression(expression), true).finish()
 }
 
@@ -72,39 +74,83 @@ fn all_narrowing_constraints_for_expression<'db>(
 fn all_negative_narrowing_constraints_for_expression<'db>(
     db: &'db dyn Db,
     expression: Expression<'db>,
-) -> NarrowingConstraints<'db> {
+) -> Option<NarrowingConstraints<'db>> {
     NarrowingConstraintsBuilder::new(db, ConstraintNode::Expression(expression), false).finish()
 }
 
-/// Generate a constraint from the *type* of the second argument of an `isinstance` call.
+/// Generate a constraint from the type of a `classinfo` argument to `isinstance` or `issubclass`.
 ///
-/// Example: for `isinstance(…, str)`, we would infer `Type::ClassLiteral(str)` from the
-/// second argument, but we need to generate a `Type::Instance(str)` constraint that can
-/// be used to narrow down the type of the first argument.
-fn generate_isinstance_constraint<'db>(
+/// The `classinfo` argument can be a class literal, a tuple of (tuples of) class literals. PEP 604
+/// union types are not yet supported. Returns `None` if the `classinfo` argument has a wrong type.
+fn generate_classinfo_constraint<'db, F>(
     db: &'db dyn Db,
     classinfo: &Type<'db>,
-) -> Option<Type<'db>> {
+    to_constraint: F,
+) -> Option<Type<'db>>
+where
+    F: Fn(ClassLiteralType<'db>) -> Type<'db> + Copy,
+{
     match classinfo {
-        Type::ClassLiteral(class) => Some(Type::Instance(*class)),
         Type::Tuple(tuple) => {
             let mut builder = UnionBuilder::new(db);
             for element in tuple.elements(db) {
-                builder = builder.add(generate_isinstance_constraint(db, element)?);
+                builder = builder.add(generate_classinfo_constraint(db, element, to_constraint)?);
             }
             Some(builder.build())
         }
+        Type::ClassLiteral(class_literal_type) => Some(to_constraint(*class_literal_type)),
         _ => None,
     }
 }
 
 type NarrowingConstraints<'db> = FxHashMap<ScopedSymbolId, Type<'db>>;
 
+fn merge_constraints_and<'db>(
+    into: &mut NarrowingConstraints<'db>,
+    from: NarrowingConstraints<'db>,
+    db: &'db dyn Db,
+) {
+    for (key, value) in from {
+        match into.entry(key) {
+            Entry::Occupied(mut entry) => {
+                *entry.get_mut() = IntersectionBuilder::new(db)
+                    .add_positive(*entry.get())
+                    .add_positive(value)
+                    .build();
+            }
+            Entry::Vacant(entry) => {
+                entry.insert(value);
+            }
+        }
+    }
+}
+
+fn merge_constraints_or<'db>(
+    into: &mut NarrowingConstraints<'db>,
+    from: &NarrowingConstraints<'db>,
+    db: &'db dyn Db,
+) {
+    for (key, value) in from {
+        match into.entry(*key) {
+            Entry::Occupied(mut entry) => {
+                *entry.get_mut() = UnionBuilder::new(db).add(*entry.get()).add(*value).build();
+            }
+            Entry::Vacant(entry) => {
+                entry.insert(KnownClass::Object.to_instance(db));
+            }
+        }
+    }
+    for (key, value) in into.iter_mut() {
+        if !from.contains_key(key) {
+            *value = KnownClass::Object.to_instance(db);
+        }
+    }
+}
+
 struct NarrowingConstraintsBuilder<'db> {
     db: &'db dyn Db,
     constraint: ConstraintNode<'db>,
     is_positive: bool,
-    constraints: NarrowingConstraints<'db>,
 }
 
 impl<'db> NarrowingConstraintsBuilder<'db> {
@@ -113,24 +159,31 @@ impl<'db> NarrowingConstraintsBuilder<'db> {
             db,
             constraint,
             is_positive,
-            constraints: NarrowingConstraints::default(),
         }
     }
 
-    fn finish(mut self) -> NarrowingConstraints<'db> {
-        match self.constraint {
+    fn finish(mut self) -> Option<NarrowingConstraints<'db>> {
+        let constraints: Option<NarrowingConstraints<'db>> = match self.constraint {
             ConstraintNode::Expression(expression) => {
-                self.evaluate_expression_constraint(expression, self.is_positive);
+                self.evaluate_expression_constraint(expression, self.is_positive)
             }
             ConstraintNode::Pattern(pattern) => self.evaluate_pattern_constraint(pattern),
+        };
+        if let Some(mut constraints) = constraints {
+            constraints.shrink_to_fit();
+            Some(constraints)
+        } else {
+            None
         }
-        self.constraints.shrink_to_fit();
-        self.constraints
     }
 
-    fn evaluate_expression_constraint(&mut self, expression: Expression<'db>, is_positive: bool) {
+    fn evaluate_expression_constraint(
+        &mut self,
+        expression: Expression<'db>,
+        is_positive: bool,
+    ) -> Option<NarrowingConstraints<'db>> {
         let expression_node = expression.node_ref(self.db).node();
-        self.evaluate_expression_node_constraint(expression_node, expression, is_positive);
+        self.evaluate_expression_node_constraint(expression_node, expression, is_positive)
     }
 
     fn evaluate_expression_node_constraint(
@@ -138,52 +191,51 @@ impl<'db> NarrowingConstraintsBuilder<'db> {
         expression_node: &ruff_python_ast::Expr,
         expression: Expression<'db>,
         is_positive: bool,
-    ) {
+    ) -> Option<NarrowingConstraints<'db>> {
         match expression_node {
             ast::Expr::Compare(expr_compare) => {
-                self.add_expr_compare(expr_compare, expression, is_positive);
+                self.evaluate_expr_compare(expr_compare, expression, is_positive)
             }
             ast::Expr::Call(expr_call) => {
-                self.add_expr_call(expr_call, expression, is_positive);
+                self.evaluate_expr_call(expr_call, expression, is_positive)
             }
-            ast::Expr::UnaryOp(unary_op) if unary_op.op == ast::UnaryOp::Not => {
-                self.evaluate_expression_node_constraint(
-                    &unary_op.operand,
-                    expression,
-                    !is_positive,
-                );
-            }
-            _ => {} // TODO other test expression kinds
+            ast::Expr::UnaryOp(unary_op) if unary_op.op == ast::UnaryOp::Not => self
+                .evaluate_expression_node_constraint(&unary_op.operand, expression, !is_positive),
+            ast::Expr::BoolOp(bool_op) => self.evaluate_bool_op(bool_op, expression, is_positive),
+            _ => None, // TODO other test expression kinds
         }
     }
 
-    fn evaluate_pattern_constraint(&mut self, pattern: PatternConstraint<'db>) {
+    fn evaluate_pattern_constraint(
+        &mut self,
+        pattern: PatternConstraint<'db>,
+    ) -> Option<NarrowingConstraints<'db>> {
         let subject = pattern.subject(self.db);
 
         match pattern.pattern(self.db).node() {
             ast::Pattern::MatchValue(_) => {
-                // TODO
+                None // TODO
             }
             ast::Pattern::MatchSingleton(singleton_pattern) => {
-                self.add_match_pattern_singleton(subject, singleton_pattern);
+                self.evaluate_match_pattern_singleton(subject, singleton_pattern)
             }
             ast::Pattern::MatchSequence(_) => {
-                // TODO
+                None // TODO
             }
             ast::Pattern::MatchMapping(_) => {
-                // TODO
+                None // TODO
             }
             ast::Pattern::MatchClass(_) => {
-                // TODO
+                None // TODO
             }
             ast::Pattern::MatchStar(_) => {
-                // TODO
+                None // TODO
             }
             ast::Pattern::MatchAs(_) => {
-                // TODO
+                None // TODO
             }
             ast::Pattern::MatchOr(_) => {
-                // TODO
+                None // TODO
             }
         }
     }
@@ -199,12 +251,12 @@ impl<'db> NarrowingConstraintsBuilder<'db> {
         }
     }
 
-    fn add_expr_compare(
+    fn evaluate_expr_compare(
         &mut self,
         expr_compare: &ast::ExprCompare,
         expression: Expression<'db>,
         is_positive: bool,
-    ) {
+    ) -> Option<NarrowingConstraints<'db>> {
         let ast::ExprCompare {
             range: _,
             left,
@@ -214,14 +266,14 @@ impl<'db> NarrowingConstraintsBuilder<'db> {
         if !left.is_name_expr() && comparators.iter().all(|c| !c.is_name_expr()) {
             // If none of the comparators are name expressions,
             // we have no symbol to narrow down the type of.
-            return;
+            return None;
         }
         if !is_positive && comparators.len() > 1 {
             // We can't negate a constraint made by a multi-comparator expression, since we can't
             // know which comparison part is the one being negated.
             // For example, the negation of  `x is 1 is y is 2`, would be `(x is not 1) or (y is not 1) or (y is not 2)`
             // and that requires cross-symbol constraints, which we don't support yet.
-            return;
+            return None;
         }
         let scope = self.scope();
         let inference = infer_expression_types(self.db, expression);
@@ -229,6 +281,7 @@ impl<'db> NarrowingConstraintsBuilder<'db> {
         let comparator_tuples = std::iter::once(&**left)
             .chain(comparators)
             .tuple_windows::<(&ruff_python_ast::Expr, &ruff_python_ast::Expr)>();
+        let mut constraints = NarrowingConstraints::default();
         for (op, (left, right)) in std::iter::zip(&**ops, comparator_tuples) {
             if let ast::Expr::Name(ast::ExprName {
                 range: _,
@@ -242,24 +295,24 @@ impl<'db> NarrowingConstraintsBuilder<'db> {
 
                 match if is_positive { *op } else { op.negate() } {
                     ast::CmpOp::IsNot => {
-                        if rhs_ty.is_singleton() {
+                        if rhs_ty.is_singleton(self.db) {
                             let ty = IntersectionBuilder::new(self.db)
                                 .add_negative(rhs_ty)
                                 .build();
-                            self.constraints.insert(symbol, ty);
+                            constraints.insert(symbol, ty);
                         } else {
                             // Non-singletons cannot be safely narrowed using `is not`
                         }
                     }
                     ast::CmpOp::Is => {
-                        self.constraints.insert(symbol, rhs_ty);
+                        constraints.insert(symbol, rhs_ty);
                     }
                     ast::CmpOp::NotEq => {
                         if rhs_ty.is_single_valued(self.db) {
                             let ty = IntersectionBuilder::new(self.db)
                                 .add_negative(rhs_ty)
                                 .build();
-                            self.constraints.insert(symbol, ty);
+                            constraints.insert(symbol, ty);
                         }
                     }
                     _ => {
@@ -268,59 +321,137 @@ impl<'db> NarrowingConstraintsBuilder<'db> {
                 }
             }
         }
+        Some(constraints)
     }
 
-    fn add_expr_call(
+    fn evaluate_expr_call(
         &mut self,
         expr_call: &ast::ExprCall,
         expression: Expression<'db>,
         is_positive: bool,
-    ) {
+    ) -> Option<NarrowingConstraints<'db>> {
         let scope = self.scope();
         let inference = infer_expression_types(self.db, expression);
 
-        if let Some(func_type) = inference
+        // TODO: add support for PEP 604 union types on the right hand side of `isinstance`
+        // and `issubclass`, for example `isinstance(x, str | (int | float))`.
+        match inference
             .expression_ty(expr_call.func.scoped_ast_id(self.db, scope))
-            .into_function_literal_type()
+            .into_function_literal()
+            .and_then(|f| f.known(self.db))
+            .and_then(KnownFunction::constraint_function)
         {
-            if func_type.is_known(self.db, KnownFunction::IsInstance)
-                && expr_call.arguments.keywords.is_empty()
-            {
-                if let [ast::Expr::Name(ast::ExprName { id, .. }), rhs] = &*expr_call.arguments.args
+            Some(function) if expr_call.arguments.keywords.is_empty() => {
+                if let [ast::Expr::Name(ast::ExprName { id, .. }), class_info] =
+                    &*expr_call.arguments.args
                 {
                     let symbol = self.symbols().symbol_id_by_name(id).unwrap();
 
-                    let rhs_type = inference.expression_ty(rhs.scoped_ast_id(self.db, scope));
+                    let class_info_ty =
+                        inference.expression_ty(class_info.scoped_ast_id(self.db, scope));
 
-                    // TODO: add support for PEP 604 union types on the right hand side:
-                    // isinstance(x, str | (int | float))
-                    if let Some(mut constraint) = generate_isinstance_constraint(self.db, &rhs_type)
-                    {
-                        if !is_positive {
-                            constraint = constraint.negate(self.db);
+                    let to_constraint = match function {
+                        KnownConstraintFunction::IsInstance => {
+                            |class_literal: ClassLiteralType<'db>| {
+                                Type::Instance(InstanceType {
+                                    class: class_literal.class,
+                                })
+                            }
                         }
-                        self.constraints.insert(symbol, constraint);
-                    }
+                        KnownConstraintFunction::IsSubclass => {
+                            |class_literal: ClassLiteralType<'db>| {
+                                Type::SubclassOf(class_literal.to_subclass_of_type())
+                            }
+                        }
+                    };
+
+                    generate_classinfo_constraint(self.db, &class_info_ty, to_constraint).map(
+                        |constraint| {
+                            let mut constraints = NarrowingConstraints::default();
+                            constraints.insert(symbol, constraint.negate_if(self.db, !is_positive));
+                            constraints
+                        },
+                    )
+                } else {
+                    None
                 }
             }
+            _ => None,
         }
     }
 
-    fn add_match_pattern_singleton(
+    fn evaluate_match_pattern_singleton(
         &mut self,
         subject: &ast::Expr,
         pattern: &ast::PatternMatchSingleton,
-    ) {
+    ) -> Option<NarrowingConstraints<'db>> {
         if let Some(ast::ExprName { id, .. }) = subject.as_name_expr() {
             // SAFETY: we should always have a symbol for every Name node.
             let symbol = self.symbols().symbol_id_by_name(id).unwrap();
 
             let ty = match pattern.value {
-                ast::Singleton::None => Type::None,
+                ast::Singleton::None => Type::none(self.db),
                 ast::Singleton::True => Type::BooleanLiteral(true),
                 ast::Singleton::False => Type::BooleanLiteral(false),
             };
-            self.constraints.insert(symbol, ty);
+            let mut constraints = NarrowingConstraints::default();
+            constraints.insert(symbol, ty);
+            Some(constraints)
+        } else {
+            None
+        }
+    }
+
+    fn evaluate_bool_op(
+        &mut self,
+        expr_bool_op: &ExprBoolOp,
+        expression: Expression<'db>,
+        is_positive: bool,
+    ) -> Option<NarrowingConstraints<'db>> {
+        let inference = infer_expression_types(self.db, expression);
+        let scope = self.scope();
+        let mut sub_constraints = expr_bool_op
+            .values
+            .iter()
+            // filter our arms with statically known truthiness
+            .filter(|expr| {
+                inference
+                    .expression_ty(expr.scoped_ast_id(self.db, scope))
+                    .bool(self.db)
+                    != match expr_bool_op.op {
+                        BoolOp::And => Truthiness::AlwaysTrue,
+                        BoolOp::Or => Truthiness::AlwaysFalse,
+                    }
+            })
+            .map(|sub_expr| {
+                self.evaluate_expression_node_constraint(sub_expr, expression, is_positive)
+            })
+            .collect::<Vec<_>>();
+        match (expr_bool_op.op, is_positive) {
+            (BoolOp::And, true) | (BoolOp::Or, false) => {
+                let mut aggregation: Option<NarrowingConstraints> = None;
+                for sub_constraint in sub_constraints.into_iter().flatten() {
+                    if let Some(ref mut some_aggregation) = aggregation {
+                        merge_constraints_and(some_aggregation, sub_constraint, self.db);
+                    } else {
+                        aggregation = Some(sub_constraint);
+                    }
+                }
+                aggregation
+            }
+            (BoolOp::Or, true) | (BoolOp::And, false) => {
+                let (first, rest) = sub_constraints.split_first_mut()?;
+                if let Some(ref mut first) = first {
+                    for rest_constraint in rest {
+                        if let Some(rest_constraint) = rest_constraint {
+                            merge_constraints_or(first, rest_constraint, self.db);
+                        } else {
+                            return None;
+                        }
+                    }
+                }
+                first.clone()
+            }
         }
     }
 }

crates/red_knot_python_semantic/src/types/unpacker.rs

@@ -0,0 +1,143 @@
+use std::borrow::Cow;
+
+use ruff_db::files::File;
+use ruff_python_ast::{self as ast, AnyNodeRef};
+use rustc_hash::FxHashMap;
+
+use crate::semantic_index::ast_ids::{HasScopedAstId, ScopedExpressionId};
+use crate::semantic_index::symbol::ScopeId;
+use crate::types::{TupleType, Type, TypeCheckDiagnostics, TypeCheckDiagnosticsBuilder};
+use crate::Db;
+
+/// Unpacks the value expression type to their respective targets.
+pub(crate) struct Unpacker<'db> {
+    db: &'db dyn Db,
+    targets: FxHashMap<ScopedExpressionId, Type<'db>>,
+    diagnostics: TypeCheckDiagnosticsBuilder<'db>,
+}
+
+impl<'db> Unpacker<'db> {
+    pub(crate) fn new(db: &'db dyn Db, file: File) -> Self {
+        Self {
+            db,
+            targets: FxHashMap::default(),
+            diagnostics: TypeCheckDiagnosticsBuilder::new(db, file),
+        }
+    }
+
+    pub(crate) fn unpack(&mut self, target: &ast::Expr, value_ty: Type<'db>, scope: ScopeId<'db>) {
+        match target {
+            ast::Expr::Name(target_name) => {
+                self.targets
+                    .insert(target_name.scoped_ast_id(self.db, scope), value_ty);
+            }
+            ast::Expr::Starred(ast::ExprStarred { value, .. }) => {
+                self.unpack(value, value_ty, scope);
+            }
+            ast::Expr::List(ast::ExprList { elts, .. })
+            | ast::Expr::Tuple(ast::ExprTuple { elts, .. }) => match value_ty {
+                Type::Tuple(tuple_ty) => {
+                    let starred_index = elts.iter().position(ast::Expr::is_starred_expr);
+
+                    let element_types = if let Some(starred_index) = starred_index {
+                        if tuple_ty.len(self.db) >= elts.len() - 1 {
+                            let mut element_types = Vec::with_capacity(elts.len());
+                            element_types.extend_from_slice(
+                                // SAFETY: Safe because of the length check above.
+                                &tuple_ty.elements(self.db)[..starred_index],
+                            );
+
+                            // E.g., in `(a, *b, c, d) = ...`, the index of starred element `b`
+                            // is 1 and the remaining elements after that are 2.
+                            let remaining = elts.len() - (starred_index + 1);
+                            // This index represents the type of the last element that belongs
+                            // to the starred expression, in an exclusive manner.
+                            let starred_end_index = tuple_ty.len(self.db) - remaining;
+                            // SAFETY: Safe because of the length check above.
+                            let _starred_element_types =
+                                &tuple_ty.elements(self.db)[starred_index..starred_end_index];
+                            // TODO: Combine the types into a list type. If the
+                            // starred_element_types is empty, then it should be `List[Any]`.
+                            // combine_types(starred_element_types);
+                            element_types.push(Type::Todo);
+
+                            element_types.extend_from_slice(
+                                // SAFETY: Safe because of the length check above.
+                                &tuple_ty.elements(self.db)[starred_end_index..],
+                            );
+                            Cow::Owned(element_types)
+                        } else {
+                            let mut element_types = tuple_ty.elements(self.db).to_vec();
+                            // Subtract 1 to insert the starred expression type at the correct
+                            // index.
+                            element_types.resize(elts.len() - 1, Type::Unknown);
+                            // TODO: This should be `list[Unknown]`
+                            element_types.insert(starred_index, Type::Todo);
+                            Cow::Owned(element_types)
+                        }
+                    } else {
+                        Cow::Borrowed(tuple_ty.elements(self.db).as_ref())
+                    };
+
+                    for (index, element) in elts.iter().enumerate() {
+                        self.unpack(
+                            element,
+                            element_types.get(index).copied().unwrap_or(Type::Unknown),
+                            scope,
+                        );
+                    }
+                }
+                Type::StringLiteral(string_literal_ty) => {
+                    // Deconstruct the string literal to delegate the inference back to the
+                    // tuple type for correct handling of starred expressions. We could go
+                    // further and deconstruct to an array of `StringLiteral` with each
+                    // individual character, instead of just an array of `LiteralString`, but
+                    // there would be a cost and it's not clear that it's worth it.
+                    let value_ty = Type::Tuple(TupleType::new(
+                        self.db,
+                        vec![Type::LiteralString; string_literal_ty.len(self.db)]
+                            .into_boxed_slice(),
+                    ));
+                    self.unpack(target, value_ty, scope);
+                }
+                _ => {
+                    let value_ty = if value_ty.is_literal_string() {
+                        Type::LiteralString
+                    } else {
+                        value_ty
+                            .iterate(self.db)
+                            .unwrap_with_diagnostic(AnyNodeRef::from(target), &mut self.diagnostics)
+                    };
+                    for element in elts {
+                        self.unpack(element, value_ty, scope);
+                    }
+                }
+            },
+            _ => {}
+        }
+    }
+
+    pub(crate) fn finish(mut self) -> UnpackResult<'db> {
+        self.targets.shrink_to_fit();
+        UnpackResult {
+            diagnostics: self.diagnostics.finish(),
+            targets: self.targets,
+        }
+    }
+}
+
+#[derive(Debug, Default, PartialEq, Eq)]
+pub(crate) struct UnpackResult<'db> {
+    targets: FxHashMap<ScopedExpressionId, Type<'db>>,
+    diagnostics: TypeCheckDiagnostics,
+}
+
+impl<'db> UnpackResult<'db> {
+    pub(crate) fn get(&self, expr_id: ScopedExpressionId) -> Option<Type<'db>> {
+        self.targets.get(&expr_id).copied()
+    }
+
+    pub(crate) fn diagnostics(&self) -> &TypeCheckDiagnostics {
+        &self.diagnostics
+    }
+}

crates/red_knot_python_semantic/src/unpack.rs

@@ -0,0 +1,55 @@
+use ruff_db::files::File;
+use ruff_python_ast::{self as ast};
+
+use crate::ast_node_ref::AstNodeRef;
+use crate::semantic_index::expression::Expression;
+use crate::semantic_index::symbol::{FileScopeId, ScopeId};
+use crate::Db;
+
+/// This ingredient represents a single unpacking.
+///
+/// This is required to make use of salsa to cache the complete unpacking of multiple variables
+/// involved. It allows us to:
+/// 1. Avoid doing structural match multiple times for each definition
+/// 2. Avoid highlighting the same error multiple times
+///
+/// ## Module-local type
+/// This type should not be used as part of any cross-module API because
+/// it holds a reference to the AST node. Range-offset changes
+/// then propagate through all usages, and deserialization requires
+/// reparsing the entire module.
+///
+/// E.g. don't use this type in:
+///
+/// * a return type of a cross-module query
+/// * a field of a type that is a return type of a cross-module query
+/// * an argument of a cross-module query
+#[salsa::tracked]
+pub(crate) struct Unpack<'db> {
+    #[id]
+    pub(crate) file: File,
+
+    #[id]
+    pub(crate) file_scope: FileScopeId,
+
+    /// The target expression that is being unpacked. For example, in `(a, b) = (1, 2)`, the target
+    /// expression is `(a, b)`.
+    #[no_eq]
+    #[return_ref]
+    pub(crate) target: AstNodeRef<ast::Expr>,
+
+    /// The ingredient representing the value expression of the unpacking. For example, in
+    /// `(a, b) = (1, 2)`, the value expression is `(1, 2)`.
+    #[no_eq]
+    pub(crate) value: Expression<'db>,
+
+    #[no_eq]
+    count: countme::Count<Unpack<'static>>,
+}
+
+impl<'db> Unpack<'db> {
+    /// Returns the scope where the unpacking is happening.
+    pub(crate) fn scope(self, db: &'db dyn Db) -> ScopeId<'db> {
+        self.file_scope(db).to_scope_id(db, self.file(db))
+    }
+}

crates/red_knot_server/src/edit.rs

@@ -6,7 +6,7 @@ mod text_document;
 
 use lsp_types::{PositionEncodingKind, Url};
 pub use notebook::NotebookDocument;
-pub(crate) use range::RangeExt;
+pub(crate) use range::{RangeExt, ToRangeExt};
 pub(crate) use text_document::DocumentVersion;
 pub use text_document::TextDocument;
 

crates/red_knot_server/src/edit/range.rs

@@ -1,13 +1,32 @@
+use super::notebook;
 use super::PositionEncoding;
-use ruff_source_file::LineIndex;
+use lsp_types as types;
+use ruff_notebook::NotebookIndex;
 use ruff_source_file::OneIndexed;
+use ruff_source_file::{LineIndex, SourceLocation};
 use ruff_text_size::{TextRange, TextSize};
 
+pub(crate) struct NotebookRange {
+    pub(crate) cell: notebook::CellId,
+    pub(crate) range: types::Range,
+}
+
 pub(crate) trait RangeExt {
     fn to_text_range(&self, text: &str, index: &LineIndex, encoding: PositionEncoding)
         -> TextRange;
 }
 
+pub(crate) trait ToRangeExt {
+    fn to_range(&self, text: &str, index: &LineIndex, encoding: PositionEncoding) -> types::Range;
+    fn to_notebook_range(
+        &self,
+        text: &str,
+        source_index: &LineIndex,
+        notebook_index: &NotebookIndex,
+        encoding: PositionEncoding,
+    ) -> NotebookRange;
+}
+
 fn u32_index_to_usize(index: u32) -> usize {
     usize::try_from(index).expect("u32 fits in usize")
 }
@@ -75,6 +94,61 @@ impl RangeExt for lsp_types::Range {
     }
 }
 
+impl ToRangeExt for TextRange {
+    fn to_range(&self, text: &str, index: &LineIndex, encoding: PositionEncoding) -> types::Range {
+        types::Range {
+            start: source_location_to_position(&offset_to_source_location(
+                self.start(),
+                text,
+                index,
+                encoding,
+            )),
+            end: source_location_to_position(&offset_to_source_location(
+                self.end(),
+                text,
+                index,
+                encoding,
+            )),
+        }
+    }
+
+    fn to_notebook_range(
+        &self,
+        text: &str,
+        source_index: &LineIndex,
+        notebook_index: &NotebookIndex,
+        encoding: PositionEncoding,
+    ) -> NotebookRange {
+        let start = offset_to_source_location(self.start(), text, source_index, encoding);
+        let mut end = offset_to_source_location(self.end(), text, source_index, encoding);
+        let starting_cell = notebook_index.cell(start.row);
+
+        // weird edge case here - if the end of the range is where the newline after the cell got added (making it 'out of bounds')
+        // we need to move it one character back (which should place it at the end of the last line).
+        // we test this by checking if the ending offset is in a different (or nonexistent) cell compared to the cell of the starting offset.
+        if notebook_index.cell(end.row) != starting_cell {
+            end.row = end.row.saturating_sub(1);
+            end.column = offset_to_source_location(
+                self.end().checked_sub(1.into()).unwrap_or_default(),
+                text,
+                source_index,
+                encoding,
+            )
+            .column;
+        }
+
+        let start = source_location_to_position(&notebook_index.translate_location(&start));
+        let end = source_location_to_position(&notebook_index.translate_location(&end));
+
+        NotebookRange {
+            cell: starting_cell
+                .map(OneIndexed::to_zero_indexed)
+                .unwrap_or_default(),
+            range: types::Range { start, end },
+        }
+    }
+}
+
 /// Converts a UTF-16 code unit offset for a given line into a UTF-8 column number.
 fn utf8_column_offset(utf16_code_unit_offset: u32, line: &str) -> TextSize {
     let mut utf8_code_unit_offset = TextSize::new(0);
@@ -96,3 +170,46 @@ fn utf8_column_offset(utf16_code_unit_offset: u32, line: &str) -> TextSize {
 
     utf8_code_unit_offset
 }
+
+fn offset_to_source_location(
+    offset: TextSize,
+    text: &str,
+    index: &LineIndex,
+    encoding: PositionEncoding,
+) -> SourceLocation {
+    match encoding {
+        PositionEncoding::UTF8 => {
+            let row = index.line_index(offset);
+            let column = offset - index.line_start(row, text);
+
+            SourceLocation {
+                column: OneIndexed::from_zero_indexed(column.to_usize()),
+                row,
+            }
+        }
+        PositionEncoding::UTF16 => {
+            let row = index.line_index(offset);
+
+            let column = if index.is_ascii() {
+                (offset - index.line_start(row, text)).to_usize()
+            } else {
+                let up_to_line = &text[TextRange::new(index.line_start(row, text), offset)];
+                up_to_line.encode_utf16().count()
+            };
+
+            SourceLocation {
+                column: OneIndexed::from_zero_indexed(column),
+                row,
+            }
+        }
+        PositionEncoding::UTF32 => index.source_location(offset, text),
+    }
+}
+
+fn source_location_to_position(location: &SourceLocation) -> types::Position {
+    types::Position {
+        line: u32::try_from(location.row.to_zero_indexed()).expect("row usize fits in u32"),
+        character: u32::try_from(location.column.to_zero_indexed())
+            .expect("character usize fits in u32"),
+    }
+}

crates/red_knot_server/src/server/api/requests/diagnostic.rs

@@ -3,15 +3,17 @@ use std::borrow::Cow;
 use lsp_types::request::DocumentDiagnosticRequest;
 use lsp_types::{
     Diagnostic, DiagnosticSeverity, DocumentDiagnosticParams, DocumentDiagnosticReport,
-    DocumentDiagnosticReportResult, FullDocumentDiagnosticReport, Position, Range,
+    DocumentDiagnosticReportResult, FullDocumentDiagnosticReport, NumberOrString, Range,
     RelatedFullDocumentDiagnosticReport, Url,
 };
 
-use red_knot_workspace::db::RootDatabase;
-
+use crate::edit::ToRangeExt;
 use crate::server::api::traits::{BackgroundDocumentRequestHandler, RequestHandler};
 use crate::server::{client::Notifier, Result};
 use crate::session::DocumentSnapshot;
+use red_knot_workspace::db::{Db, RootDatabase};
+use ruff_db::diagnostic::Severity;
+use ruff_db::source::{line_index, source_text};
 
 pub(crate) struct DocumentDiagnosticRequestHandler;
 
@@ -64,36 +66,37 @@ fn compute_diagnostics(snapshot: &DocumentSnapshot, db: &RootDatabase) -> Vec<Di
     diagnostics
         .as_slice()
         .iter()
-        .map(|message| to_lsp_diagnostic(message))
+        .map(|message| to_lsp_diagnostic(db, message, snapshot.encoding()))
         .collect()
 }
 
-fn to_lsp_diagnostic(message: &str) -> Diagnostic {
-    let words = message.split(':').collect::<Vec<_>>();
+fn to_lsp_diagnostic(
+    db: &dyn Db,
+    diagnostic: &dyn ruff_db::diagnostic::Diagnostic,
+    encoding: crate::PositionEncoding,
+) -> Diagnostic {
+    let range = if let Some(range) = diagnostic.range() {
+        let index = line_index(db.upcast(), diagnostic.file());
+        let source = source_text(db.upcast(), diagnostic.file());
 
-    let (range, message) = match words.as_slice() {
-        [_, _, line, column, message] | [_, line, column, message] => {
-            let line = line.parse::<u32>().unwrap_or_default().saturating_sub(1);
-            let column = column.parse::<u32>().unwrap_or_default();
-            (
-                Range::new(
-                    Position::new(line, column.saturating_sub(1)),
-                    Position::new(line, column),
-                ),
-                message.trim(),
-            )
-        }
-        _ => (Range::default(), message),
+        range.to_range(&source, &index, encoding)
+    } else {
+        Range::default()
+    };
+
+    let severity = match diagnostic.severity() {
+        Severity::Info => DiagnosticSeverity::INFORMATION,
+        Severity::Error => DiagnosticSeverity::ERROR,
     };
 
     Diagnostic {
         range,
-        severity: Some(DiagnosticSeverity::ERROR),
+        severity: Some(severity),
         tags: None,
-        code: None,
+        code: Some(NumberOrString::String(diagnostic.rule().to_string())),
         code_description: None,
         source: Some("red-knot".into()),
-        message: message.to_string(),
+        message: diagnostic.message().into_owned(),
         related_information: None,
         data: None,
     }

crates/red_knot_test/src/diagnostic.rs

@@ -2,8 +2,8 @@
 //!
 //! We don't assume that we will get the diagnostics in source order.
 
+use ruff_db::diagnostic::Diagnostic;
 use ruff_source_file::{LineIndex, OneIndexed};
-use ruff_text_size::Ranged;
 use std::ops::{Deref, Range};
 
 /// All diagnostics for one embedded Python file, sorted and grouped by start line number.
@@ -19,13 +19,17 @@ pub(crate) struct SortedDiagnostics<T> {
 
 impl<T> SortedDiagnostics<T>
 where
-    T: Ranged + Clone,
+    T: Diagnostic,
 {
     pub(crate) fn new(diagnostics: impl IntoIterator<Item = T>, line_index: &LineIndex) -> Self {
         let mut diagnostics: Vec<_> = diagnostics
             .into_iter()
             .map(|diagnostic| DiagnosticWithLine {
-                line_number: line_index.line_index(diagnostic.start()),
+                line_number: diagnostic
+                    .range()
+                    .map_or(OneIndexed::from_zero_indexed(0), |range| {
+                        line_index.line_index(range.start())
+                    }),
                 diagnostic,
             })
             .collect();
@@ -94,7 +98,7 @@ pub(crate) struct LineDiagnosticsIterator<'a, T> {
 
 impl<'a, T> Iterator for LineDiagnosticsIterator<'a, T>
 where
-    T: Ranged + Clone,
+    T: Diagnostic,
 {
     type Item = LineDiagnostics<'a, T>;
 
@@ -110,7 +114,7 @@ where
     }
 }
 
-impl<T> std::iter::FusedIterator for LineDiagnosticsIterator<'_, T> where T: Clone + Ranged {}
+impl<T> std::iter::FusedIterator for LineDiagnosticsIterator<'_, T> where T: Diagnostic {}
 
 /// All diagnostics that start on a single line of source code in one embedded Python file.
 #[derive(Debug)]
@@ -139,11 +143,14 @@ struct DiagnosticWithLine<T> {
 #[cfg(test)]
 mod tests {
     use crate::db::Db;
-    use ruff_db::files::system_path_to_file;
+    use crate::diagnostic::Diagnostic;
+    use ruff_db::diagnostic::Severity;
+    use ruff_db::files::{system_path_to_file, File};
     use ruff_db::source::line_index;
     use ruff_db::system::{DbWithTestSystem, SystemPathBuf};
     use ruff_source_file::OneIndexed;
     use ruff_text_size::{TextRange, TextSize};
+    use std::borrow::Cow;
 
     #[test]
     fn sort_and_group() {
@@ -152,13 +159,18 @@ mod tests {
         let file = system_path_to_file(&db, "/src/test.py").unwrap();
         let lines = line_index(&db, file);
 
-        let ranges = vec![
+        let ranges = [
             TextRange::new(TextSize::new(0), TextSize::new(1)),
             TextRange::new(TextSize::new(5), TextSize::new(10)),
             TextRange::new(TextSize::new(1), TextSize::new(7)),
         ];
 
-        let sorted = super::SortedDiagnostics::new(&ranges, &lines);
+        let diagnostics: Vec<_> = ranges
+            .into_iter()
+            .map(|range| DummyDiagnostic { range, file })
+            .collect();
+
+        let sorted = super::SortedDiagnostics::new(diagnostics, &lines);
         let grouped = sorted.iter_lines().collect::<Vec<_>>();
 
         let [line1, line2] = &grouped[..] else {
@@ -170,4 +182,32 @@ mod tests {
         assert_eq!(line2.line_number, OneIndexed::from_zero_indexed(1));
         assert_eq!(line2.diagnostics.len(), 1);
     }
+
+    #[derive(Debug)]
+    struct DummyDiagnostic {
+        range: TextRange,
+        file: File,
+    }
+
+    impl Diagnostic for DummyDiagnostic {
+        fn rule(&self) -> &str {
+            "dummy"
+        }
+
+        fn message(&self) -> Cow<str> {
+            "dummy".into()
+        }
+
+        fn file(&self) -> File {
+            self.file
+        }
+
+        fn range(&self) -> Option<TextRange> {
+            Some(self.range)
+        }
+
+        fn severity(&self) -> Severity {
+            Severity::Error
+        }
+    }
 }

crates/red_knot_test/src/lib.rs

@@ -1,6 +1,7 @@
 use colored::Colorize;
 use parser as test_parser;
 use red_knot_python_semantic::types::check_types;
+use ruff_db::diagnostic::{Diagnostic, ParseDiagnostic};
 use ruff_db::files::{system_path_to_file, File, Files};
 use ruff_db::parsed::parsed_module;
 use ruff_db::system::{DbWithTestSystem, SystemPathBuf};
@@ -87,16 +88,24 @@ fn run_test(db: &mut db::Db, test: &parser::MarkdownTest) -> Result<(), Failures
         .filter_map(|test_file| {
             let parsed = parsed_module(db, test_file.file);
 
-            // TODO allow testing against code with syntax errors
-            assert!(
-                parsed.errors().is_empty(),
-                "Python syntax errors in {}, {}: {:?}",
-                test.name(),
-                test_file.file.path(db),
-                parsed.errors()
-            );
+            let mut diagnostics: Vec<Box<_>> = parsed
+                .errors()
+                .iter()
+                .cloned()
+                .map(|error| {
+                    let diagnostic: Box<dyn Diagnostic> =
+                        Box::new(ParseDiagnostic::new(test_file.file, error));
+                    diagnostic
+                })
+                .collect();
 
-            match matcher::match_file(db, test_file.file, check_types(db, test_file.file)) {
+            let type_diagnostics = check_types(db, test_file.file);
+            diagnostics.extend(type_diagnostics.into_iter().map(|diagnostic| {
+                let diagnostic: Box<dyn Diagnostic> = Box::new((*diagnostic).clone());
+                diagnostic
+            }));
+
+            match matcher::match_file(db, test_file.file, diagnostics) {
                 Ok(()) => None,
                 Err(line_failures) => Some(FileFailures {
                     backtick_offset: test_file.backtick_offset,

crates/red_knot_test/src/matcher.rs

@@ -1,17 +1,15 @@
-//! Match [`TypeCheckDiagnostic`]s against [`Assertion`]s and produce test failure messages for any
+//! Match [`Diagnostic`]s against [`Assertion`]s and produce test failure messages for any
 //! mismatches.
 use crate::assertion::{Assertion, ErrorAssertion, InlineFileAssertions};
 use crate::db::Db;
 use crate::diagnostic::SortedDiagnostics;
 use colored::Colorize;
-use red_knot_python_semantic::types::TypeCheckDiagnostic;
+use ruff_db::diagnostic::Diagnostic;
 use ruff_db::files::File;
 use ruff_db::source::{line_index, source_text, SourceText};
 use ruff_source_file::{LineIndex, OneIndexed};
-use ruff_text_size::Ranged;
 use std::cmp::Ordering;
 use std::ops::Range;
-use std::sync::Arc;
 
 #[derive(Debug, Default)]
 pub(super) struct FailuresByLine {
@@ -55,7 +53,7 @@ pub(super) fn match_file<T>(
     diagnostics: impl IntoIterator<Item = T>,
 ) -> Result<(), FailuresByLine>
 where
-    T: Diagnostic + Clone,
+    T: Diagnostic,
 {
     // Parse assertions from comments in the file, and get diagnostics from the file; both
     // ordered by line number.
@@ -126,22 +124,6 @@ where
     }
 }
 
-pub(super) trait Diagnostic: Ranged {
-    fn rule(&self) -> &str;
-
-    fn message(&self) -> &str;
-}
-
-impl Diagnostic for Arc<TypeCheckDiagnostic> {
-    fn rule(&self) -> &str {
-        self.as_ref().rule()
-    }
-
-    fn message(&self) -> &str {
-        self.as_ref().message()
-    }
-}
-
 trait Unmatched {
     fn unmatched(&self) -> String;
 }
@@ -253,10 +235,15 @@ impl Matcher {
         }
     }
 
-    fn column<T: Ranged>(&self, ranged: &T) -> OneIndexed {
-        self.line_index
-            .source_location(ranged.start(), &self.source)
-            .column
+    fn column<T: Diagnostic>(&self, diagnostic: &T) -> OneIndexed {
+        diagnostic
+            .range()
+            .map(|range| {
+                self.line_index
+                    .source_location(range.start(), &self.source)
+                    .column
+            })
+            .unwrap_or(OneIndexed::from_zero_indexed(0))
     }
 
     /// Check if `assertion` matches any [`Diagnostic`]s in `unmatched`.
@@ -323,20 +310,21 @@ impl Matcher {
 #[cfg(test)]
 mod tests {
     use super::FailuresByLine;
-    use ruff_db::files::system_path_to_file;
+    use ruff_db::diagnostic::{Diagnostic, Severity};
+    use ruff_db::files::{system_path_to_file, File};
     use ruff_db::system::{DbWithTestSystem, SystemPathBuf};
     use ruff_python_trivia::textwrap::dedent;
     use ruff_source_file::OneIndexed;
-    use ruff_text_size::{Ranged, TextRange};
+    use ruff_text_size::TextRange;
+    use std::borrow::Cow;
 
-    #[derive(Clone, Debug)]
-    struct TestDiagnostic {
+    struct ExpectedDiagnostic {
         rule: &'static str,
         message: &'static str,
         range: TextRange,
     }
 
-    impl TestDiagnostic {
+    impl ExpectedDiagnostic {
         fn new(rule: &'static str, message: &'static str, offset: usize) -> Self {
             let offset: u32 = offset.try_into().unwrap();
             Self {
@@ -345,32 +333,64 @@ mod tests {
                 range: TextRange::new(offset.into(), (offset + 1).into()),
             }
         }
+
+        fn into_diagnostic(self, file: File) -> TestDiagnostic {
+            TestDiagnostic {
+                rule: self.rule,
+                message: self.message,
+                range: self.range,
+                file,
+            }
+        }
     }
 
-    impl super::Diagnostic for TestDiagnostic {
+    #[derive(Debug)]
+    struct TestDiagnostic {
+        rule: &'static str,
+        message: &'static str,
+        range: TextRange,
+        file: File,
+    }
+
+    impl Diagnostic for TestDiagnostic {
         fn rule(&self) -> &str {
             self.rule
         }
 
-        fn message(&self) -> &str {
-            self.message
+        fn message(&self) -> Cow<str> {
+            self.message.into()
+        }
+
+        fn file(&self) -> File {
+            self.file
+        }
+
+        fn range(&self) -> Option<TextRange> {
+            Some(self.range)
+        }
+
+        fn severity(&self) -> Severity {
+            Severity::Error
         }
     }
 
-    impl Ranged for TestDiagnostic {
-        fn range(&self) -> ruff_text_size::TextRange {
-            self.range
-        }
-    }
-
-    fn get_result(source: &str, diagnostics: Vec<TestDiagnostic>) -> Result<(), FailuresByLine> {
+    fn get_result(
+        source: &str,
+        diagnostics: Vec<ExpectedDiagnostic>,
+    ) -> Result<(), FailuresByLine> {
         colored::control::set_override(false);
 
         let mut db = crate::db::Db::setup(SystemPathBuf::from("/src"));
         db.write_file("/src/test.py", source).unwrap();
         let file = system_path_to_file(&db, "/src/test.py").unwrap();
 
-        super::match_file(&db, file, diagnostics)
+        super::match_file(
+            &db,
+            file,
+            diagnostics
+                .into_iter()
+                .map(|diagnostic| diagnostic.into_diagnostic(file)),
+        )
     }
 
     fn assert_fail(result: Result<(), FailuresByLine>, messages: &[(usize, &[&str])]) {
@@ -403,7 +423,7 @@ mod tests {
     fn revealed_match() {
         let result = get_result(
             "x # revealed: Foo",
-            vec![TestDiagnostic::new(
+            vec![ExpectedDiagnostic::new(
                 "revealed-type",
                 "Revealed type is `Foo`",
                 0,
@@ -417,7 +437,7 @@ mod tests {
     fn revealed_wrong_rule() {
         let result = get_result(
             "x # revealed: Foo",
-            vec![TestDiagnostic::new(
+            vec![ExpectedDiagnostic::new(
                 "not-revealed-type",
                 "Revealed type is `Foo`",
                 0,
@@ -440,7 +460,11 @@ mod tests {
     fn revealed_wrong_message() {
         let result = get_result(
             "x # revealed: Foo",
-            vec![TestDiagnostic::new("revealed-type", "Something else", 0)],
+            vec![ExpectedDiagnostic::new(
+                "revealed-type",
+                "Something else",
+                0,
+            )],
         );
 
         assert_fail(
@@ -467,8 +491,8 @@ mod tests {
         let result = get_result(
             "x # revealed: Foo",
             vec![
-                TestDiagnostic::new("revealed-type", "Revealed type is `Foo`", 0),
-                TestDiagnostic::new("undefined-reveal", "Doesn't matter", 0),
+                ExpectedDiagnostic::new("revealed-type", "Revealed type is `Foo`", 0),
+                ExpectedDiagnostic::new("undefined-reveal", "Doesn't matter", 0),
             ],
         );
 
@@ -479,7 +503,11 @@ mod tests {
     fn revealed_match_with_only_undefined() {
         let result = get_result(
             "x # revealed: Foo",
-            vec![TestDiagnostic::new("undefined-reveal", "Doesn't matter", 0)],
+            vec![ExpectedDiagnostic::new(
+                "undefined-reveal",
+                "Doesn't matter",
+                0,
+            )],
         );
 
         assert_fail(result, &[(0, &["unmatched assertion: revealed: Foo"])]);
@@ -490,8 +518,8 @@ mod tests {
         let result = get_result(
             "x # revealed: Foo",
             vec![
-                TestDiagnostic::new("revealed-type", "Revealed type is `Bar`", 0),
-                TestDiagnostic::new("undefined-reveal", "Doesn't matter", 0),
+                ExpectedDiagnostic::new("revealed-type", "Revealed type is `Bar`", 0),
+                ExpectedDiagnostic::new("undefined-reveal", "Doesn't matter", 0),
             ],
         );
 
@@ -512,8 +540,8 @@ mod tests {
         let result = get_result(
             "reveal_type(1)",
             vec![
-                TestDiagnostic::new("undefined-reveal", "undefined reveal message", 0),
-                TestDiagnostic::new("revealed-type", "Revealed type is `Literal[1]`", 12),
+                ExpectedDiagnostic::new("undefined-reveal", "undefined reveal message", 0),
+                ExpectedDiagnostic::new("revealed-type", "Revealed type is `Literal[1]`", 12),
             ],
         );
 
@@ -535,8 +563,8 @@ mod tests {
         let result = get_result(
             "reveal_type(1) # error: [something-else]",
             vec![
-                TestDiagnostic::new("undefined-reveal", "undefined reveal message", 0),
-                TestDiagnostic::new("revealed-type", "Revealed type is `Literal[1]`", 12),
+                ExpectedDiagnostic::new("undefined-reveal", "undefined reveal message", 0),
+                ExpectedDiagnostic::new("revealed-type", "Revealed type is `Literal[1]`", 12),
             ],
         );
 
@@ -565,7 +593,7 @@ mod tests {
     fn error_match_rule() {
         let result = get_result(
             "x # error: [some-rule]",
-            vec![TestDiagnostic::new("some-rule", "Any message", 0)],
+            vec![ExpectedDiagnostic::new("some-rule", "Any message", 0)],
         );
 
         assert_ok(&result);
@@ -575,7 +603,7 @@ mod tests {
     fn error_wrong_rule() {
         let result = get_result(
             "x # error: [some-rule]",
-            vec![TestDiagnostic::new("anything", "Any message", 0)],
+            vec![ExpectedDiagnostic::new("anything", "Any message", 0)],
         );
 
         assert_fail(
@@ -594,7 +622,11 @@ mod tests {
     fn error_match_message() {
         let result = get_result(
             r#"x # error: "contains this""#,
-            vec![TestDiagnostic::new("anything", "message contains this", 0)],
+            vec![ExpectedDiagnostic::new(
+                "anything",
+                "message contains this",
+                0,
+            )],
         );
 
         assert_ok(&result);
@@ -604,7 +636,7 @@ mod tests {
     fn error_wrong_message() {
         let result = get_result(
             r#"x # error: "contains this""#,
-            vec![TestDiagnostic::new("anything", "Any message", 0)],
+            vec![ExpectedDiagnostic::new("anything", "Any message", 0)],
         );
 
         assert_fail(
@@ -623,7 +655,7 @@ mod tests {
     fn error_match_column_and_rule() {
         let result = get_result(
             "x # error: 1 [some-rule]",
-            vec![TestDiagnostic::new("some-rule", "Any message", 0)],
+            vec![ExpectedDiagnostic::new("some-rule", "Any message", 0)],
         );
 
         assert_ok(&result);
@@ -633,7 +665,7 @@ mod tests {
     fn error_wrong_column() {
         let result = get_result(
             "x # error: 2 [rule]",
-            vec![TestDiagnostic::new("rule", "Any message", 0)],
+            vec![ExpectedDiagnostic::new("rule", "Any message", 0)],
         );
 
         assert_fail(
@@ -652,7 +684,11 @@ mod tests {
     fn error_match_column_and_message() {
         let result = get_result(
             r#"x # error: 1 "contains this""#,
-            vec![TestDiagnostic::new("anything", "message contains this", 0)],
+            vec![ExpectedDiagnostic::new(
+                "anything",
+                "message contains this",
+                0,
+            )],
         );
 
         assert_ok(&result);
@@ -662,7 +698,11 @@ mod tests {
     fn error_match_rule_and_message() {
         let result = get_result(
             r#"x # error: [a-rule] "contains this""#,
-            vec![TestDiagnostic::new("a-rule", "message contains this", 0)],
+            vec![ExpectedDiagnostic::new(
+                "a-rule",
+                "message contains this",
+                0,
+            )],
         );
 
         assert_ok(&result);
@@ -672,7 +712,11 @@ mod tests {
     fn error_match_all() {
         let result = get_result(
             r#"x # error: 1 [a-rule] "contains this""#,
-            vec![TestDiagnostic::new("a-rule", "message contains this", 0)],
+            vec![ExpectedDiagnostic::new(
+                "a-rule",
+                "message contains this",
+                0,
+            )],
         );
 
         assert_ok(&result);
@@ -682,7 +726,11 @@ mod tests {
     fn error_match_all_wrong_column() {
         let result = get_result(
             r#"x # error: 2 [some-rule] "contains this""#,
-            vec![TestDiagnostic::new("some-rule", "message contains this", 0)],
+            vec![ExpectedDiagnostic::new(
+                "some-rule",
+                "message contains this",
+                0,
+            )],
         );
 
         assert_fail(
@@ -701,7 +749,7 @@ mod tests {
     fn error_match_all_wrong_rule() {
         let result = get_result(
             r#"x # error: 1 [some-rule] "contains this""#,
-            vec![TestDiagnostic::new(
+            vec![ExpectedDiagnostic::new(
                 "other-rule",
                 "message contains this",
                 0,
@@ -724,7 +772,7 @@ mod tests {
     fn error_match_all_wrong_message() {
         let result = get_result(
             r#"x # error: 1 [some-rule] "contains this""#,
-            vec![TestDiagnostic::new("some-rule", "Any message", 0)],
+            vec![ExpectedDiagnostic::new("some-rule", "Any message", 0)],
         );
 
         assert_fail(
@@ -757,9 +805,9 @@ mod tests {
         let result = get_result(
             &source,
             vec![
-                TestDiagnostic::new("line-two", "msg", two),
-                TestDiagnostic::new("line-three", "msg", three),
-                TestDiagnostic::new("line-five", "msg", five),
+                ExpectedDiagnostic::new("line-two", "msg", two),
+                ExpectedDiagnostic::new("line-three", "msg", three),
+                ExpectedDiagnostic::new("line-five", "msg", five),
             ],
         );
 
@@ -788,8 +836,8 @@ mod tests {
         let result = get_result(
             &source,
             vec![
-                TestDiagnostic::new("line-one", "msg", one),
-                TestDiagnostic::new("line-two", "msg", two),
+                ExpectedDiagnostic::new("line-one", "msg", one),
+                ExpectedDiagnostic::new("line-two", "msg", two),
             ],
         );
 
@@ -809,8 +857,8 @@ mod tests {
         let result = get_result(
             &source,
             vec![
-                TestDiagnostic::new("one-rule", "msg", x),
-                TestDiagnostic::new("other-rule", "msg", x),
+                ExpectedDiagnostic::new("one-rule", "msg", x),
+                ExpectedDiagnostic::new("other-rule", "msg", x),
             ],
         );
 
@@ -830,8 +878,8 @@ mod tests {
         let result = get_result(
             &source,
             vec![
-                TestDiagnostic::new("one-rule", "msg", x),
-                TestDiagnostic::new("one-rule", "msg", x),
+                ExpectedDiagnostic::new("one-rule", "msg", x),
+                ExpectedDiagnostic::new("one-rule", "msg", x),
             ],
         );
 
@@ -851,9 +899,9 @@ mod tests {
         let result = get_result(
             &source,
             vec![
-                TestDiagnostic::new("one-rule", "msg", x),
-                TestDiagnostic::new("other-rule", "msg", x),
-                TestDiagnostic::new("third-rule", "msg", x),
+                ExpectedDiagnostic::new("one-rule", "msg", x),
+                ExpectedDiagnostic::new("other-rule", "msg", x),
+                ExpectedDiagnostic::new("third-rule", "msg", x),
             ],
         );
 
@@ -877,8 +925,8 @@ mod tests {
         let result = get_result(
             &source,
             vec![
-                TestDiagnostic::new("undefined-reveal", "msg", reveal),
-                TestDiagnostic::new("revealed-type", "Revealed type is `Literal[5]`", reveal),
+                ExpectedDiagnostic::new("undefined-reveal", "msg", reveal),
+                ExpectedDiagnostic::new("revealed-type", "Revealed type is `Literal[5]`", reveal),
             ],
         );
 
@@ -891,7 +939,7 @@ mod tests {
         let x = source.find('x').unwrap();
         let result = get_result(
             source,
-            vec![TestDiagnostic::new("some-rule", "some message", x)],
+            vec![ExpectedDiagnostic::new("some-rule", "some message", x)],
         );
 
         assert_fail(
@@ -912,7 +960,7 @@ mod tests {
         let x = source.find('x').unwrap();
         let result = get_result(
             source,
-            vec![TestDiagnostic::new("some-rule", "some message", x)],
+            vec![ExpectedDiagnostic::new("some-rule", "some message", x)],
         );
 
         assert_fail(

crates/red_knot_wasm/src/lib.rs

@@ -6,6 +6,7 @@ use wasm_bindgen::prelude::*;
 use red_knot_workspace::db::RootDatabase;
 use red_knot_workspace::workspace::settings::Configuration;
 use red_knot_workspace::workspace::WorkspaceMetadata;
+use ruff_db::diagnostic::Diagnostic;
 use ruff_db::files::{system_path_to_file, File};
 use ruff_db::system::walk_directory::WalkDirectoryBuilder;
 use ruff_db::system::{
@@ -110,14 +111,20 @@ impl Workspace {
     pub fn check_file(&self, file_id: &FileHandle) -> Result<Vec<String>, Error> {
         let result = self.db.check_file(file_id.file).map_err(into_error)?;
 
-        Ok(result)
+        Ok(result
+            .into_iter()
+            .map(|diagnostic| diagnostic.display(&self.db).to_string())
+            .collect())
     }
 
     /// Checks all open files
     pub fn check(&self) -> Result<Vec<String>, Error> {
         let result = self.db.check().map_err(into_error)?;
 
-        Ok(result)
+        Ok(result
+            .into_iter()
+            .map(|diagnostic| diagnostic.display(&self.db).to_string())
+            .collect())
     }
 
     /// Returns the parsed AST for `path`

crates/red_knot_wasm/tests/api.rs

@@ -19,6 +19,6 @@ fn check() {
 
     assert_eq!(
         result,
-        vec!["/test.py:1:8: Cannot resolve import `random22`"]
+        vec!["error[unresolved-import] /test.py:1:8 Cannot resolve import `random22`"]
     );
 }

crates/red_knot_workspace/resources/test/corpus/25_func_annotations_starred.py

@@ -0,0 +1,8 @@
+# Regression test for https://github.com/astral-sh/ruff/issues/14115
+#
+# This is invalid syntax, but should not lead to a crash.
+
+def f() -> *int: ...
+
+
+f()

crates/red_knot_workspace/src/db.rs

@@ -4,14 +4,14 @@ use std::sync::Arc;
 use salsa::plumbing::ZalsaDatabase;
 use salsa::{Cancelled, Event};
 
+use crate::workspace::{check_file, Workspace, WorkspaceMetadata};
 use red_knot_python_semantic::{Db as SemanticDb, Program};
+use ruff_db::diagnostic::Diagnostic;
 use ruff_db::files::{File, Files};
 use ruff_db::system::System;
 use ruff_db::vendored::VendoredFileSystem;
 use ruff_db::{Db as SourceDb, Upcast};
 
-use crate::workspace::{check_file, Workspace, WorkspaceMetadata};
-
 mod changes;
 
 #[salsa::db]
@@ -51,11 +51,11 @@ impl RootDatabase {
     }
 
     /// Checks all open files in the workspace and its dependencies.
-    pub fn check(&self) -> Result<Vec<String>, Cancelled> {
+    pub fn check(&self) -> Result<Vec<Box<dyn Diagnostic>>, Cancelled> {
         self.with_db(|db| db.workspace().check(db))
     }
 
-    pub fn check_file(&self, file: File) -> Result<Vec<String>, Cancelled> {
+    pub fn check_file(&self, file: File) -> Result<Vec<Box<dyn Diagnostic>>, Cancelled> {
         let _span = tracing::debug_span!("check_file", file=%file.path(self)).entered();
 
         self.with_db(|db| check_file(db, file))

crates/red_knot_workspace/src/workspace.rs

@@ -1,23 +1,23 @@
-use std::{collections::BTreeMap, sync::Arc};
-
 use rustc_hash::{FxBuildHasher, FxHashSet};
 use salsa::{Durability, Setter as _};
+use std::borrow::Cow;
+use std::{collections::BTreeMap, sync::Arc};
 
+use crate::db::Db;
+use crate::db::RootDatabase;
+use crate::workspace::files::{Index, Indexed, IndexedIter, PackageFiles};
 pub use metadata::{PackageMetadata, WorkspaceMetadata};
 use red_knot_python_semantic::types::check_types;
 use red_knot_python_semantic::SearchPathSettings;
+use ruff_db::diagnostic::{Diagnostic, ParseDiagnostic, Severity};
 use ruff_db::parsed::parsed_module;
-use ruff_db::source::{line_index, source_text, SourceDiagnostic};
+use ruff_db::source::{source_text, SourceTextError};
 use ruff_db::{
     files::{system_path_to_file, File},
     system::{walk_directory::WalkState, SystemPath, SystemPathBuf},
 };
 use ruff_python_ast::{name::Name, PySourceType};
-use ruff_text_size::Ranged;
-
-use crate::db::Db;
-use crate::db::RootDatabase;
-use crate::workspace::files::{Index, Indexed, IndexedIter, PackageFiles};
+use ruff_text_size::TextRange;
 
 mod files;
 mod metadata;
@@ -188,7 +188,7 @@ impl Workspace {
     }
 
     /// Checks all open files in the workspace and its dependencies.
-    pub fn check(self, db: &RootDatabase) -> Vec<String> {
+    pub fn check(self, db: &RootDatabase) -> Vec<Box<dyn Diagnostic>> {
         let workspace_span = tracing::debug_span!("check_workspace");
         let _span = workspace_span.enter();
 
@@ -378,47 +378,31 @@ impl Package {
     }
 }
 
-#[salsa::tracked]
-pub(super) fn check_file(db: &dyn Db, file: File) -> Vec<String> {
-    tracing::debug!("Checking file '{path}'", path = file.path(db));
-
-    let mut diagnostics = Vec::new();
-
-    let source_diagnostics = source_text::accumulated::<SourceDiagnostic>(db.upcast(), file);
-    // TODO(micha): Consider using a single accumulator for all diagnostics
-    diagnostics.extend(
-        source_diagnostics
-            .iter()
-            .map(std::string::ToString::to_string),
-    );
-
+pub(super) fn check_file(db: &dyn Db, file: File) -> Vec<Box<dyn Diagnostic>> {
+    let mut diagnostics: Vec<Box<dyn Diagnostic>> = Vec::new();
     // Abort checking if there are IO errors.
     let source = source_text(db.upcast(), file);
 
-    if source.has_read_error() {
+    if let Some(read_error) = source.read_error() {
+        diagnostics.push(Box::new(IOErrorDiagnostic {
+            file,
+            error: read_error.clone(),
+        }));
         return diagnostics;
     }
 
     let parsed = parsed_module(db.upcast(), file);
+    diagnostics.extend(parsed.errors().iter().map(|error| {
+        let diagnostic: Box<dyn Diagnostic> = Box::new(ParseDiagnostic::new(file, error.clone()));
+        diagnostic
+    }));
 
-    if !parsed.errors().is_empty() {
-        let path = file.path(db);
-        let line_index = line_index(db.upcast(), file);
-        diagnostics.extend(parsed.errors().iter().map(|err| {
-            let source_location = line_index.source_location(err.location.start(), source.as_str());
-            format!("{path}:{source_location}: {message}", message = err.error)
-        }));
-    }
+    diagnostics.extend(check_types(db.upcast(), file).iter().map(|diagnostic| {
+        let boxed: Box<dyn Diagnostic> = Box::new(diagnostic.clone());
+        boxed
+    }));
 
-    for diagnostic in check_types(db.upcast(), file) {
-        let index = line_index(db.upcast(), diagnostic.file());
-        let location = index.source_location(diagnostic.start(), source.as_str());
-        diagnostics.push(format!(
-            "{path}:{location}: {message}",
-            path = file.path(db),
-            message = diagnostic.message()
-        ));
-    }
+    diagnostics.sort_unstable_by_key(|diagnostic| diagnostic.range().unwrap_or_default().start());
 
     diagnostics
 }
@@ -533,17 +517,45 @@ impl Iterator for WorkspaceFilesIter<'_> {
     }
 }
 
+#[derive(Debug)]
+pub struct IOErrorDiagnostic {
+    file: File,
+    error: SourceTextError,
+}
+
+impl Diagnostic for IOErrorDiagnostic {
+    fn rule(&self) -> &str {
+        "io"
+    }
+
+    fn message(&self) -> Cow<str> {
+        self.error.to_string().into()
+    }
+
+    fn file(&self) -> File {
+        self.file
+    }
+
+    fn range(&self) -> Option<TextRange> {
+        None
+    }
+
+    fn severity(&self) -> Severity {
+        Severity::Error
+    }
+}
+
 #[cfg(test)]
 mod tests {
+    use crate::db::tests::TestDb;
+    use crate::workspace::check_file;
     use red_knot_python_semantic::types::check_types;
+    use ruff_db::diagnostic::Diagnostic;
     use ruff_db::files::system_path_to_file;
     use ruff_db::source::source_text;
     use ruff_db::system::{DbWithTestSystem, SystemPath};
     use ruff_db::testing::assert_function_query_was_not_run;
 
-    use crate::db::tests::TestDb;
-    use crate::workspace::check_file;
-
     #[test]
     fn check_file_skips_type_checking_when_file_cant_be_read() -> ruff_db::system::Result<()> {
         let mut db = TestDb::new();
@@ -558,7 +570,10 @@ mod tests {
 
         assert_eq!(source_text(&db, file).as_str(), "");
         assert_eq!(
-            check_file(&db, file),
+            check_file(&db, file)
+                .into_iter()
+                .map(|diagnostic| diagnostic.message().into_owned())
+                .collect::<Vec<_>>(),
             vec!["Failed to read file: No such file or directory".to_string()]
         );
 
@@ -570,7 +585,13 @@ mod tests {
         db.write_file(path, "").unwrap();
 
         assert_eq!(source_text(&db, file).as_str(), "");
-        assert_eq!(check_file(&db, file), vec![] as Vec<String>);
+        assert_eq!(
+            check_file(&db, file)
+                .into_iter()
+                .map(|diagnostic| diagnostic.message().into_owned())
+                .collect::<Vec<_>>(),
+            vec![] as Vec<String>
+        );
 
         Ok(())
     }

crates/ruff/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ruff"
-version = "0.7.2"
+version = "0.7.3"
 publish = true
 authors = { workspace = true }
 edition = { workspace = true }

crates/ruff/tests/snapshots/integration_test__rule_f401.snap

@@ -85,6 +85,7 @@ else:
 
 ## Options
 - `lint.ignore-init-module-imports`
+- `lint.pyflakes.allowed-unused-imports`
 
 ## References
 - [Python documentation: `import`](https://docs.python.org/3/reference/simple_stmts.html#the-import-statement)

crates/ruff_benchmark/benches/red_knot.rs

@@ -8,6 +8,7 @@ use red_knot_workspace::workspace::settings::Configuration;
 use red_knot_workspace::workspace::WorkspaceMetadata;
 use ruff_benchmark::criterion::{criterion_group, criterion_main, BatchSize, Criterion};
 use ruff_benchmark::TestFile;
+use ruff_db::diagnostic::Diagnostic;
 use ruff_db::files::{system_path_to_file, File};
 use ruff_db::source::source_text;
 use ruff_db::system::{MemoryFileSystem, SystemPath, SystemPathBuf, TestSystem};
@@ -24,31 +25,32 @@ const TOMLLIB_312_URL: &str = "https://raw.githubusercontent.com/python/cpython/
 
 static EXPECTED_DIAGNOSTICS: &[&str] = &[
     // We don't support `*` imports yet:
-    "/src/tomllib/_parser.py:7:29: Module `collections.abc` has no member `Iterable`",
+    "error[unresolved-import] /src/tomllib/_parser.py:7:29 Module `collections.abc` has no member `Iterable`",
     // We don't support terminal statements in control flow yet:
-    "/src/tomllib/_parser.py:246:15: Method `__class_getitem__` of type `Literal[frozenset]` is possibly unbound",
-    "/src/tomllib/_parser.py:66:18: Name `s` used when possibly not defined",
-    "/src/tomllib/_parser.py:98:12: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:101:12: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:104:14: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:108:17: Conflicting declared types for `second_char`: Unknown, str | None",
-    "/src/tomllib/_parser.py:115:14: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:126:12: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:267:9: Conflicting declared types for `char`: Unknown, str | None",
-    "/src/tomllib/_parser.py:348:20: Name `nest` used when possibly not defined",
-    "/src/tomllib/_parser.py:353:5: Name `nest` used when possibly not defined",
-    "/src/tomllib/_parser.py:364:9: Conflicting declared types for `char`: Unknown, str | None",
-    "/src/tomllib/_parser.py:381:13: Conflicting declared types for `char`: Unknown, str | None",
-    "/src/tomllib/_parser.py:395:9: Conflicting declared types for `char`: Unknown, str | None",
-    "/src/tomllib/_parser.py:453:24: Name `nest` used when possibly not defined",
-    "/src/tomllib/_parser.py:455:9: Name `nest` used when possibly not defined",
-    "/src/tomllib/_parser.py:482:16: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:566:12: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:573:12: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:579:12: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:580:63: Name `char` used when possibly not defined",
-    "/src/tomllib/_parser.py:590:9: Conflicting declared types for `char`: Unknown, str | None",
-    "/src/tomllib/_parser.py:629:38: Name `datetime_obj` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:66:18 Name `s` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:98:12 Name `char` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:101:12 Name `char` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:104:14 Name `char` used when possibly not defined",
+    "error[conflicting-declarations] /src/tomllib/_parser.py:108:17 Conflicting declared types for `second_char`: Unknown, str | None",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:115:14 Name `char` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:126:12 Name `char` used when possibly not defined",
+    "error[call-possibly-unbound-method] /src/tomllib/_parser.py:246:15 Method `__class_getitem__` of type `Literal[frozenset]` is possibly unbound",
+    "error[conflicting-declarations] /src/tomllib/_parser.py:267:9 Conflicting declared types for `char`: Unknown, str | None",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:348:20 Name `nest` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:353:5 Name `nest` used when possibly not defined",
+    "error[conflicting-declarations] /src/tomllib/_parser.py:364:9 Conflicting declared types for `char`: Unknown, str | None",
+    "error[conflicting-declarations] /src/tomllib/_parser.py:381:13 Conflicting declared types for `char`: Unknown, str | None",
+    "error[conflicting-declarations] /src/tomllib/_parser.py:395:9 Conflicting declared types for `char`: Unknown, str | None",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:453:24 Name `nest` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:455:9 Name `nest` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:482:16 Name `char` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:566:12 Name `char` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:573:12 Name `char` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:579:12 Name `char` used when possibly not defined",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:580:63 Name `char` used when possibly not defined",
+    "error[conflicting-declarations] /src/tomllib/_parser.py:590:9 Conflicting declared types for `char`: Unknown, str | None",
+    "error[possibly-unresolved-reference] /src/tomllib/_parser.py:629:38 Name `datetime_obj` used when possibly not defined",
+    "error[invalid-base] /src/tomllib/_parser.py:692:8354 Invalid class base with type `GenericAlias` (all bases must be a class, `Any`, `Unknown` or `Todo`)",
 ];
 
 fn get_test_file(name: &str) -> TestFile {
@@ -122,7 +124,14 @@ fn setup_rayon() {
 fn benchmark_incremental(criterion: &mut Criterion) {
     fn setup() -> Case {
         let case = setup_case();
-        let result = case.db.check().unwrap();
+
+        let result: Vec<_> = case
+            .db
+            .check()
+            .unwrap()
+            .into_iter()
+            .map(|diagnostic| diagnostic.display(&case.db).to_string())
+            .collect();
 
         assert_eq!(result, EXPECTED_DIAGNOSTICS);
 
@@ -149,7 +158,7 @@ fn benchmark_incremental(criterion: &mut Criterion) {
 
         let result = db.check().unwrap();
 
-        assert_eq!(result, EXPECTED_DIAGNOSTICS);
+        assert_eq!(result.len(), EXPECTED_DIAGNOSTICS.len());
     }
 
     setup_rayon();
@@ -167,7 +176,12 @@ fn benchmark_cold(criterion: &mut Criterion) {
             setup_case,
             |case| {
                 let Case { db, .. } = case;
-                let result = db.check().unwrap();
+                let result: Vec<_> = db
+                    .check()
+                    .unwrap()
+                    .into_iter()
+                    .map(|diagnostic| diagnostic.display(db).to_string())
+                    .collect();
 
                 assert_eq!(result, EXPECTED_DIAGNOSTICS);
             },

crates/ruff_db/src/diagnostic.rs

@@ -0,0 +1,180 @@
+use crate::{
+    files::File,
+    source::{line_index, source_text},
+    Db,
+};
+use ruff_python_parser::ParseError;
+use ruff_text_size::TextRange;
+use std::borrow::Cow;
+
+pub trait Diagnostic: Send + Sync + std::fmt::Debug {
+    fn rule(&self) -> &str;
+
+    fn message(&self) -> std::borrow::Cow<str>;
+
+    fn file(&self) -> File;
+
+    fn range(&self) -> Option<TextRange>;
+
+    fn severity(&self) -> Severity;
+
+    fn display<'a>(&'a self, db: &'a dyn Db) -> DisplayDiagnostic<'a>
+    where
+        Self: Sized,
+    {
+        DisplayDiagnostic {
+            db,
+            diagnostic: self,
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy)]
+pub enum Severity {
+    Info,
+    Error,
+}
+
+pub struct DisplayDiagnostic<'db> {
+    db: &'db dyn Db,
+    diagnostic: &'db dyn Diagnostic,
+}
+
+impl<'db> DisplayDiagnostic<'db> {
+    pub fn new(db: &'db dyn Db, diagnostic: &'db dyn Diagnostic) -> Self {
+        Self { db, diagnostic }
+    }
+}
+
+impl std::fmt::Display for DisplayDiagnostic<'_> {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self.diagnostic.severity() {
+            Severity::Info => f.write_str("info")?,
+            Severity::Error => f.write_str("error")?,
+        }
+
+        write!(
+            f,
+            "[{rule}] {path}",
+            rule = self.diagnostic.rule(),
+            path = self.diagnostic.file().path(self.db)
+        )?;
+
+        if let Some(range) = self.diagnostic.range() {
+            let index = line_index(self.db, self.diagnostic.file());
+            let source = source_text(self.db, self.diagnostic.file());
+
+            let start = index.source_location(range.start(), &source);
+
+            write!(f, ":{line}:{col}", line = start.row, col = start.column)?;
+        }
+
+        write!(f, " {message}", message = self.diagnostic.message())
+    }
+}
+
+impl<T> Diagnostic for Box<T>
+where
+    T: Diagnostic,
+{
+    fn rule(&self) -> &str {
+        (**self).rule()
+    }
+
+    fn message(&self) -> Cow<str> {
+        (**self).message()
+    }
+
+    fn file(&self) -> File {
+        (**self).file()
+    }
+
+    fn range(&self) -> Option<TextRange> {
+        (**self).range()
+    }
+
+    fn severity(&self) -> Severity {
+        (**self).severity()
+    }
+}
+
+impl<T> Diagnostic for std::sync::Arc<T>
+where
+    T: Diagnostic,
+{
+    fn rule(&self) -> &str {
+        (**self).rule()
+    }
+
+    fn message(&self) -> std::borrow::Cow<str> {
+        (**self).message()
+    }
+
+    fn file(&self) -> File {
+        (**self).file()
+    }
+
+    fn range(&self) -> Option<TextRange> {
+        (**self).range()
+    }
+
+    fn severity(&self) -> Severity {
+        (**self).severity()
+    }
+}
+
+impl Diagnostic for Box<dyn Diagnostic> {
+    fn rule(&self) -> &str {
+        (**self).rule()
+    }
+
+    fn message(&self) -> Cow<str> {
+        (**self).message()
+    }
+
+    fn file(&self) -> File {
+        (**self).file()
+    }
+
+    fn range(&self) -> Option<TextRange> {
+        (**self).range()
+    }
+
+    fn severity(&self) -> Severity {
+        (**self).severity()
+    }
+}
+
+#[derive(Debug)]
+pub struct ParseDiagnostic {
+    file: File,
+    error: ParseError,
+}
+
+impl ParseDiagnostic {
+    pub fn new(file: File, error: ParseError) -> Self {
+        Self { file, error }
+    }
+}
+
+impl Diagnostic for ParseDiagnostic {
+    fn rule(&self) -> &str {
+        "invalid-syntax"
+    }
+
+    fn message(&self) -> Cow<str> {
+        self.error.error.to_string().into()
+    }
+
+    fn file(&self) -> File {
+        self.file
+    }
+
+    fn range(&self) -> Option<TextRange> {
+        Some(self.error.location)
+    }
+
+    fn severity(&self) -> Severity {
+        Severity::Error
+    }
+}

crates/ruff_db/src/lib.rs

@@ -6,6 +6,7 @@ use crate::files::Files;
 use crate::system::System;
 use crate::vendored::VendoredFileSystem;
 
+pub mod diagnostic;
 pub mod display;
 pub mod file_revision;
 pub mod files;

crates/ruff_db/src/source.rs

@@ -1,9 +1,7 @@
-use std::fmt::Formatter;
 use std::ops::Deref;
 use std::sync::Arc;
 
 use countme::Count;
-use salsa::Accumulator;
 
 use ruff_notebook::Notebook;
 use ruff_python_ast::PySourceType;
@@ -17,16 +15,14 @@ use crate::Db;
 pub fn source_text(db: &dyn Db, file: File) -> SourceText {
     let path = file.path(db);
     let _span = tracing::trace_span!("source_text", file = %path).entered();
-    let mut has_read_error = false;
+    let mut read_error = None;
 
     let kind = if is_notebook(file.path(db)) {
         file.read_to_notebook(db)
             .unwrap_or_else(|error| {
                 tracing::debug!("Failed to read notebook '{path}': {error}");
 
-                has_read_error = true;
-                SourceDiagnostic(Arc::new(SourceTextError::FailedToReadNotebook(error)))
-                    .accumulate(db);
+                read_error = Some(SourceTextError::FailedToReadNotebook(error.to_string()));
                 Notebook::empty()
             })
             .into()
@@ -35,8 +31,7 @@ pub fn source_text(db: &dyn Db, file: File) -> SourceText {
             .unwrap_or_else(|error| {
                 tracing::debug!("Failed to read file '{path}': {error}");
 
-                has_read_error = true;
-                SourceDiagnostic(Arc::new(SourceTextError::FailedToReadFile(error))).accumulate(db);
+                read_error = Some(SourceTextError::FailedToReadFile(error.to_string()));
                 String::new()
             })
             .into()
@@ -45,7 +40,7 @@ pub fn source_text(db: &dyn Db, file: File) -> SourceText {
     SourceText {
         inner: Arc::new(SourceTextInner {
             kind,
-            has_read_error,
+            read_error,
             count: Count::new(),
         }),
     }
@@ -98,8 +93,8 @@ impl SourceText {
     }
 
     /// Returns `true` if there was an error when reading the content of the file.
-    pub fn has_read_error(&self) -> bool {
-        self.inner.has_read_error
+    pub fn read_error(&self) -> Option<&SourceTextError> {
+        self.inner.read_error.as_ref()
     }
 }
 
@@ -132,7 +127,7 @@ impl std::fmt::Debug for SourceText {
 struct SourceTextInner {
     count: Count<SourceText>,
     kind: SourceTextKind,
-    has_read_error: bool,
+    read_error: Option<SourceTextError>,
 }
 
 #[derive(Eq, PartialEq)]
@@ -153,21 +148,12 @@ impl From<Notebook> for SourceTextKind {
     }
 }
 
-#[salsa::accumulator]
-pub struct SourceDiagnostic(Arc<SourceTextError>);
-
-impl std::fmt::Display for SourceDiagnostic {
-    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
-        std::fmt::Display::fmt(&self.0, f)
-    }
-}
-
-#[derive(Debug, thiserror::Error)]
+#[derive(Debug, thiserror::Error, PartialEq, Eq, Clone)]
 pub enum SourceTextError {
     #[error("Failed to read notebook: {0}`")]
-    FailedToReadNotebook(#[from] ruff_notebook::NotebookError),
+    FailedToReadNotebook(String),
     #[error("Failed to read file: {0}")]
-    FailedToReadFile(#[from] std::io::Error),
+    FailedToReadFile(String),
 }
 
 /// Computes the [`LineIndex`] for `file`.

crates/ruff_linter/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ruff_linter"
-version = "0.7.2"
+version = "0.7.3"
 publish = false
 authors = { workspace = true }
 edition = { workspace = true }

crates/ruff_linter/resources/test/fixtures/flake8_bugbear/B006_1.pyi

@@ -0,0 +1,10 @@
+# we disable this rule for pyi files
+
+def mquantiles(
+    a: _ArrayLikeFloat_co,
+    prob: _ArrayLikeFloat_co = [0.25, 0.5, 0.75],
+    alphap: AnyReal = 0.4,
+    betap: AnyReal = 0.4,
+    axis: CanIndex | None = None,
+    limit: tuple[AnyReal, AnyReal] | tuple[()] = (),
+) -> _MArrayND: ...

crates/ruff_linter/resources/test/fixtures/flake8_bugbear/B033.py

@@ -19,6 +19,7 @@ incorrect_set = {
     1,
     1,
 }
+incorrect_set = {False, 1, 0}
 
 ###
 # Non-errors.

crates/ruff_linter/resources/test/fixtures/flake8_builtins/A002.py

@@ -10,6 +10,8 @@ async def func3(id, dir):
     pass
 
 
+# this is Ok for A002 (trigger A005 instead)
+# https://github.com/astral-sh/ruff/issues/14135
 map([], lambda float: ...)
 
 from typing import override, overload

crates/ruff_linter/resources/test/fixtures/flake8_builtins/A006.py

@@ -3,3 +3,8 @@ lambda x, float, y: x + y
 lambda min, max: min
 lambda id: id
 lambda dir: dir
+
+# Ok for A006 - should trigger A002 instead
+# https://github.com/astral-sh/ruff/issues/14135
+def func1(str, /, type, *complex, Exception, **getattr):
+    pass

crates/ruff_linter/resources/test/fixtures/flake8_pyi/PYI006.py

@@ -7,8 +7,14 @@ if sys.version_info >= (3, 9): ...  # OK
 
 if sys.version_info == (3, 9): ...  # OK
 
-if sys.version_info <= (3, 10): ...  # OK
+if sys.version_info == (3, 9): ...  # Error: PYI006 Use only `<` and `>=` for version info comparisons
 
-if sys.version_info > (3, 10): ...  # OK
+if sys.version_info <= (3, 10): ...  # Error: PYI006 Use only `<` and `>=` for version info comparisons
 
-if python_version > (3, 10): ...  # OK
+if sys.version_info <= (3, 10): ...  # Error: PYI006 Use only `<` and `>=` for version info comparisons
+
+if sys.version_info > (3, 10): ...  # Error: PYI006 Use only `<` and `>=` for version info comparisons
+elif sys.version_info > (3, 11): ...  # Error: PYI006 Use only `<` and `>=` for version info comparisons
+
+if python_version > (3, 10): ...  # Error: PYI006 Use only `<` and `>=` for version info comparisons
+elif python_version == (3, 11): ...  # Error: PYI006 Use only `<` and `>=` for version info comparisons

crates/ruff_linter/resources/test/fixtures/flake8_pyi/PYI021.pyi

@@ -6,6 +6,19 @@ def foo():
 class Bar:
     """bar"""  # ERROR PYI021
 
+class Qux:
+    """qux"""  # ERROR PYI021
+
+    def __init__(self) -> None: ...
+
+class Baz:
+    """Multiline docstring
+
+    Lorem ipsum dolor sit amet
+    """
+
+    def __init__(self) -> None: ...
+
 def bar():
     x = 1
     """foo"""  # OK, not a doc string

crates/ruff_linter/resources/test/fixtures/flake8_simplify/SIM115.py

@@ -256,3 +256,11 @@ dbm.sqlite3.open("foo.db").close()
 # SIM115
 f = dbm.sqlite3.open("foo.db")
 f.close()
+
+# OK
+def func(filepath, encoding):
+    return open(filepath, mode="rt", encoding=encoding)
+
+# OK
+def func(filepath, encoding):
+    return f(open(filepath, mode="rt", encoding=encoding))

crates/ruff_linter/resources/test/fixtures/flake8_simplify/SIM905.py

@@ -0,0 +1,99 @@
+# setup
+sep = ","
+no_sep = None
+
+# positives
+"""
+	itemA
+	itemB
+	itemC
+""".split()
+
+"a,b,c,d".split(",")
+"a,b,c,d".split(None)
+"a,b,c,d".split(",", 1)
+"a,b,c,d".split(None, 1)
+"a,b,c,d".split(sep=",")
+"a,b,c,d".split(sep=None)
+"a,b,c,d".split(sep=",", maxsplit=1)
+"a,b,c,d".split(sep=None, maxsplit=1)
+"a,b,c,d".split(maxsplit=1, sep=",")
+"a,b,c,d".split(maxsplit=1, sep=None)
+"a,b,c,d".split(",", maxsplit=1)
+"a,b,c,d".split(None, maxsplit=1)
+"a,b,c,d".split(maxsplit=1)
+"a,b,c,d".split(maxsplit=1.0)
+"a,b,c,d".split(maxsplit=1)
+"a,b,c,d".split(maxsplit=0)
+"VERB AUX PRON ADP DET".split(" ")
+'   1   2   3   '.split()
+'1<>2<>3<4'.split('<>')
+
+" a*a a*a a ".split("*", -1)  # [' a', 'a a', 'a a ']
+"".split()  # []
+""" 	
+""".split()  # []
+"   	".split()  # []
+"/abc/".split() # ['/abc/']
+("a,b,c"
+# comment
+.split()
+)  # ['a,b,c']
+("a,b,c"
+# comment1
+.split(",")
+) # ['a', 'b', 'c']
+("a,"
+# comment
+"b,"
+"c"
+.split(",")
+) # ['a', 'b', 'c']
+
+"hello "\
+	"world".split()
+# ['hello', 'world']
+
+# prefixes and isc
+u"a b".split()  # ['a', 'b']
+r"a \n b".split()  # ['a', '\\n', 'b']
+("a " "b").split()  # ['a', 'b']
+"a " "b".split()  # ['a', 'b']
+u"a " "b".split()  # ['a', 'b']
+"a " u"b".split()  # ['a', 'b']
+u"a " r"\n".split()  # ['a', '\\n']
+r"\n " u"\n".split()  # ['\\n']
+r"\n " "\n".split()  # ['\\n']
+"a " r"\n".split()  # ['a', '\\n']
+
+"a,b,c".split(',', maxsplit=0) # ['a,b,c']
+"a,b,c".split(',', maxsplit=-1)  # ['a', 'b', 'c']
+"a,b,c".split(',', maxsplit=-2)  # ['a', 'b', 'c']
+"a,b,c".split(',', maxsplit=-0)  # ['a,b,c']
+
+# negatives
+
+# invalid values should not cause panic 
+"a,b,c,d".split(maxsplit="hello")
+"a,b,c,d".split(maxsplit=-"hello")
+
+# variable names not implemented
+"a,b,c,d".split(sep)
+"a,b,c,d".split(no_sep)
+for n in range(3):
+	"a,b,c,d".split(",", maxsplit=n)
+
+# f-strings not yet implemented
+world = "world"
+_ = f"{world}_hello_world".split("_")
+
+hello = "hello_world"
+_ = f"{hello}_world".split("_")
+
+# split on bytes not yet implemented, much less frequent
+b"TesT.WwW.ExamplE.CoM".split(b".")
+
+# str.splitlines not yet implemented
+"hello\nworld".splitlines()
+"hello\nworld".splitlines(keepends=True)
+"hello\nworld".splitlines(keepends=False)

crates/ruff_linter/resources/test/fixtures/pyflakes/F601.py

@@ -58,3 +58,13 @@ x = {
 t={"x":"test123", "x":("test123")}
 
 t={"x":("test123"), "x":"test123"}
+
+# Regression test for: https://github.com/astral-sh/ruff/issues/12772
+x = {
+    1: "abc",
+    1: "def",
+    True: "ghi",
+    0: "foo",
+    0: "bar",
+    False: "baz",
+}

crates/ruff_linter/resources/test/fixtures/pylint/iteration_over_set.py

@@ -60,3 +60,6 @@ for item in {1, 2, 3, 4, 5, 6, int("7")}:  # calls in set literals are fine
 for item in {1, 2, 2}:  # duplicate literals will be ignored
     # B033 catches this
     print(f"I like {item}.")
+
+for item in {False, 0, 0.0, 0j, True, 1, 1.0}:
+    print(item)

crates/ruff_linter/resources/test/fixtures/pyupgrade/UP044.py

@@ -17,3 +17,11 @@ def f(*args: Unpack[other.Type]): pass
 def foo(*args: Unpack[int | str]) -> None: pass
 def foo(*args: Unpack[int and str]) -> None: pass
 def foo(*args: Unpack[int > str]) -> None: pass
+
+# We do not use the shorthand unpacking syntax in the following cases
+from typing import TypedDict
+class KwargsDict(TypedDict):
+    x: int
+    y: int
+
+def foo(name: str, /, **kwargs: Unpack[KwargsDict]) -> None: pass

crates/ruff_linter/resources/test/fixtures/refurb/FURB157.py

@@ -15,3 +15,23 @@ decimal.Decimal("0")
 Decimal(0)
 Decimal("Infinity")
 decimal.Decimal(0)
+
+# Handle Python's Decimal parsing
+# See https://github.com/astral-sh/ruff/issues/13807
+
+# Errors
+Decimal("1_000")
+Decimal("__1____000") 
+
+# Ok
+Decimal("2e-4")
+Decimal("2E-4")
+Decimal("_1.234__")
+Decimal("2e4")
+Decimal("2e+4")
+Decimal("2E4")
+Decimal("1.2") 
+# Ok: even though this is equal to `Decimal(123)`,
+# we assume that a developer would
+# only write it this way if they meant to.
+Decimal("١٢٣") 

crates/ruff_linter/resources/test/fixtures/refurb/FURB189.py

@@ -0,0 +1,42 @@
+# setup
+from enum import Enum, EnumMeta
+from collections import UserList as UL
+
+class SetOnceMappingMixin:
+    __slots__ = ()
+    def __setitem__(self, key, value):
+        if key in self:
+            raise KeyError(str(key) + ' already set')
+        return super().__setitem__(key, value)
+    
+
+class CaseInsensitiveEnumMeta(EnumMeta):
+    pass
+
+# positives
+class D(dict):
+    pass
+
+class L(list):
+    pass
+
+class S(str):
+    pass
+
+# currently not detected
+class SetOnceDict(SetOnceMappingMixin, dict):
+    pass
+
+# negatives
+class C:
+    pass
+
+class I(int):
+    pass
+
+class ActivityState(str, Enum, metaclass=CaseInsensitiveEnumMeta):
+    """Activity state. This is an optional property and if not provided, the state will be Active by
+    default.
+    """
+    ACTIVE = "Active"
+    INACTIVE = "Inactive"

crates/ruff_linter/resources/test/fixtures/ruff/RUF028.py

@@ -87,3 +87,17 @@ def match_case_and_elif():
                 pass
             elif string == "Hello": # fmt: skip
                 pass
+
+
+# Regression test for decorators
+import pytest
+
+@pytest.mark.parametrize(
+    "test_input,expected",
+     [
+          ("3+5",  8 ),
+          ("17+2", 19),
+     ],
+)  # fmt: skip
+def test_eval(test_input, expected):
+    assert eval(test_input) == expected

crates/ruff_linter/resources/test/fixtures/ruff/RUF100_3.py

@@ -23,6 +23,9 @@ print(a)  # noqa: E501, F821 # comment
 print(a)  # noqa: E501, F821  # comment
 print(a)  # noqa: E501, F821 comment
 print(a)  # noqa: E501, F821  comment
+print(a)  # noqa: E501,,F821  comment
+print(a)  # noqa: E501, ,F821  comment
+print(a)  # noqa: E501 F821  comment
 
 print(a)  # comment with unicode µ    # noqa: E501
 print(a)  # comment with unicode µ    # noqa: E501, F821

crates/ruff_linter/src/checkers/ast/analyze/definitions.rs

@@ -155,7 +155,7 @@ pub(crate) fn definitions(checker: &mut Checker) {
 
         // flake8-pyi
         if enforce_stubs {
-            flake8_pyi::rules::docstring_in_stubs(checker, docstring);
+            flake8_pyi::rules::docstring_in_stubs(checker, definition, docstring);
         }
         if enforce_stubs_and_runtime {
             flake8_pyi::rules::iter_method_return_iterable(checker, definition);

crates/ruff_linter/src/checkers/ast/analyze/expression.rs

@@ -388,6 +388,8 @@ pub(crate) fn expression(expr: &Expr, checker: &mut Checker) {
                 Rule::StaticJoinToFString,
                 // refurb
                 Rule::HashlibDigestHex,
+                // flake8-simplify
+                Rule::SplitStaticString,
             ]) {
                 if let Expr::Attribute(ast::ExprAttribute { value, attr, .. }) = func.as_ref() {
                     let attr = attr.as_str();
@@ -405,6 +407,16 @@ pub(crate) fn expression(expr: &Expr, checker: &mut Checker) {
                                     string_value.to_str(),
                                 );
                             }
+                        } else if matches!(attr, "split" | "rsplit") {
+                            // "...".split(...) call
+                            if checker.enabled(Rule::SplitStaticString) {
+                                flake8_simplify::rules::split_static_string(
+                                    checker,
+                                    attr,
+                                    call,
+                                    string_value.to_str(),
+                                );
+                            }
                         } else if attr == "format" {
                             // "...".format(...) call
                             let location = expr.range();

crates/ruff_linter/src/checkers/ast/analyze/statement.rs

@@ -549,6 +549,9 @@ pub(crate) fn statement(stmt: &Stmt, checker: &mut Checker) {
             if checker.enabled(Rule::WhitespaceAfterDecorator) {
                 pycodestyle::rules::whitespace_after_decorator(checker, decorator_list);
             }
+            if checker.enabled(Rule::SubclassBuiltin) {
+                refurb::rules::subclass_builtin(checker, class_def);
+            }
         }
         Stmt::Import(ast::StmtImport { names, range: _ }) => {
             if checker.enabled(Rule::MultipleImportsOnOneLine) {
@@ -1213,8 +1216,18 @@ pub(crate) fn statement(stmt: &Stmt, checker: &mut Checker) {
                         flake8_pyi::rules::unrecognized_platform(checker, test);
                     }
                 }
-                if checker.any_enabled(&[Rule::BadVersionInfoComparison, Rule::BadVersionInfoOrder])
-                {
+                if checker.enabled(Rule::ComplexIfStatementInStub) {
+                    if let Expr::BoolOp(ast::ExprBoolOp { values, .. }) = test.as_ref() {
+                        for value in values {
+                            flake8_pyi::rules::complex_if_statement_in_stub(checker, value);
+                        }
+                    } else {
+                        flake8_pyi::rules::complex_if_statement_in_stub(checker, test);
+                    }
+                }
+            }
+            if checker.any_enabled(&[Rule::BadVersionInfoComparison, Rule::BadVersionInfoOrder]) {
+                if checker.source_type.is_stub() || checker.settings.preview.is_enabled() {
                     fn bad_version_info_comparison(
                         checker: &mut Checker,
                         test: &Expr,
@@ -1247,15 +1260,6 @@ pub(crate) fn statement(stmt: &Stmt, checker: &mut Checker) {
                         }
                     }
                 }
-                if checker.enabled(Rule::ComplexIfStatementInStub) {
-                    if let Expr::BoolOp(ast::ExprBoolOp { values, .. }) = test.as_ref() {
-                        for value in values {
-                            flake8_pyi::rules::complex_if_statement_in_stub(checker, value);
-                        }
-                    } else {
-                        flake8_pyi::rules::complex_if_statement_in_stub(checker, test);
-                    }
-                }
             }
         }
         Stmt::Assert(

crates/ruff_linter/src/checkers/ast/mod.rs

@@ -1535,7 +1535,6 @@ impl<'a> Visitor<'a> for Checker<'a> {
         };
 
         // Step 4: Analysis
-        analyze::expression(expr, self);
         match expr {
             Expr::StringLiteral(string_literal) => {
                 analyze::string_like(string_literal.into(), self);
@@ -1546,6 +1545,7 @@ impl<'a> Visitor<'a> for Checker<'a> {
         }
 
         self.semantic.flags = flags_snapshot;
+        analyze::expression(expr, self);
         self.semantic.pop_node();
     }
 

crates/ruff_linter/src/codes.rs

@@ -480,6 +480,7 @@ pub fn code_to_rule(linter: Linter, code: &str) -> Option<(RuleGroup, Rule)> {
         (Flake8Simplify, "223") => (RuleGroup::Stable, rules::flake8_simplify::rules::ExprAndFalse),
         (Flake8Simplify, "300") => (RuleGroup::Stable, rules::flake8_simplify::rules::YodaConditions),
         (Flake8Simplify, "401") => (RuleGroup::Stable, rules::flake8_simplify::rules::IfElseBlockInsteadOfDictGet),
+        (Flake8Simplify, "905") => (RuleGroup::Preview, rules::flake8_simplify::rules::SplitStaticString),
         (Flake8Simplify, "910") => (RuleGroup::Stable, rules::flake8_simplify::rules::DictGetWithNoneDefault),
         (Flake8Simplify, "911") => (RuleGroup::Stable, rules::flake8_simplify::rules::ZipDictKeysAndValues),
 
@@ -1071,6 +1072,7 @@ pub fn code_to_rule(linter: Linter, code: &str) -> Option<(RuleGroup, Rule)> {
         (Refurb, "181") => (RuleGroup::Stable, rules::refurb::rules::HashlibDigestHex),
         (Refurb, "187") => (RuleGroup::Stable, rules::refurb::rules::ListReverseCopy),
         (Refurb, "188") => (RuleGroup::Preview, rules::refurb::rules::SliceToRemovePrefixOrSuffix),
+        (Refurb, "189") => (RuleGroup::Preview, rules::refurb::rules::SubclassBuiltin),
         (Refurb, "192") => (RuleGroup::Preview, rules::refurb::rules::SortedMinMax),
 
         // flake8-logging

crates/ruff_linter/src/message/snapshots/ruff_linter__message__sarif__tests__results.snap

@@ -81,7 +81,7 @@ expression: value
           "rules": [
             {
               "fullDescription": {
-                "text": "## What it does\nChecks for unused imports.\n\n## Why is this bad?\nUnused imports add a performance overhead at runtime, and risk creating\nimport cycles. They also increase the cognitive load of reading the code.\n\nIf an import statement is used to check for the availability or existence\nof a module, consider using `importlib.util.find_spec` instead.\n\nIf an import statement is used to re-export a symbol as part of a module's\npublic interface, consider using a \"redundant\" import alias, which\ninstructs Ruff (and other tools) to respect the re-export, and avoid\nmarking it as unused, as in:\n\n```python\nfrom module import member as member\n```\n\nAlternatively, you can use `__all__` to declare a symbol as part of the module's\ninterface, as in:\n\n```python\n# __init__.py\nimport some_module\n\n__all__ = [\"some_module\"]\n```\n\n## Fix safety\n\nFixes to remove unused imports are safe, except in `__init__.py` files.\n\nApplying fixes to `__init__.py` files is currently in preview. The fix offered depends on the\ntype of the unused import. Ruff will suggest a safe fix to export first-party imports with\neither a redundant alias or, if already present in the file, an `__all__` entry. If multiple\n`__all__` declarations are present, Ruff will not offer a fix. Ruff will suggest an unsafe fix\nto remove third-party and standard library imports -- the fix is unsafe because the module's\ninterface changes.\n\n## Example\n\n```python\nimport numpy as np  # unused import\n\n\ndef area(radius):\n    return 3.14 * radius**2\n```\n\nUse instead:\n\n```python\ndef area(radius):\n    return 3.14 * radius**2\n```\n\nTo check the availability of a module, use `importlib.util.find_spec`:\n\n```python\nfrom importlib.util import find_spec\n\nif find_spec(\"numpy\") is not None:\n    print(\"numpy is installed\")\nelse:\n    print(\"numpy is not installed\")\n```\n\n## Options\n- `lint.ignore-init-module-imports`\n\n## References\n- [Python documentation: `import`](https://docs.python.org/3/reference/simple_stmts.html#the-import-statement)\n- [Python documentation: `importlib.util.find_spec`](https://docs.python.org/3/library/importlib.html#importlib.util.find_spec)\n- [Typing documentation: interface conventions](https://typing.readthedocs.io/en/latest/source/libraries.html#library-interface-public-and-private-symbols)\n"
+                "text": "## What it does\nChecks for unused imports.\n\n## Why is this bad?\nUnused imports add a performance overhead at runtime, and risk creating\nimport cycles. They also increase the cognitive load of reading the code.\n\nIf an import statement is used to check for the availability or existence\nof a module, consider using `importlib.util.find_spec` instead.\n\nIf an import statement is used to re-export a symbol as part of a module's\npublic interface, consider using a \"redundant\" import alias, which\ninstructs Ruff (and other tools) to respect the re-export, and avoid\nmarking it as unused, as in:\n\n```python\nfrom module import member as member\n```\n\nAlternatively, you can use `__all__` to declare a symbol as part of the module's\ninterface, as in:\n\n```python\n# __init__.py\nimport some_module\n\n__all__ = [\"some_module\"]\n```\n\n## Fix safety\n\nFixes to remove unused imports are safe, except in `__init__.py` files.\n\nApplying fixes to `__init__.py` files is currently in preview. The fix offered depends on the\ntype of the unused import. Ruff will suggest a safe fix to export first-party imports with\neither a redundant alias or, if already present in the file, an `__all__` entry. If multiple\n`__all__` declarations are present, Ruff will not offer a fix. Ruff will suggest an unsafe fix\nto remove third-party and standard library imports -- the fix is unsafe because the module's\ninterface changes.\n\n## Example\n\n```python\nimport numpy as np  # unused import\n\n\ndef area(radius):\n    return 3.14 * radius**2\n```\n\nUse instead:\n\n```python\ndef area(radius):\n    return 3.14 * radius**2\n```\n\nTo check the availability of a module, use `importlib.util.find_spec`:\n\n```python\nfrom importlib.util import find_spec\n\nif find_spec(\"numpy\") is not None:\n    print(\"numpy is installed\")\nelse:\n    print(\"numpy is not installed\")\n```\n\n## Options\n- `lint.ignore-init-module-imports`\n- `lint.pyflakes.allowed-unused-imports`\n\n## References\n- [Python documentation: `import`](https://docs.python.org/3/reference/simple_stmts.html#the-import-statement)\n- [Python documentation: `importlib.util.find_spec`](https://docs.python.org/3/library/importlib.html#importlib.util.find_spec)\n- [Typing documentation: interface conventions](https://typing.readthedocs.io/en/latest/source/libraries.html#library-interface-public-and-private-symbols)\n"
               },
               "help": {
                 "text": "`{name}` imported but unused; consider using `importlib.util.find_spec` to test for availability"

crates/ruff_linter/src/noqa.rs

@@ -183,7 +183,7 @@ impl<'a> Directive<'a> {
         // Extract, e.g., the `401` in `F401`.
         let suffix = line[prefix..]
             .chars()
-            .take_while(char::is_ascii_digit)
+            .take_while(char::is_ascii_alphanumeric)
             .count();
         if prefix > 0 && suffix > 0 {
             Some(&line[..prefix + suffix])
@@ -549,7 +549,7 @@ impl<'a> ParsedFileExemption<'a> {
         // Extract, e.g., the `401` in `F401`.
         let suffix = line[prefix..]
             .chars()
-            .take_while(char::is_ascii_digit)
+            .take_while(char::is_ascii_alphanumeric)
             .count();
         if prefix > 0 && suffix > 0 {
             Some(&line[..prefix + suffix])
@@ -895,7 +895,7 @@ pub(crate) struct NoqaDirectiveLine<'a> {
     pub(crate) directive: Directive<'a>,
     /// The codes that are ignored by the directive.
     pub(crate) matches: Vec<NoqaCode>,
-    // Whether the directive applies to range.end
+    /// Whether the directive applies to `range.end`.
     pub(crate) includes_end: bool,
 }
 
@@ -1191,6 +1191,24 @@ mod tests {
         assert_debug_snapshot!(Directive::try_extract(source, TextSize::default()));
     }
 
+    #[test]
+    fn noqa_squashed_codes() {
+        let source = "# noqa: F401F841";
+        assert_debug_snapshot!(Directive::try_extract(source, TextSize::default()));
+    }
+
+    #[test]
+    fn noqa_empty_comma() {
+        let source = "# noqa: F401,,F841";
+        assert_debug_snapshot!(Directive::try_extract(source, TextSize::default()));
+    }
+
+    #[test]
+    fn noqa_empty_comma_space() {
+        let source = "# noqa: F401, ,F841";
+        assert_debug_snapshot!(Directive::try_extract(source, TextSize::default()));
+    }
+
     #[test]
     fn noqa_invalid_suffix() {
         let source = "# noqa[F401]";

crates/ruff_linter/src/rules/eradicate/detection.rs

@@ -16,8 +16,43 @@ static CODE_INDICATORS: LazyLock<AhoCorasick> = LazyLock::new(|| {
 
 static ALLOWLIST_REGEX: LazyLock<Regex> = LazyLock::new(|| {
     Regex::new(
-        r"^(?i)(?:pylint|pyright|noqa|nosec|region|endregion|type:\s*ignore|fmt:\s*(on|off)|isort:\s*(on|off|skip|skip_file|split|dont-add-imports(:\s*\[.*?])?)|mypy:|SPDX-License-Identifier:|(?:en)?coding[:=][ \t]*([-_.a-zA-Z0-9]+))",
-    ).unwrap()
+        r"(?x)
+        ^
+        (?:
+            # Case-sensitive
+            pyright
+        |   mypy:
+        |   type:\s*ignore
+        |   SPDX-License-Identifier:
+        |   fmt:\s*(on|off|skip)
+        |   region|endregion
+
+            # Case-insensitive
+        |   (?i:
+                noqa
+            )
+
+            # Unknown case sensitivity
+        |   (?i:
+                pylint
+            |   nosec
+            |   isort:\s*(on|off|skip|skip_file|split|dont-add-imports(:\s*\[.*?])?)
+            |   (?:en)?coding[:=][\x20\t]*([-_.A-Z0-9]+)
+            )
+
+            # IntelliJ language injection comments:
+            # * `language` must be lowercase.
+            # * No spaces around `=`.
+            # * Language IDs as used in comments must have no spaces,
+            #   though to IntelliJ they can be anything.
+            # * May optionally contain `prefix=` and/or `suffix=`,
+            #   not declared here since we use `.is_match()`.
+        |   language=[-_.a-zA-Z0-9]+
+
+        )
+        ",
+    )
+    .unwrap()
 });
 
 static HASH_NUMBER: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"#\d").unwrap());
@@ -297,6 +332,48 @@ mod tests {
         ));
     }
 
+    #[test]
+    fn comment_contains_language_injection() {
+        // `language` with bad casing
+        assert!(comment_contains_code("# Language=C#", &[]));
+        assert!(comment_contains_code("# lAngUAgE=inI", &[]));
+
+        // Unreasonable language IDs, possibly literals
+        assert!(comment_contains_code("# language=\"pt\"", &[]));
+        assert!(comment_contains_code("# language='en'", &[]));
+
+        // Spaces around equal sign
+        assert!(comment_contains_code("# language =xml", &[]));
+        assert!(comment_contains_code("# language= html", &[]));
+        assert!(comment_contains_code("# language = RegExp", &[]));
+
+        // Leading whitespace
+        assert!(!comment_contains_code("#language=CSS", &[]));
+        assert!(!comment_contains_code("#   \t language=C++", &[]));
+
+        // Human language false negatives
+        assert!(!comment_contains_code("# language=en", &[]));
+        assert!(!comment_contains_code("# language=en-US", &[]));
+
+        // Casing (fine because such IDs cannot be validated)
+        assert!(!comment_contains_code("# language=PytHoN", &[]));
+        assert!(!comment_contains_code("# language=jaVaScrIpt", &[]));
+
+        // Space within ID (fine because `Shell` is considered the ID)
+        assert!(!comment_contains_code("#    language=Shell Script", &[]));
+
+        // With prefix and/or suffix
+        assert!(!comment_contains_code("# language=HTML prefix=<body>", &[]));
+        assert!(!comment_contains_code(
+            r"# language=Requirements suffix=\n",
+            &[]
+        ));
+        assert!(!comment_contains_code(
+            "language=javascript prefix=(function(){ suffix=})()",
+            &[]
+        ));
+    }
+
     #[test]
     fn comment_contains_todo() {
         let task_tags = TASK_TAGS

crates/ruff_linter/src/rules/eradicate/rules/commented_out_code.rs

@@ -37,11 +37,11 @@ impl Violation for CommentedOutCode {
 
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("Found commented-out code")
+        "Found commented-out code".to_string()
     }
 
     fn fix_title(&self) -> Option<String> {
-        Some(format!("Remove commented-out code"))
+        Some("Remove commented-out code".to_string())
     }
 }
 

crates/ruff_linter/src/rules/fastapi/rules/fastapi_non_annotated_dependency.rs

@@ -64,7 +64,7 @@ impl Violation for FastApiNonAnnotatedDependency {
 
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("FastAPI dependency without `Annotated`")
+        "FastAPI dependency without `Annotated`".to_string()
     }
 
     fn fix_title(&self) -> Option<String> {

crates/ruff_linter/src/rules/fastapi/rules/fastapi_redundant_response_model.rs

@@ -65,7 +65,7 @@ pub struct FastApiRedundantResponseModel;
 impl AlwaysFixableViolation for FastApiRedundantResponseModel {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("FastAPI route with redundant `response_model` argument")
+        "FastAPI route with redundant `response_model` argument".to_string()
     }
 
     fn fix_title(&self) -> String {

crates/ruff_linter/src/rules/flake8_2020/rules/compare.rs

@@ -46,7 +46,7 @@ pub struct SysVersionCmpStr3;
 impl Violation for SysVersionCmpStr3 {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("`sys.version` compared to string (python3.10), use `sys.version_info`")
+        "`sys.version` compared to string (python3.10), use `sys.version_info`".to_string()
     }
 }
 
@@ -93,7 +93,7 @@ pub struct SysVersionInfo0Eq3;
 impl Violation for SysVersionInfo0Eq3 {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("`sys.version_info[0] == 3` referenced (python4), use `>=`")
+        "`sys.version_info[0] == 3` referenced (python4), use `>=`".to_string()
     }
 }
 
@@ -133,10 +133,9 @@ pub struct SysVersionInfo1CmpInt;
 impl Violation for SysVersionInfo1CmpInt {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!(
-            "`sys.version_info[1]` compared to integer (python4), compare `sys.version_info` to \
+        "`sys.version_info[1]` compared to integer (python4), compare `sys.version_info` to \
              tuple"
-        )
+            .to_string()
     }
 }
 
@@ -176,10 +175,9 @@ pub struct SysVersionInfoMinorCmpInt;
 impl Violation for SysVersionInfoMinorCmpInt {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!(
-            "`sys.version_info.minor` compared to integer (python4), compare `sys.version_info` \
+        "`sys.version_info.minor` compared to integer (python4), compare `sys.version_info` \
              to tuple"
-        )
+            .to_string()
     }
 }
 
@@ -220,7 +218,7 @@ pub struct SysVersionCmpStr10;
 impl Violation for SysVersionCmpStr10 {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("`sys.version` compared to string (python10), use `sys.version_info`")
+        "`sys.version` compared to string (python10), use `sys.version_info`".to_string()
     }
 }
 

crates/ruff_linter/src/rules/flake8_2020/rules/name_or_attribute.rs

@@ -41,7 +41,7 @@ pub struct SixPY3;
 impl Violation for SixPY3 {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("`six.PY3` referenced (python4), use `not six.PY2`")
+        "`six.PY3` referenced (python4), use `not six.PY2`".to_string()
     }
 }
 

crates/ruff_linter/src/rules/flake8_2020/rules/subscript.rs

@@ -43,7 +43,7 @@ pub struct SysVersionSlice3;
 impl Violation for SysVersionSlice3 {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("`sys.version[:3]` referenced (python3.10), use `sys.version_info`")
+        "`sys.version[:3]` referenced (python3.10), use `sys.version_info`".to_string()
     }
 }
 
@@ -83,7 +83,7 @@ pub struct SysVersion2;
 impl Violation for SysVersion2 {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("`sys.version[2]` referenced (python3.10), use `sys.version_info`")
+        "`sys.version[2]` referenced (python3.10), use `sys.version_info`".to_string()
     }
 }
 
@@ -123,7 +123,7 @@ pub struct SysVersion0;
 impl Violation for SysVersion0 {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("`sys.version[0]` referenced (python10), use `sys.version_info`")
+        "`sys.version[0]` referenced (python10), use `sys.version_info`".to_string()
     }
 }
 
@@ -163,7 +163,7 @@ pub struct SysVersionSlice1;
 impl Violation for SysVersionSlice1 {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("`sys.version[:1]` referenced (python10), use `sys.version_info`")
+        "`sys.version[:1]` referenced (python10), use `sys.version_info`".to_string()
     }
 }
 

crates/ruff_linter/src/rules/flake8_annotations/rules/definition.rs

@@ -229,12 +229,11 @@ impl Violation for MissingReturnTypeUndocumentedPublicFunction {
     }
 
     fn fix_title(&self) -> Option<String> {
-        let Self { annotation, .. } = self;
-        if let Some(annotation) = annotation {
-            Some(format!("Add return type annotation: `{annotation}`"))
-        } else {
-            Some(format!("Add return type annotation"))
-        }
+        let title = match &self.annotation {
+            Some(annotation) => format!("Add return type annotation: `{annotation}`"),
+            None => "Add return type annotation".to_string(),
+        };
+        Some(title)
     }
 }
 
@@ -273,12 +272,11 @@ impl Violation for MissingReturnTypePrivateFunction {
     }
 
     fn fix_title(&self) -> Option<String> {
-        let Self { annotation, .. } = self;
-        if let Some(annotation) = annotation {
-            Some(format!("Add return type annotation: `{annotation}`"))
-        } else {
-            Some(format!("Add return type annotation"))
-        }
+        let title = match &self.annotation {
+            Some(annotation) => format!("Add return type annotation: `{annotation}`"),
+            None => "Add return type annotation".to_string(),
+        };
+        Some(title)
     }
 }
 
@@ -330,12 +328,11 @@ impl Violation for MissingReturnTypeSpecialMethod {
     }
 
     fn fix_title(&self) -> Option<String> {
-        let Self { annotation, .. } = self;
-        if let Some(annotation) = annotation {
-            Some(format!("Add return type annotation: `{annotation}`"))
-        } else {
-            Some(format!("Add return type annotation"))
-        }
+        let title = match &self.annotation {
+            Some(annotation) => format!("Add return type annotation: `{annotation}`"),
+            None => "Add return type annotation".to_string(),
+        };
+        Some(title)
     }
 }
 
@@ -378,12 +375,11 @@ impl Violation for MissingReturnTypeStaticMethod {
     }
 
     fn fix_title(&self) -> Option<String> {
-        let Self { annotation, .. } = self;
-        if let Some(annotation) = annotation {
-            Some(format!("Add return type annotation: `{annotation}`"))
-        } else {
-            Some(format!("Add return type annotation"))
-        }
+        let title = match &self.annotation {
+            Some(annotation) => format!("Add return type annotation: `{annotation}`"),
+            None => "Add return type annotation".to_string(),
+        };
+        Some(title)
     }
 }
 
@@ -426,12 +422,11 @@ impl Violation for MissingReturnTypeClassMethod {
     }
 
     fn fix_title(&self) -> Option<String> {
-        let Self { annotation, .. } = self;
-        if let Some(annotation) = annotation {
-            Some(format!("Add return type annotation: `{annotation}`"))
-        } else {
-            Some(format!("Add return type annotation"))
-        }
+        let title = match &self.annotation {
+            Some(annotation) => format!("Add return type annotation: `{annotation}`"),
+            None => "Add return type annotation".to_string(),
+        };
+        Some(title)
     }
 }
 
@@ -474,7 +469,7 @@ impl Violation for MissingReturnTypeClassMethod {
 /// ```
 ///
 /// ## References
-/// - [PEP 484](https://www.python.org/dev/peps/pep-0484/#the-any-type)
+/// - [Typing spec: `Any`](https://typing.readthedocs.io/en/latest/spec/special-types.html#any)
 /// - [Python documentation: `typing.Any`](https://docs.python.org/3/library/typing.html#typing.Any)
 /// - [Mypy documentation: The Any type](https://mypy.readthedocs.io/en/stable/kinds_of_types.html#the-any-type)
 #[violation]

crates/ruff_linter/src/rules/flake8_async/rules/async_function_with_timeout.rs

@@ -71,7 +71,7 @@ pub struct AsyncFunctionWithTimeout {
 impl Violation for AsyncFunctionWithTimeout {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("Async function definition with a `timeout` parameter")
+        "Async function definition with a `timeout` parameter".to_string()
     }
 
     fn fix_title(&self) -> Option<String> {

crates/ruff_linter/src/rules/flake8_async/rules/blocking_http_call.rs

@@ -37,7 +37,7 @@ pub struct BlockingHttpCallInAsyncFunction;
 impl Violation for BlockingHttpCallInAsyncFunction {
     #[derive_message_formats]
     fn message(&self) -> String {
-        format!("Async functions should not call blocking HTTP methods")
+        "Async functions should not call blocking HTTP methods".to_string()
     }
 }