Remove inner abstraction layer from string nodes

## Summary

Part of #7595 

This PR moves the `RUF001` and `RUF002` rules to the AST checker. This
removes the use of docstring detection from these rules.

## Test Plan

As this is just a refactor, make sure existing test cases pass.

.config/nextest.toml

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

.github/workflows/ci.yaml

@@ -111,13 +111,23 @@ jobs:
       - uses: actions/checkout@v4
       - name: "Install Rust toolchain"
         run: rustup show
+      - name: "Install mold"
+        uses: rui314/setup-mold@v1
+      - name: "Install cargo nextest"
+        uses: taiki-e/install-action@v2
+        with:
+          tool: cargo-nextest
       - name: "Install cargo insta"
         uses: taiki-e/install-action@v2
         with:
           tool: cargo-insta
       - uses: Swatinem/rust-cache@v2
       - name: "Run tests"
-        run: cargo insta test --all --all-features --unreferenced reject
+        shell: bash
+        env:
+          NEXTEST_PROFILE: "ci"
+        run: cargo insta test --all-features --unreferenced reject --test-runner nextest
+
       # Check for broken links in the documentation.
       - run: cargo doc --all --no-deps
         env:
@@ -138,15 +148,16 @@ jobs:
       - uses: actions/checkout@v4
       - name: "Install Rust toolchain"
         run: rustup show
-      - name: "Install cargo insta"
+      - name: "Install cargo nextest"
         uses: taiki-e/install-action@v2
         with:
-          tool: cargo-insta
+          tool: cargo-nextest
       - uses: Swatinem/rust-cache@v2
       - name: "Run tests"
         shell: bash
-        # We can't reject unreferenced snapshots on windows because flake8_executable can't run on windows
-        run: cargo insta test --all --exclude ruff_dev --all-features
+        run: |
+          cargo nextest run --all-features --profile ci
+          cargo test --all-features --doc
 
   cargo-test-wasm:
     name: "cargo test (wasm)"
@@ -407,7 +418,7 @@ jobs:
       - uses: actions/setup-python@v5
       - name: "Add SSH key"
         if: ${{ env.MKDOCS_INSIDERS_SSH_KEY_EXISTS == 'true' }}
-        uses: webfactory/ssh-agent@v0.8.0
+        uses: webfactory/ssh-agent@v0.9.0
         with:
           ssh-private-key: ${{ secrets.MKDOCS_INSIDERS_SSH_KEY }}
       - name: "Install Rust toolchain"

.github/workflows/docs.yaml

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

CHANGELOG.md

@@ -1,5 +1,66 @@
 # Changelog
 
+## 0.2.2
+
+Highlights include:
+
+- Initial support formatting f-strings (in `--preview`).
+- Support for overriding arbitrary configuration options via the CLI through an expanded `--config`
+    argument (e.g., `--config "lint.isort.combine-as-imports=false"`).
+- Significant performance improvements in Ruff's lexer, parser, and lint rules.
+
+### Preview features
+
+- Implement minimal f-string formatting ([#9642](https://github.com/astral-sh/ruff/pull/9642))
+- \[`pycodestyle`\] Add blank line(s) rules (`E301`, `E302`, `E303`, `E304`, `E305`, `E306`) ([#9266](https://github.com/astral-sh/ruff/pull/9266))
+- \[`refurb`\] Implement `readlines_in_for` (`FURB129`) ([#9880](https://github.com/astral-sh/ruff/pull/9880))
+
+### Rule changes
+
+- \[`ruff`\] Ensure closing parentheses for multiline sequences are always on their own line (`RUF022`, `RUF023`) ([#9793](https://github.com/astral-sh/ruff/pull/9793))
+- \[`numpy`\] Add missing deprecation violations (`NPY002`) ([#9862](https://github.com/astral-sh/ruff/pull/9862))
+- \[`flake8-bandit`\] Detect `mark_safe` usages in decorators ([#9887](https://github.com/astral-sh/ruff/pull/9887))
+- \[`ruff`\] Expand `asyncio-dangling-task` (`RUF006`) to include `new_event_loop` ([#9976](https://github.com/astral-sh/ruff/pull/9976))
+- \[`flake8-pyi`\] Ignore 'unused' private type dicts in class scopes ([#9952](https://github.com/astral-sh/ruff/pull/9952))
+
+### Formatter
+
+- Docstring formatting: Preserve tab indentation when using `indent-style=tabs` ([#9915](https://github.com/astral-sh/ruff/pull/9915))
+- Disable top-level docstring formatting for notebooks ([#9957](https://github.com/astral-sh/ruff/pull/9957))
+- Stabilize quote-style's `preserve` mode ([#9922](https://github.com/astral-sh/ruff/pull/9922))
+
+### CLI
+
+- Allow arbitrary configuration options to be overridden via the CLI ([#9599](https://github.com/astral-sh/ruff/pull/9599))
+
+### Bug fixes
+
+- Make `show-settings` filters directory-agnostic ([#9866](https://github.com/astral-sh/ruff/pull/9866))
+- Respect duplicates when rewriting type aliases ([#9905](https://github.com/astral-sh/ruff/pull/9905))
+- Respect tuple assignments in typing analyzer ([#9969](https://github.com/astral-sh/ruff/pull/9969))
+- Use atomic write when persisting cache ([#9981](https://github.com/astral-sh/ruff/pull/9981))
+- Use non-parenthesized range for `DebugText` ([#9953](https://github.com/astral-sh/ruff/pull/9953))
+- \[`flake8-simplify`\] Avoid false positive with `async` for loops (`SIM113`) ([#9996](https://github.com/astral-sh/ruff/pull/9996))
+- \[`flake8-trio`\] Respect `async with` in `timeout-without-await` ([#9859](https://github.com/astral-sh/ruff/pull/9859))
+- \[`perflint`\] Catch a wider range of mutations in `PERF101` ([#9955](https://github.com/astral-sh/ruff/pull/9955))
+- \[`pycodestyle`\] Fix `E30X` panics on blank lines with trailing white spaces ([#9907](https://github.com/astral-sh/ruff/pull/9907))
+- \[`pydocstyle`\] Allow using `parameters` as a subsection header (`D405`) ([#9894](https://github.com/astral-sh/ruff/pull/9894))
+- \[`pydocstyle`\] Fix blank-line docstring rules for module-level docstrings ([#9878](https://github.com/astral-sh/ruff/pull/9878))
+- \[`pylint`\] Accept 0.0 and 1.0 as common magic values (`PLR2004`) ([#9964](https://github.com/astral-sh/ruff/pull/9964))
+- \[`pylint`\] Avoid suggesting set rewrites for non-hashable types ([#9956](https://github.com/astral-sh/ruff/pull/9956))
+- \[`ruff`\] Avoid false negatives with string literals inside of method calls (`RUF027`) ([#9865](https://github.com/astral-sh/ruff/pull/9865))
+- \[`ruff`\] Fix panic on with f-string detection (`RUF027`) ([#9990](https://github.com/astral-sh/ruff/pull/9990))
+- \[`ruff`\] Ignore builtins when detecting missing f-strings ([#9849](https://github.com/astral-sh/ruff/pull/9849))
+
+### Performance
+
+- Use `memchr` for string lexing ([#9888](https://github.com/astral-sh/ruff/pull/9888))
+- Use `memchr` for tab-indentation detection ([#9853](https://github.com/astral-sh/ruff/pull/9853))
+- Reduce `Result<Tok, LexicalError>` size by using `Box<str>` instead of `String` ([#9885](https://github.com/astral-sh/ruff/pull/9885))
+- Reduce size of `Expr` from 80 to 64 bytes ([#9900](https://github.com/astral-sh/ruff/pull/9900))
+- Improve trailing comma rule performance ([#9867](https://github.com/astral-sh/ruff/pull/9867))
+- Remove unnecessary string cloning from the parser ([#9884](https://github.com/astral-sh/ruff/pull/9884))
+
 ## 0.2.1
 
 This release includes support for range formatting (i.e., the ability to format specific lines

CONTRIBUTING.md

@@ -26,6 +26,10 @@ Welcome! We're happy to have you here. Thank you in advance for your contributio
 - [`cargo dev`](#cargo-dev)
 - [Subsystems](#subsystems)
     - [Compilation Pipeline](#compilation-pipeline)
+    - [Import Categorization](#import-categorization)
+        - [Project root](#project-root)
+        - [Package root](#package-root)
+        - [Import categorization](#import-categorization-1)
 
 ## The Basics
 
@@ -35,7 +39,7 @@ For small changes (e.g., bug fixes), feel free to submit a PR.
 
 For larger changes (e.g., new lint rules, new functionality, new configuration options), consider
 creating an [**issue**](https://github.com/astral-sh/ruff/issues) outlining your proposed change.
-You can also join us on [**Discord**](https://discord.gg/c9MhzV8aU5) to discuss your idea with the
+You can also join us on [**Discord**](https://discord.com/invite/astral-sh) to discuss your idea with the
 community. We've labeled [beginner-friendly tasks](https://github.com/astral-sh/ruff/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22)
 in the issue tracker, along with [bugs](https://github.com/astral-sh/ruff/issues?q=is%3Aissue+is%3Aopen+label%3Abug)
 and [improvements](https://github.com/astral-sh/ruff/issues?q=is%3Aissue+is%3Aopen+label%3Aaccepted)
@@ -63,7 +67,7 @@ You'll also need [Insta](https://insta.rs/docs/) to update snapshot tests:
 cargo install cargo-insta
 ```
 
-and pre-commit to run some validation checks:
+And you'll need pre-commit to run some validation checks:
 
 ```shell
 pipx install pre-commit  # or `pip install pre-commit` if you have a virtualenv
@@ -76,6 +80,16 @@ when making a commit:
 pre-commit install
 ```
 
+We recommend [nextest](https://nexte.st/) to run Ruff's test suite (via `cargo nextest run`),
+though it's not strictly necessary:
+
+```shell
+cargo install cargo-nextest --locked
+```
+
+Throughout this guide, any usages of `cargo test` can be replaced with `cargo nextest run`,
+if you choose to install `nextest`.
+
 ### Development
 
 After cloning the repository, run Ruff locally from the repository root with:
@@ -373,6 +387,11 @@ We have several ways of benchmarking and profiling Ruff:
 - Microbenchmarks which run the linter or the formatter on individual files. These run on pull requests.
 - Profiling the linter on either the microbenchmarks or entire projects
 
+> \[!NOTE\]
+> When running benchmarks, ensure that your CPU is otherwise idle (e.g., close any background
+> applications, like web browsers). You may also want to switch your CPU to a "performance"
+> mode, if it exists, especially when benchmarking short-lived processes.
+
 ### CPython Benchmark
 
 First, clone [CPython](https://github.com/python/cpython). It's a large and diverse Python codebase,

Cargo.lock

@@ -273,9 +273,9 @@ dependencies = [
 
 [[package]]
 name = "chrono"
-version = "0.4.33"
+version = "0.4.34"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9f13690e35a5e4ace198e7beea2895d29f3a9cc55015fcebe6336bd2010af9eb"
+checksum = "5bc015644b92d5890fab7489e49d21f879d5c990186827d42ec511919404f38b"
 dependencies = [
  "android-tzdata",
  "iana-time-zone",
@@ -1037,9 +1037,9 @@ dependencies = [
 
 [[package]]
 name = "indicatif"
-version = "0.17.7"
+version = "0.17.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fb28741c9db9a713d93deb3bb9515c20788cef5815265bee4980e87bde7e0f25"
+checksum = "763a5a8f45087d6bcea4222e7b72c291a054edf80e4ef6efd2a4979878c7bea3"
 dependencies = [
  "console",
  "instant",
@@ -1979,7 +1979,7 @@ dependencies = [
 
 [[package]]
 name = "ruff"
-version = "0.2.1"
+version = "0.2.2"
 dependencies = [
  "anyhow",
  "argfile",
@@ -2140,7 +2140,7 @@ dependencies = [
 
 [[package]]
 name = "ruff_linter"
-version = "0.2.1"
+version = "0.2.2"
 dependencies = [
  "aho-corasick",
  "annotate-snippets 0.9.2",
@@ -2394,7 +2394,7 @@ dependencies = [
 
 [[package]]
 name = "ruff_shrinking"
-version = "0.2.1"
+version = "0.2.2"
 dependencies = [
  "anyhow",
  "clap",
@@ -2944,18 +2944,18 @@ dependencies = [
 
 [[package]]
 name = "thiserror"
-version = "1.0.56"
+version = "1.0.57"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d54378c645627613241d077a3a79db965db602882668f9136ac42af9ecb730ad"
+checksum = "1e45bcbe8ed29775f228095caf2cd67af7a4ccf756ebff23a306bf3e8b47b24b"
 dependencies = [
  "thiserror-impl",
 ]
 
 [[package]]
 name = "thiserror-impl"
-version = "1.0.56"
+version = "1.0.57"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fa0faa943b50f3db30a20aa7e265dbc66076993efed8463e8de414e5d06d3471"
+checksum = "a953cb265bef375dae3de6663da4d3804eee9682ea80d8e2542529b73c531c81"
 dependencies = [
  "proc-macro2",
  "quote",

Cargo.toml

@@ -21,7 +21,7 @@ bincode = { version = "1.3.3" }
 bitflags = { version = "2.4.1" }
 bstr = { version = "1.9.0" }
 cachedir = { version = "0.3.1" }
-chrono = { version = "0.4.33", default-features = false, features = ["clock"] }
+chrono = { version = "0.4.34", default-features = false, features = ["clock"] }
 clap = { version = "4.4.18", features = ["derive"] }
 clap_complete_command = { version = "0.5.1" }
 clearscreen = { version = "2.0.0" }
@@ -44,7 +44,7 @@ hexf-parse = { version ="0.2.1"}
 ignore = { version = "0.4.22" }
 imara-diff ={ version = "0.1.5"}
 imperative = { version = "1.0.4" }
-indicatif ={ version = "0.17.7"}
+indicatif ={ version = "0.17.8"}
 indoc ={ version = "2.0.4"}
 insta = { version = "1.34.0", feature = ["filters", "glob"] }
 insta-cmd = { version = "0.4.0" }
@@ -92,7 +92,7 @@ strum_macros = { version = "0.25.3" }
 syn = { version = "2.0.40" }
 tempfile = { version ="3.9.0"}
 test-case = { version = "3.3.1" }
-thiserror = { version = "1.0.51" }
+thiserror = { version = "1.0.57" }
 tikv-jemallocator = { version ="0.5.0"}
 toml = { version = "0.8.9" }
 tracing = { version = "0.1.40" }

README.md

@@ -8,7 +8,7 @@
 [![image](https://img.shields.io/pypi/pyversions/ruff.svg)](https://pypi.python.org/pypi/ruff)
 [![Actions status](https://github.com/astral-sh/ruff/workflows/CI/badge.svg)](https://github.com/astral-sh/ruff/actions)
 
-[**Discord**](https://discord.gg/c9MhzV8aU5) | [**Docs**](https://docs.astral.sh/ruff/) | [**Playground**](https://play.ruff.rs/)
+[**Discord**](https://discord.com/invite/astral-sh) | [**Docs**](https://docs.astral.sh/ruff/) | [**Playground**](https://play.ruff.rs/)
 
 An extremely fast Python linter and code formatter, written in Rust.
 
@@ -150,7 +150,7 @@ Ruff can also be used as a [pre-commit](https://pre-commit.com/) hook via [`ruff
 ```yaml
 - repo: https://github.com/astral-sh/ruff-pre-commit
   # Ruff version.
-  rev: v0.2.1
+  rev: v0.2.2
   hooks:
     # Run the linter.
     - id: ruff
@@ -341,14 +341,14 @@ For a complete enumeration of the supported rules, see [_Rules_](https://docs.as
 Contributions are welcome and highly appreciated. To get started, check out the
 [**contributing guidelines**](https://docs.astral.sh/ruff/contributing/).
 
-You can also join us on [**Discord**](https://discord.gg/c9MhzV8aU5).
+You can also join us on [**Discord**](https://discord.com/invite/astral-sh).
 
 ## Support
 
 Having trouble? Check out the existing issues on [**GitHub**](https://github.com/astral-sh/ruff/issues),
 or feel free to [**open a new one**](https://github.com/astral-sh/ruff/issues/new).
 
-You can also ask for help on [**Discord**](https://discord.gg/c9MhzV8aU5).
+You can also ask for help on [**Discord**](https://discord.com/invite/astral-sh).
 
 ## Acknowledgements
 

crates/ruff/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ruff"
-version = "0.2.1"
+version = "0.2.2"
 publish = false
 authors = { workspace = true }
 edition = { workspace = true }
@@ -48,6 +48,7 @@ serde = { workspace = true }
 serde_json = { workspace = true }
 shellexpand = { workspace = true }
 strum = { workspace = true, features = [] }
+tempfile = { workspace = true }
 thiserror = { workspace = true }
 toml = { workspace = true }
 tracing = { workspace = true, features = ["log"] }

crates/ruff/src/cache.rs

@@ -1,7 +1,7 @@
 use std::fmt::Debug;
 use std::fs::{self, File};
 use std::hash::Hasher;
-use std::io::{self, BufReader, BufWriter, Write};
+use std::io::{self, BufReader, Write};
 use std::path::{Path, PathBuf};
 use std::sync::atomic::{AtomicU64, Ordering};
 use std::sync::Mutex;
@@ -15,6 +15,7 @@ use rayon::iter::ParallelIterator;
 use rayon::iter::{IntoParallelIterator, ParallelBridge};
 use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
+use tempfile::NamedTempFile;
 
 use ruff_cache::{CacheKey, CacheKeyHasher};
 use ruff_diagnostics::{DiagnosticKind, Fix};
@@ -165,15 +166,29 @@ impl Cache {
             return Ok(());
         }
 
-        let file = File::create(&self.path)
-            .with_context(|| format!("Failed to create cache file '{}'", self.path.display()))?;
-        let writer = BufWriter::new(file);
-        bincode::serialize_into(writer, &self.package).with_context(|| {
+        // Write the cache to a temporary file first and then rename it for an "atomic" write.
+        // Protects against data loss if the process is killed during the write and races between different ruff
+        // processes, resulting in a corrupted cache file. https://github.com/astral-sh/ruff/issues/8147#issuecomment-1943345964
+        let mut temp_file =
+            NamedTempFile::new_in(self.path.parent().expect("Write path must have a parent"))
+                .context("Failed to create temporary file")?;
+
+        // Serialize to in-memory buffer because hyperfine benchmark showed that it's faster than
+        // using a `BufWriter` and our cache files are small enough that streaming isn't necessary.
+        let serialized =
+            bincode::serialize(&self.package).context("Failed to serialize cache data")?;
+        temp_file
+            .write_all(&serialized)
+            .context("Failed to write serialized cache to temporary file.")?;
+
+        temp_file.persist(&self.path).with_context(|| {
             format!(
-                "Failed to serialise cache to file '{}'",
+                "Failed to rename temporary cache file to {}",
                 self.path.display()
             )
-        })
+        })?;
+
+        Ok(())
     }
 
     /// Applies the pending changes without storing the cache to disk.

crates/ruff/tests/snapshots/integration_test__explain_status_codes_f401.snap

@@ -25,6 +25,15 @@ import cycles. They also increase the cognitive load of reading the code.
 If an import statement is used to check for the availability or existence
 of a module, consider using `importlib.util.find_spec` instead.
 
+If an import statement is used to re-export a symbol as part of a module's
+public interface, consider using a "redundant" import alias, which
+instructs Ruff (and other tools) to respect the re-export, and avoid
+marking it as unused, as in:
+
+```python
+from module import member as member
+```
+
 ## Example
 ```python
 import numpy as np  # unused import
@@ -51,11 +60,12 @@ else:
 ```
 
 ## Options
-- `lint.pyflakes.extend-generics`
+- `lint.ignore-init-module-imports`
 
 ## References
 - [Python documentation: `import`](https://docs.python.org/3/reference/simple_stmts.html#the-import-statement)
 - [Python documentation: `importlib.util.find_spec`](https://docs.python.org/3/library/importlib.html#importlib.util.find_spec)
+- [Typing documentation: interface conventions](https://typing.readthedocs.io/en/latest/source/libraries.html#library-interface-public-and-private-symbols)
 
 ----- stderr -----
 

crates/ruff_benchmark/Cargo.toml

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

crates/ruff_diagnostics/Cargo.toml

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

crates/ruff_index/Cargo.toml

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

crates/ruff_linter/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ruff_linter"
-version = "0.2.1"
+version = "0.2.2"
 publish = false
 authors = { workspace = true }
 edition = { workspace = true }

crates/ruff_linter/resources/test/fixtures/flake8_pyi/PYI049.py

@@ -11,13 +11,25 @@ class _UnusedTypedDict2(typing.TypedDict):
 
 
 class _UsedTypedDict(TypedDict):
-   foo: bytes
+    foo: bytes
 
 
 class _CustomClass(_UsedTypedDict):
     bar: list[int]
 
+
 _UnusedTypedDict3 = TypedDict("_UnusedTypedDict3", {"foo": int})
 _UsedTypedDict3 = TypedDict("_UsedTypedDict3", {"bar": bytes})
 
+
 def uses_UsedTypedDict3(arg: _UsedTypedDict3) -> None: ...
+
+
+# In `.py` files, we don't flag unused definitions in class scopes (unlike in `.pyi`
+# files).
+class _CustomClass3:
+    class _UnusedTypeDict4(TypedDict):
+        pass
+
+    def method(self) -> None:
+        _CustomClass3._UnusedTypeDict4()

crates/ruff_linter/resources/test/fixtures/flake8_pyi/PYI049.pyi

@@ -35,3 +35,13 @@ _UnusedTypedDict3 = TypedDict("_UnusedTypedDict3", {"foo": int})
 _UsedTypedDict3 = TypedDict("_UsedTypedDict3", {"bar": bytes})
 
 def uses_UsedTypedDict3(arg: _UsedTypedDict3) -> None: ...
+
+
+# In `.pyi` files, we flag unused definitions in class scopes as well as in the global
+# scope (unlike in `.py` files).
+class _CustomClass3:
+    class _UnusedTypeDict4(TypedDict):
+        pass
+
+    def method(self) -> None:
+        _CustomClass3._UnusedTypeDict4()

crates/ruff_linter/resources/test/fixtures/flake8_simplify/SIM113.py

@@ -193,3 +193,11 @@ def func():
         for y in range(5):
             g(x, idx)
             idx += 1
+
+async def func():
+    # OK (for loop is async)
+    idx = 0
+
+    async for x in async_gen():
+        g(x, idx)
+        idx += 1

crates/ruff_linter/resources/test/fixtures/perflint/PERF101.py

@@ -36,35 +36,47 @@ for i in list( # Comment
 ):  # PERF101
     pass
 
-for i in list(foo_dict):  # Ok
+for i in list(foo_dict):  # OK
     pass
 
-for i in list(1):  # Ok
+for i in list(1):  # OK
     pass
 
-for i in list(foo_int):  # Ok
+for i in list(foo_int):  # OK
     pass
 
 
 import itertools
 
-for i in itertools.product(foo_int):  # Ok
+for i in itertools.product(foo_int):  # OK
     pass
 
-for i in list(foo_list):  # Ok
+for i in list(foo_list):  # OK
     foo_list.append(i + 1)
 
 for i in list(foo_list):  # PERF101
     # Make sure we match the correct list
     other_list.append(i + 1)
 
-for i in list(foo_tuple):  # Ok
+for i in list(foo_tuple):  # OK
     foo_tuple.append(i + 1)
 
-for i in list(foo_set):  # Ok
+for i in list(foo_set):  # OK
     foo_set.append(i + 1)
 
 x, y, nested_tuple = (1, 2, (3, 4, 5))
 
 for i in list(nested_tuple):  # PERF101
     pass
+
+for i in list(foo_list):  # OK
+    if True:
+        foo_list.append(i + 1)
+
+for i in list(foo_list):  # OK
+    if True:
+        foo_list[i] = i + 1
+
+for i in list(foo_list):  # OK
+    if True:
+        del foo_list[i + 1]

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

@@ -4,7 +4,12 @@
 1 in (
     1, 2, 3
 )
-
-# OK
 fruits = ["cherry", "grapes"]
 "cherry" in fruits
+_ = {key: value for key, value in {"a": 1, "b": 2}.items() if key in ("a", "b")}
+
+# OK
+fruits in [[1, 2, 3], [4, 5, 6]]
+fruits in [1, 2, 3]
+1 in [[1, 2, 3], [4, 5, 6]]
+_ = {key: value for key, value in {"a": 1, "b": 2}.items() if key in (["a", "b"], ["c", "d"])}

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

@@ -35,6 +35,15 @@ if argc != 0:  # correct
 if argc != 1:  # correct
     pass
 
+if argc != -1.0:  # correct
+    pass
+
+if argc != 0.0:  # correct
+    pass
+
+if argc != 1.0:  # correct
+    pass
+
 if argc != 2:  # [magic-value-comparison]
     pass
 
@@ -44,6 +53,12 @@ if argc != -2:  # [magic-value-comparison]
 if argc != +2:  # [magic-value-comparison]
     pass
 
+if argc != -2.0:  # [magic-value-comparison]
+    pass
+
+if argc != +2.0:  # [magic-value-comparison]
+    pass
+
 if __name__ == "__main__":  # correct
     pass
 

crates/ruff_linter/resources/test/fixtures/refurb/FURB129.py

@@ -0,0 +1,75 @@
+import codecs
+import io
+from pathlib import Path
+
+# Errors
+with open("FURB129.py") as f:
+    for _line in f.readlines():
+        pass
+    a = [line.lower() for line in f.readlines()]
+    b = {line.upper() for line in f.readlines()}
+    c = {line.lower(): line.upper() for line in f.readlines()}
+
+with Path("FURB129.py").open() as f:
+    for _line in f.readlines():
+        pass
+
+for _line in open("FURB129.py").readlines():
+    pass
+
+for _line in Path("FURB129.py").open().readlines():
+    pass
+
+
+def func():
+    f = Path("FURB129.py").open()
+    for _line in f.readlines():
+        pass
+    f.close()
+
+
+def func(f: io.BytesIO):
+    for _line in f.readlines():
+        pass
+
+
+def func():
+    with (open("FURB129.py") as f, foo as bar):
+        for _line in f.readlines():
+            pass
+        for _line in bar.readlines():
+            pass
+
+
+# False positives
+def func(f):
+    for _line in f.readlines():
+        pass
+
+
+def func(f: codecs.StreamReader):
+    for _line in f.readlines():
+        pass
+
+
+def func():
+    class A:
+        def readlines(self) -> list[str]:
+            return ["a", "b", "c"]
+
+    return A()
+
+
+for _line in func().readlines():
+    pass
+
+# OK
+for _line in ["a", "b", "c"]:
+    pass
+with open("FURB129.py") as f:
+    for _line in f:
+        pass
+    for _line in f.readlines(10):
+        pass
+    for _not_line in f.readline():
+        pass

crates/ruff_linter/resources/test/fixtures/ruff/RUF006.py

@@ -162,3 +162,26 @@ async def f(x: bool):
         T = asyncio.create_task(asyncio.sleep(1))
     else:
         T = None
+
+
+# Error
+def f():
+    loop = asyncio.new_event_loop()
+    loop.create_task(main()) # Error
+
+# Error
+def f():
+    loop = asyncio.get_event_loop()
+    loop.create_task(main()) # Error
+
+# OK
+def f():
+    global task
+    loop = asyncio.new_event_loop()
+    task = loop.create_task(main()) # Error
+
+# OK
+def f():
+    global task
+    loop = asyncio.get_event_loop()
+    task = loop.create_task(main()) # Error

crates/ruff_linter/resources/test/fixtures/ruff/RUF027_0.py

@@ -68,3 +68,7 @@ def method_calls():
     first = "Wendy"
     last = "Appleseed"
     value.method("{first} {last}")  # RUF027
+
+def format_specifiers():
+    a = 4
+    b = "{a:b} {a:^5}"

crates/ruff_linter/resources/test/fixtures/ruff/RUF027_2.py

@@ -0,0 +1,2 @@
+# 测试eval函数,eval()函数用来执行一个字符串表达式，并返t表达式的值。另外，可以讲字符串转换成列表或元组或字典
+a = "{1: 'a', 2: 'b'}"

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

@@ -2,11 +2,14 @@ use ruff_python_ast::Comprehension;
 
 use crate::checkers::ast::Checker;
 use crate::codes::Rule;
-use crate::rules::flake8_simplify;
+use crate::rules::{flake8_simplify, refurb};
 
 /// Run lint rules over a [`Comprehension`] syntax nodes.
 pub(crate) fn comprehension(comprehension: &Comprehension, checker: &mut Checker) {
     if checker.enabled(Rule::InDictKeys) {
         flake8_simplify::rules::key_in_dict_comprehension(checker, comprehension);
     }
+    if checker.enabled(Rule::ReadlinesInFor) {
+        refurb::rules::readlines_in_comprehension(checker, comprehension);
+    }
 }

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

@@ -281,17 +281,21 @@ pub(crate) fn deferred_scopes(checker: &mut Checker) {
             }
         }
 
-        if checker.enabled(Rule::UnusedPrivateTypeVar) {
-            flake8_pyi::rules::unused_private_type_var(checker, scope, &mut diagnostics);
-        }
-        if checker.enabled(Rule::UnusedPrivateProtocol) {
-            flake8_pyi::rules::unused_private_protocol(checker, scope, &mut diagnostics);
-        }
-        if checker.enabled(Rule::UnusedPrivateTypeAlias) {
-            flake8_pyi::rules::unused_private_type_alias(checker, scope, &mut diagnostics);
-        }
-        if checker.enabled(Rule::UnusedPrivateTypedDict) {
-            flake8_pyi::rules::unused_private_typed_dict(checker, scope, &mut diagnostics);
+        if checker.source_type.is_stub()
+            || matches!(scope.kind, ScopeKind::Module | ScopeKind::Function(_))
+        {
+            if checker.enabled(Rule::UnusedPrivateTypeVar) {
+                flake8_pyi::rules::unused_private_type_var(checker, scope, &mut diagnostics);
+            }
+            if checker.enabled(Rule::UnusedPrivateProtocol) {
+                flake8_pyi::rules::unused_private_protocol(checker, scope, &mut diagnostics);
+            }
+            if checker.enabled(Rule::UnusedPrivateTypeAlias) {
+                flake8_pyi::rules::unused_private_type_alias(checker, scope, &mut diagnostics);
+            }
+            if checker.enabled(Rule::UnusedPrivateTypedDict) {
+                flake8_pyi::rules::unused_private_typed_dict(checker, scope, &mut diagnostics);
+            }
         }
 
         if checker.enabled(Rule::AsyncioDanglingTask) {

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

@@ -1317,6 +1317,9 @@ pub(crate) fn statement(stmt: &Stmt, checker: &mut Checker) {
             if checker.enabled(Rule::UnnecessaryDictIndexLookup) {
                 pylint::rules::unnecessary_dict_index_lookup(checker, for_stmt);
             }
+            if checker.enabled(Rule::ReadlinesInFor) {
+                refurb::rules::readlines_in_for(checker, for_stmt);
+            }
             if !is_async {
                 if checker.enabled(Rule::ReimplementedBuiltin) {
                     flake8_simplify::rules::convert_for_loop_to_any_all(checker, stmt);

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

@@ -2,10 +2,16 @@ use ruff_python_ast::StringLike;
 
 use crate::checkers::ast::Checker;
 use crate::codes::Rule;
-use crate::rules::{flake8_bandit, flake8_pyi};
+use crate::rules::{flake8_bandit, flake8_pyi, ruff};
 
 /// Run lint rules over a [`StringLike`] syntax nodes.
 pub(crate) fn string_like(string_like: StringLike, checker: &mut Checker) {
+    if checker.any_enabled(&[
+        Rule::AmbiguousUnicodeCharacterString,
+        Rule::AmbiguousUnicodeCharacterDocstring,
+    ]) {
+        ruff::rules::ambiguous_unicode_character_string(checker, string_like);
+    }
     if checker.enabled(Rule::HardcodedBindAllInterfaces) {
         flake8_bandit::rules::hardcoded_bind_all_interfaces(checker, string_like);
     }

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

@@ -40,7 +40,7 @@ use ruff_diagnostics::{Diagnostic, IsolationLevel};
 use ruff_notebook::{CellOffsets, NotebookIndex};
 use ruff_python_ast::all::{extract_all_names, DunderAllFlags};
 use ruff_python_ast::helpers::{
-    collect_import_from_member, extract_handled_exceptions, to_module_path,
+    collect_import_from_member, extract_handled_exceptions, is_docstring_stmt, to_module_path,
 };
 use ruff_python_ast::identifier::Identifier;
 use ruff_python_ast::str::trailing_quote;
@@ -71,6 +71,38 @@ mod analyze;
 mod annotation;
 mod deferred;
 
+/// State representing whether a docstring is expected or not for the next statement.
+#[derive(Default, Debug, Copy, Clone, PartialEq)]
+enum DocstringState {
+    /// The next statement is expected to be a docstring, but not necessarily so.
+    ///
+    /// For example, in the following code:
+    ///
+    /// ```python
+    /// class Foo:
+    ///     pass
+    ///
+    ///
+    /// def bar(x, y):
+    ///     """Docstring."""
+    ///     return x +  y
+    /// ```
+    ///
+    /// For `Foo`, the state is expected when the checker is visiting the class
+    /// body but isn't going to be present. While, for `bar` function, the docstring
+    /// is expected and present.
+    #[default]
+    Expected,
+    Other,
+}
+
+impl DocstringState {
+    /// Returns `true` if the next statement is expected to be a docstring.
+    const fn is_expected(self) -> bool {
+        matches!(self, DocstringState::Expected)
+    }
+}
+
 pub(crate) struct Checker<'a> {
     /// The [`Path`] to the file under analysis.
     path: &'a Path,
@@ -114,6 +146,8 @@ pub(crate) struct Checker<'a> {
     pub(crate) flake8_bugbear_seen: Vec<TextRange>,
     /// The end offset of the last visited statement.
     last_stmt_end: TextSize,
+    /// A state describing if a docstring is expected or not.
+    docstring_state: DocstringState,
 }
 
 impl<'a> Checker<'a> {
@@ -153,6 +187,7 @@ impl<'a> Checker<'a> {
             cell_offsets,
             notebook_index,
             last_stmt_end: TextSize::default(),
+            docstring_state: DocstringState::default(),
         }
     }
 }
@@ -305,19 +340,16 @@ where
             self.semantic.flags -= SemanticModelFlags::IMPORT_BOUNDARY;
         }
 
-        // Track whether we've seen docstrings, non-imports, etc.
+        // Track whether we've seen module docstrings, non-imports, etc.
         match stmt {
             Stmt::Expr(ast::StmtExpr { value, .. })
-                if !self
-                    .semantic
-                    .flags
-                    .intersects(SemanticModelFlags::MODULE_DOCSTRING)
+                if !self.semantic.seen_module_docstring_boundary()
                     && value.is_string_literal_expr() =>
             {
-                self.semantic.flags |= SemanticModelFlags::MODULE_DOCSTRING;
+                self.semantic.flags |= SemanticModelFlags::MODULE_DOCSTRING_BOUNDARY;
             }
             Stmt::ImportFrom(ast::StmtImportFrom { module, names, .. }) => {
-                self.semantic.flags |= SemanticModelFlags::MODULE_DOCSTRING;
+                self.semantic.flags |= SemanticModelFlags::MODULE_DOCSTRING_BOUNDARY;
 
                 // Allow __future__ imports until we see a non-__future__ import.
                 if let Some("__future__") = module.as_deref() {
@@ -332,11 +364,11 @@ where
                 }
             }
             Stmt::Import(_) => {
-                self.semantic.flags |= SemanticModelFlags::MODULE_DOCSTRING;
+                self.semantic.flags |= SemanticModelFlags::MODULE_DOCSTRING_BOUNDARY;
                 self.semantic.flags |= SemanticModelFlags::FUTURES_BOUNDARY;
             }
             _ => {
-                self.semantic.flags |= SemanticModelFlags::MODULE_DOCSTRING;
+                self.semantic.flags |= SemanticModelFlags::MODULE_DOCSTRING_BOUNDARY;
                 self.semantic.flags |= SemanticModelFlags::FUTURES_BOUNDARY;
                 if !(self.semantic.seen_import_boundary()
                     || helpers::is_assignment_to_a_dunder(stmt)
@@ -353,6 +385,16 @@ where
         // the node.
         let flags_snapshot = self.semantic.flags;
 
+        // Update the semantic model if it is in a docstring. This should be done after the
+        // flags snapshot to ensure that it gets reset once the statement is analyzed.
+        if self.docstring_state.is_expected() {
+            if is_docstring_stmt(stmt) {
+                self.semantic.flags |= SemanticModelFlags::DOCSTRING;
+            }
+            // Reset the state irrespective of whether the statement is a docstring or not.
+            self.docstring_state = DocstringState::Other;
+        }
+
         // Step 1: Binding
         match stmt {
             Stmt::AugAssign(ast::StmtAugAssign {
@@ -654,6 +696,8 @@ where
                     self.semantic.set_globals(globals);
                 }
 
+                // Set the docstring state before visiting the class body.
+                self.docstring_state = DocstringState::Expected;
                 self.visit_body(body);
             }
             Stmt::TypeAlias(ast::StmtTypeAlias {
@@ -1288,6 +1332,16 @@ where
                 self.semantic.flags |= SemanticModelFlags::F_STRING;
                 visitor::walk_expr(self, expr);
             }
+            Expr::NamedExpr(ast::ExprNamedExpr {
+                target,
+                value,
+                range: _,
+            }) => {
+                self.visit_expr(value);
+
+                self.semantic.flags |= SemanticModelFlags::NAMED_EXPRESSION_ASSIGNMENT;
+                self.visit_expr(target);
+            }
             _ => visitor::walk_expr(self, expr),
         }
 
@@ -1504,6 +1558,8 @@ impl<'a> Checker<'a> {
             unreachable!("Generator expression must contain at least one generator");
         };
 
+        let flags = self.semantic.flags;
+
         // Generators are compiled as nested functions. (This may change with PEP 709.)
         // As such, the `iter` of the first generator is evaluated in the outer scope, while all
         // subsequent nodes are evaluated in the inner scope.
@@ -1533,14 +1589,22 @@ impl<'a> Checker<'a> {
         // `x` is local to `foo`, and the `T` in `y=T` skips the class scope when resolving.
         self.visit_expr(&generator.iter);
         self.semantic.push_scope(ScopeKind::Generator);
+
+        self.semantic.flags = flags | SemanticModelFlags::COMPREHENSION_ASSIGNMENT;
         self.visit_expr(&generator.target);
+        self.semantic.flags = flags;
+
         for expr in &generator.ifs {
             self.visit_boolean_test(expr);
         }
 
         for generator in iterator {
             self.visit_expr(&generator.iter);
+
+            self.semantic.flags = flags | SemanticModelFlags::COMPREHENSION_ASSIGNMENT;
             self.visit_expr(&generator.target);
+            self.semantic.flags = flags;
+
             for expr in &generator.ifs {
                 self.visit_boolean_test(expr);
             }
@@ -1739,11 +1803,21 @@ impl<'a> Checker<'a> {
             return;
         }
 
+        // A binding within a `for` must be a loop variable, as in:
+        // ```python
+        // for x in range(10):
+        //     ...
+        // ```
         if parent.is_for_stmt() {
             self.add_binding(id, expr.range(), BindingKind::LoopVar, flags);
             return;
         }
 
+        // A binding within a `with` must be an item, as in:
+        // ```python
+        // with open("file.txt") as fp:
+        //     ...
+        // ```
         if parent.is_with_stmt() {
             self.add_binding(id, expr.range(), BindingKind::WithItemVar, flags);
             return;
@@ -1799,17 +1873,26 @@ impl<'a> Checker<'a> {
         }
 
         // If the expression is the left-hand side of a walrus operator, then it's a named
-        // expression assignment.
-        if self
-            .semantic
-            .current_expressions()
-            .filter_map(Expr::as_named_expr_expr)
-            .any(|parent| parent.target.as_ref() == expr)
-        {
+        // expression assignment, as in:
+        // ```python
+        // if (x := 10) > 5:
+        //     ...
+        // ```
+        if self.semantic.in_named_expression_assignment() {
             self.add_binding(id, expr.range(), BindingKind::NamedExprAssignment, flags);
             return;
         }
 
+        // If the expression is part of a comprehension target, then it's a comprehension variable
+        // assignment, as in:
+        // ```python
+        // [x for x in range(10)]
+        // ```
+        if self.semantic.in_comprehension_assignment() {
+            self.add_binding(id, expr.range(), BindingKind::ComprehensionVar, flags);
+            return;
+        }
+
         self.add_binding(id, expr.range(), BindingKind::Assignment, flags);
     }
 
@@ -1925,6 +2008,8 @@ impl<'a> Checker<'a> {
                 };
 
                 self.visit_parameters(parameters);
+                // Set the docstring state before visiting the function body.
+                self.docstring_state = DocstringState::Expected;
                 self.visit_body(body);
             }
         }

crates/ruff_linter/src/checkers/tokens.rs

@@ -6,17 +6,14 @@ use ruff_notebook::CellOffsets;
 use ruff_python_ast::PySourceType;
 use ruff_python_codegen::Stylist;
 use ruff_python_parser::lexer::LexResult;
-use ruff_python_parser::Tok;
 
 use ruff_diagnostics::Diagnostic;
 use ruff_python_index::Indexer;
 use ruff_source_file::Locator;
 
 use crate::directives::TodoComment;
-use crate::lex::docstring_detection::StateMachine;
 use crate::registry::{AsRule, Rule};
 use crate::rules::pycodestyle::rules::BlankLinesChecker;
-use crate::rules::ruff::rules::Context;
 use crate::rules::{
     eradicate, flake8_commas, flake8_executable, flake8_fixme, flake8_implicit_str_concat,
     flake8_pyi, flake8_quotes, flake8_todos, pycodestyle, pygrep_hooks, pylint, pyupgrade, ruff,
@@ -66,31 +63,15 @@ pub(crate) fn check_tokens(
         pylint::rules::empty_comments(&mut diagnostics, indexer, locator);
     }
 
-    if settings.rules.any_enabled(&[
-        Rule::AmbiguousUnicodeCharacterString,
-        Rule::AmbiguousUnicodeCharacterDocstring,
-        Rule::AmbiguousUnicodeCharacterComment,
-    ]) {
-        let mut state_machine = StateMachine::default();
-        for &(ref tok, range) in tokens.iter().flatten() {
-            let is_docstring = state_machine.consume(tok);
-            let context = match tok {
-                Tok::String { .. } => {
-                    if is_docstring {
-                        Context::Docstring
-                    } else {
-                        Context::String
-                    }
-                }
-                Tok::FStringMiddle { .. } => Context::String,
-                Tok::Comment(_) => Context::Comment,
-                _ => continue,
-            };
-            ruff::rules::ambiguous_unicode_character(
+    if settings
+        .rules
+        .enabled(Rule::AmbiguousUnicodeCharacterComment)
+    {
+        for range in indexer.comment_ranges() {
+            ruff::rules::ambiguous_unicode_character_comment(
                 &mut diagnostics,
                 locator,
-                range,
-                context,
+                *range,
                 settings,
             );
         }

crates/ruff_linter/src/codes.rs

@@ -1025,6 +1025,7 @@ pub fn code_to_rule(linter: Linter, code: &str) -> Option<(RuleGroup, Rule)> {
         #[allow(deprecated)]
         (Refurb, "113") => (RuleGroup::Nursery, rules::refurb::rules::RepeatedAppend),
         (Refurb, "118") => (RuleGroup::Preview, rules::refurb::rules::ReimplementedOperator),
+        (Refurb, "129") => (RuleGroup::Preview, rules::refurb::rules::ReadlinesInFor),
         #[allow(deprecated)]
         (Refurb, "131") => (RuleGroup::Nursery, rules::refurb::rules::DeleteFullSlice),
         #[allow(deprecated)]

crates/ruff_linter/src/registry.rs

@@ -256,8 +256,6 @@ impl Rule {
             | Rule::MixedSpacesAndTabs
             | Rule::TrailingWhitespace => LintSource::PhysicalLines,
             Rule::AmbiguousUnicodeCharacterComment
-            | Rule::AmbiguousUnicodeCharacterDocstring
-            | Rule::AmbiguousUnicodeCharacterString
             | Rule::AvoidableEscapedQuote
             | Rule::BadQuotesDocstring
             | Rule::BadQuotesInlineString

crates/ruff_linter/src/renamer.rs

@@ -248,6 +248,7 @@ impl Renamer {
             | BindingKind::Assignment
             | BindingKind::BoundException
             | BindingKind::LoopVar
+            | BindingKind::ComprehensionVar
             | BindingKind::WithItemVar
             | BindingKind::Global
             | BindingKind::Nonlocal(_)

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

@@ -118,8 +118,7 @@ fn is_open_call_from_pathlib(func: &Expr, semantic: &SemanticModel) -> bool {
 
     let binding = semantic.binding(binding_id);
 
-    let Some(Expr::Call(call)) = analyze::typing::find_binding_value(&name.id, binding, semantic)
-    else {
+    let Some(Expr::Call(call)) = analyze::typing::find_binding_value(binding, semantic) else {
         return false;
     };
 

crates/ruff_linter/src/rules/flake8_pyi/snapshots/ruff_linter__rules__flake8_pyi__tests__PYI049_PYI049.py.snap

@@ -15,13 +15,11 @@ PYI049.py:9:7: PYI049 Private TypedDict `_UnusedTypedDict2` is never used
 10 |     bar: int
    |
 
-PYI049.py:20:1: PYI049 Private TypedDict `_UnusedTypedDict3` is never used
+PYI049.py:21:1: PYI049 Private TypedDict `_UnusedTypedDict3` is never used
    |
-18 |     bar: list[int]
-19 | 
-20 | _UnusedTypedDict3 = TypedDict("_UnusedTypedDict3", {"foo": int})
+21 | _UnusedTypedDict3 = TypedDict("_UnusedTypedDict3", {"foo": int})
    | ^^^^^^^^^^^^^^^^^ PYI049
-21 | _UsedTypedDict3 = TypedDict("_UsedTypedDict3", {"bar": bytes})
+22 | _UsedTypedDict3 = TypedDict("_UsedTypedDict3", {"bar": bytes})
    |
 
 

crates/ruff_linter/src/rules/flake8_pyi/snapshots/ruff_linter__rules__flake8_pyi__tests__PYI049_PYI049.pyi.snap

@@ -24,4 +24,13 @@ PYI049.pyi:34:1: PYI049 Private TypedDict `_UnusedTypedDict3` is never used
 35 | _UsedTypedDict3 = TypedDict("_UsedTypedDict3", {"bar": bytes})
    |
 
+PYI049.pyi:43:11: PYI049 Private TypedDict `_UnusedTypeDict4` is never used
+   |
+41 | # scope (unlike in `.py` files).
+42 | class _CustomClass3:
+43 |     class _UnusedTypeDict4(TypedDict):
+   |           ^^^^^^^^^^^^^^^^ PYI049
+44 |         pass
+   |
+
 

crates/ruff_linter/src/rules/flake8_simplify/rules/enumerate_for_loop.rs

@@ -49,6 +49,11 @@ impl Violation for EnumerateForLoop {
 
 /// SIM113
 pub(crate) fn enumerate_for_loop(checker: &mut Checker, for_stmt: &ast::StmtFor) {
+    // If the loop is async, abort.
+    if for_stmt.is_async {
+        return;
+    }
+
     // If the loop contains a `continue`, abort.
     let mut visitor = LoopControlFlowVisitor::default();
     visitor.visit_body(&for_stmt.body);
@@ -76,8 +81,7 @@ pub(crate) fn enumerate_for_loop(checker: &mut Checker, for_stmt: &ast::StmtFor)
             }
 
             // Ensure that the index variable was initialized to 0.
-            let Some(value) = typing::find_binding_value(&index.id, binding, checker.semantic())
-            else {
+            let Some(value) = typing::find_binding_value(binding, checker.semantic()) else {
                 continue;
             };
             if !matches!(

crates/ruff_linter/src/rules/isort/mod.rs

@@ -419,23 +419,20 @@ mod tests {
         Ok(())
     }
 
-    // Test currently disabled as line endings are automatically converted to
-    // platform-appropriate ones in CI/CD #[test_case(Path::new("
-    // line_ending_crlf.py"))] #[test_case(Path::new("line_ending_lf.py"))]
-    // fn source_code_style(path: &Path) -> Result<()> {
-    //     let snapshot = format!("{}", path.to_string_lossy());
-    //     let diagnostics = test_path(
-    //         Path::new("isort")
-    //             .join(path)
-    //             .as_path(),
-    //         &LinterSettings {
-    //             src: vec![test_resource_path("fixtures/isort")],
-    //             ..LinterSettings::for_rule(Rule::UnsortedImports)
-    //         },
-    //     )?;
-    //     crate::assert_messages!(snapshot, diagnostics);
-    //     Ok(())
-    // }
+    #[test_case(Path::new("line_ending_crlf.py"))]
+    #[test_case(Path::new("line_ending_lf.py"))]
+    fn source_code_style(path: &Path) -> Result<()> {
+        let snapshot = format!("{}", path.to_string_lossy());
+        let diagnostics = test_path(
+            Path::new("isort").join(path).as_path(),
+            &LinterSettings {
+                src: vec![test_resource_path("fixtures/isort")],
+                ..LinterSettings::for_rule(Rule::UnsortedImports)
+            },
+        )?;
+        crate::assert_messages!(snapshot, diagnostics);
+        Ok(())
+    }
 
     #[test_case(Path::new("separate_local_folder_imports.py"))]
     fn known_local_folder(path: &Path) -> Result<()> {

crates/ruff_linter/src/rules/isort/snapshots/ruff_linter__rules__isort__tests__line_ending_crlf.py.snap

@@ -0,0 +1,23 @@
+---
+source: crates/ruff_linter/src/rules/isort/mod.rs
+---
+line_ending_crlf.py:1:1: I001 [*] Import block is un-sorted or un-formatted
+  |
+1 | / from long_module_name import member_one, member_two, member_three, member_four, member_five
+2 | | 
+  | |_^ I001
+  |
+  = help: Organize imports
+
+ℹ Safe fix
+1   |-from long_module_name import member_one, member_two, member_three, member_four, member_five
+  1 |+from long_module_name import (
+  2 |+    member_five,
+  3 |+    member_four,
+  4 |+    member_one,
+  5 |+    member_three,
+  6 |+    member_two,
+  7 |+)
+2 8 | 
+
+

crates/ruff_linter/src/rules/isort/snapshots/ruff_linter__rules__isort__tests__line_ending_lf.py.snap

@@ -0,0 +1,23 @@
+---
+source: crates/ruff_linter/src/rules/isort/mod.rs
+---
+line_ending_lf.py:1:1: I001 [*] Import block is un-sorted or un-formatted
+  |
+1 | / from long_module_name import member_one, member_two, member_three, member_four, member_five
+2 | | 
+  | |_^ I001
+  |
+  = help: Organize imports
+
+ℹ Safe fix
+1   |-from long_module_name import member_one, member_two, member_three, member_four, member_five
+  1 |+from long_module_name import (
+  2 |+    member_five,
+  3 |+    member_four,
+  4 |+    member_one,
+  5 |+    member_three,
+  6 |+    member_two,
+  7 |+)
+2 8 | 
+
+

crates/ruff_linter/src/rules/pandas_vet/helpers.rs

@@ -47,6 +47,7 @@ pub(super) fn test_expression(expr: &Expr, semantic: &SemanticModel) -> Resoluti
                         | BindingKind::Assignment
                         | BindingKind::NamedExprAssignment
                         | BindingKind::LoopVar
+                        | BindingKind::ComprehensionVar
                         | BindingKind::Global
                         | BindingKind::Nonlocal(_) => Resolution::RelevantLocal,
                         BindingKind::Import(import) if matches!(import.call_path(), ["pandas"]) => {

crates/ruff_linter/src/rules/perflint/rules/unnecessary_list_cast.rs

@@ -1,5 +1,6 @@
 use ruff_diagnostics::{AlwaysFixableViolation, Diagnostic, Edit, Fix};
 use ruff_macros::{derive_message_formats, violation};
+use ruff_python_ast::statement_visitor::{walk_stmt, StatementVisitor};
 use ruff_python_ast::{self as ast, Arguments, Expr, Stmt};
 use ruff_python_semantic::analyze::typing::find_assigned_value;
 use ruff_text_size::TextRange;
@@ -98,22 +99,25 @@ pub(crate) fn unnecessary_list_cast(checker: &mut Checker, iter: &Expr, body: &[
             range: iterable_range,
             ..
         }) => {
-            // If the variable is being appended to, don't suggest removing the cast:
-            //
-            // ```python
-            // items = ["foo", "bar"]
-            // for item in list(items):
-            //    items.append("baz")
-            // ```
-            //
-            // Here, removing the `list()` cast would change the behavior of the code.
-            if body.iter().any(|stmt| match_append(stmt, id)) {
-                return;
-            }
             let Some(value) = find_assigned_value(id, checker.semantic()) else {
                 return;
             };
             if matches!(value, Expr::Tuple(_) | Expr::List(_) | Expr::Set(_)) {
+                // If the variable is being modified to, don't suggest removing the cast:
+                //
+                // ```python
+                // items = ["foo", "bar"]
+                // for item in list(items):
+                //    items.append("baz")
+                // ```
+                //
+                // Here, removing the `list()` cast would change the behavior of the code.
+                let mut visitor = MutationVisitor::new(id);
+                visitor.visit_body(body);
+                if visitor.is_mutated {
+                    return;
+                }
+
                 let mut diagnostic = Diagnostic::new(UnnecessaryListCast, *list_range);
                 diagnostic.set_fix(remove_cast(*list_range, *iterable_range));
                 checker.diagnostics.push(diagnostic);
@@ -123,28 +127,6 @@ pub(crate) fn unnecessary_list_cast(checker: &mut Checker, iter: &Expr, body: &[
     }
 }
 
-/// Check if a statement is an `append` call to a given identifier.
-///
-/// For example, `foo.append(bar)` would return `true` if `id` is `foo`.
-fn match_append(stmt: &Stmt, id: &str) -> bool {
-    let Some(ast::StmtExpr { value, .. }) = stmt.as_expr_stmt() else {
-        return false;
-    };
-    let Some(ast::ExprCall { func, .. }) = value.as_call_expr() else {
-        return false;
-    };
-    let Some(ast::ExprAttribute { value, attr, .. }) = func.as_attribute_expr() else {
-        return false;
-    };
-    if attr != "append" {
-        return false;
-    }
-    let Some(ast::ExprName { id: target_id, .. }) = value.as_name_expr() else {
-        return false;
-    };
-    target_id == id
-}
-
 /// Generate a [`Fix`] to remove a `list` cast from an expression.
 fn remove_cast(list_range: TextRange, iterable_range: TextRange) -> Fix {
     Fix::safe_edits(
@@ -152,3 +134,95 @@ fn remove_cast(list_range: TextRange, iterable_range: TextRange) -> Fix {
         [Edit::deletion(iterable_range.end(), list_range.end())],
     )
 }
+
+/// A [`StatementVisitor`] that (conservatively) identifies mutations to a variable.
+#[derive(Default)]
+pub(crate) struct MutationVisitor<'a> {
+    pub(crate) target: &'a str,
+    pub(crate) is_mutated: bool,
+}
+
+impl<'a> MutationVisitor<'a> {
+    pub(crate) fn new(target: &'a str) -> Self {
+        Self {
+            target,
+            is_mutated: false,
+        }
+    }
+}
+
+impl<'a, 'b> StatementVisitor<'b> for MutationVisitor<'a>
+where
+    'b: 'a,
+{
+    fn visit_stmt(&mut self, stmt: &'b Stmt) {
+        if match_mutation(stmt, self.target) {
+            self.is_mutated = true;
+        } else {
+            walk_stmt(self, stmt);
+        }
+    }
+}
+
+/// Check if a statement is (probably) a modification to the list assigned to the given identifier.
+///
+/// For example, `foo.append(bar)` would return `true` if `id` is `foo`.
+fn match_mutation(stmt: &Stmt, id: &str) -> bool {
+    match stmt {
+        // Ex) `foo.append(bar)`
+        Stmt::Expr(ast::StmtExpr { value, .. }) => {
+            let Some(ast::ExprCall { func, .. }) = value.as_call_expr() else {
+                return false;
+            };
+            let Some(ast::ExprAttribute { value, attr, .. }) = func.as_attribute_expr() else {
+                return false;
+            };
+            if !matches!(
+                attr.as_str(),
+                "append" | "insert" | "extend" | "remove" | "pop" | "clear" | "reverse" | "sort"
+            ) {
+                return false;
+            }
+            let Some(ast::ExprName { id: target_id, .. }) = value.as_name_expr() else {
+                return false;
+            };
+            target_id == id
+        }
+        // Ex) `foo[0] = bar`
+        Stmt::Assign(ast::StmtAssign { targets, .. }) => targets.iter().any(|target| {
+            if let Some(ast::ExprSubscript { value: target, .. }) = target.as_subscript_expr() {
+                if let Some(ast::ExprName { id: target_id, .. }) = target.as_name_expr() {
+                    return target_id == id;
+                }
+            }
+            false
+        }),
+        // Ex) `foo += bar`
+        Stmt::AugAssign(ast::StmtAugAssign { target, .. }) => {
+            if let Some(ast::ExprName { id: target_id, .. }) = target.as_name_expr() {
+                target_id == id
+            } else {
+                false
+            }
+        }
+        // Ex) `foo[0]: int = bar`
+        Stmt::AnnAssign(ast::StmtAnnAssign { target, .. }) => {
+            if let Some(ast::ExprSubscript { value: target, .. }) = target.as_subscript_expr() {
+                if let Some(ast::ExprName { id: target_id, .. }) = target.as_name_expr() {
+                    return target_id == id;
+                }
+            }
+            false
+        }
+        // Ex) `del foo[0]`
+        Stmt::Delete(ast::StmtDelete { targets, .. }) => targets.iter().any(|target| {
+            if let Some(ast::ExprSubscript { value: target, .. }) = target.as_subscript_expr() {
+                if let Some(ast::ExprName { id: target_id, .. }) = target.as_name_expr() {
+                    return target_id == id;
+                }
+            }
+            false
+        }),
+        _ => false,
+    }
+}

crates/ruff_linter/src/rules/perflint/snapshots/ruff_linter__rules__perflint__tests__PERF101_PERF101.py.snap

@@ -178,7 +178,7 @@ PERF101.py:34:10: PERF101 [*] Do not cast an iterable to `list` before iterating
    34 |+for i in {1, 2, 3}:  # PERF101
 37 35 |     pass
 38 36 | 
-39 37 | for i in list(foo_dict):  # Ok
+39 37 | for i in list(foo_dict):  # OK
 
 PERF101.py:57:10: PERF101 [*] Do not cast an iterable to `list` before iterating over it
    |
@@ -192,7 +192,7 @@ PERF101.py:57:10: PERF101 [*] Do not cast an iterable to `list` before iterating
    = help: Remove `list()` cast
 
 ℹ Safe fix
-54 54 | for i in list(foo_list):  # Ok
+54 54 | for i in list(foo_list):  # OK
 55 55 |     foo_list.append(i + 1)
 56 56 | 
 57    |-for i in list(foo_list):  # PERF101
@@ -218,5 +218,7 @@ PERF101.py:69:10: PERF101 [*] Do not cast an iterable to `list` before iterating
 69    |-for i in list(nested_tuple):  # PERF101
    69 |+for i in nested_tuple:  # PERF101
 70 70 |     pass
+71 71 | 
+72 72 | for i in list(foo_list):  # OK
 
 

crates/ruff_linter/src/rules/pydocstyle/rules/indent.rs

@@ -87,12 +87,17 @@ impl Violation for IndentWithSpaces {
 ///     """
 /// ```
 ///
+/// ## Formatter compatibility
+/// We recommend against using this rule alongside the [formatter]. The
+/// formatter enforces consistent indentation, making the rule redundant.
+///
 /// ## References
 /// - [PEP 257 – Docstring Conventions](https://peps.python.org/pep-0257/)
 /// - [NumPy Style Guide](https://numpydoc.readthedocs.io/en/latest/format.html)
 /// - [Google Python Style Guide - Docstrings](https://google.github.io/styleguide/pyguide.html#38-comments-and-docstrings)
 ///
 /// [PEP 257]: https://peps.python.org/pep-0257/
+/// [formatter]: https://docs.astral.sh/ruff/formatter/
 #[violation]
 pub struct UnderIndentation;
 

crates/ruff_linter/src/rules/pydocstyle/rules/triple_quotes.rs

@@ -27,10 +27,16 @@ use crate::docstrings::Docstring;
 ///     """Return the pathname of the KOS root directory."""
 /// ```
 ///
+/// ## Formatter compatibility
+/// We recommend against using this rule alongside the [formatter]. The
+/// formatter enforces consistent quotes, making the rule redundant.
+///
 /// ## References
 /// - [PEP 257 – Docstring Conventions](https://peps.python.org/pep-0257/)
 /// - [NumPy Style Guide](https://numpydoc.readthedocs.io/en/latest/format.html)
 /// - [Google Python Style Guide - Docstrings](https://google.github.io/styleguide/pyguide.html#38-comments-and-docstrings)
+///
+/// [formatter]: https://docs.astral.sh/ruff/formatter/
 #[violation]
 pub struct TripleSingleQuotes {
     expected_quote: Quote,

crates/ruff_linter/src/rules/pyflakes/rules/unused_import.rs

@@ -28,6 +28,15 @@ enum UnusedImportContext {
 /// If an import statement is used to check for the availability or existence
 /// of a module, consider using `importlib.util.find_spec` instead.
 ///
+/// If an import statement is used to re-export a symbol as part of a module's
+/// public interface, consider using a "redundant" import alias, which
+/// instructs Ruff (and other tools) to respect the re-export, and avoid
+/// marking it as unused, as in:
+///
+/// ```python
+/// from module import member as member
+/// ```
+///
 /// ## Example
 /// ```python
 /// import numpy as np  # unused import
@@ -54,11 +63,12 @@ enum UnusedImportContext {
 /// ```
 ///
 /// ## Options
-/// - `lint.pyflakes.extend-generics`
+/// - `lint.ignore-init-module-imports`
 ///
 /// ## References
 /// - [Python documentation: `import`](https://docs.python.org/3/reference/simple_stmts.html#the-import-statement)
 /// - [Python documentation: `importlib.util.find_spec`](https://docs.python.org/3/library/importlib.html#importlib.util.find_spec)
+/// - [Typing documentation: interface conventions](https://typing.readthedocs.io/en/latest/source/libraries.html#library-interface-public-and-private-symbols)
 #[violation]
 pub struct UnusedImport {
     name: String,

crates/ruff_linter/src/rules/pylint/rules/literal_membership.rs

@@ -1,6 +1,7 @@
 use ruff_diagnostics::{AlwaysFixableViolation, Diagnostic, Edit, Fix};
 use ruff_macros::{derive_message_formats, violation};
 use ruff_python_ast::{self as ast, CmpOp, Expr};
+use ruff_python_semantic::analyze::typing;
 use ruff_text_size::Ranged;
 
 use crate::checkers::ast::Checker;
@@ -25,7 +26,8 @@ use crate::checkers::ast::Checker;
 /// ## Fix safety
 /// This rule's fix is marked as unsafe, as the use of a `set` literal will
 /// error at runtime if the sequence contains unhashable elements (like lists
-/// or dictionaries).
+/// or dictionaries). While Ruff will attempt to infer the hashability of the
+/// elements, it may not always be able to do so.
 ///
 /// ## References
 /// - [What’s New In Python 3.2](https://docs.python.org/3/whatsnew/3.2.html#optimizations)
@@ -57,7 +59,40 @@ pub(crate) fn literal_membership(checker: &mut Checker, compare: &ast::ExprCompa
         return;
     };
 
-    if !matches!(right, Expr::List(_) | Expr::Tuple(_)) {
+    let elts = match right {
+        Expr::List(ast::ExprList { elts, .. }) => elts,
+        Expr::Tuple(ast::ExprTuple { elts, .. }) => elts,
+        _ => return,
+    };
+
+    // If `left`, or any of the elements in `right`, are known to _not_ be hashable, return.
+    if std::iter::once(compare.left.as_ref())
+        .chain(elts)
+        .any(|expr| match expr {
+            // Expressions that are known _not_ to be hashable.
+            Expr::List(_)
+            | Expr::Set(_)
+            | Expr::Dict(_)
+            | Expr::ListComp(_)
+            | Expr::SetComp(_)
+            | Expr::DictComp(_)
+            | Expr::GeneratorExp(_)
+            | Expr::Await(_)
+            | Expr::Yield(_)
+            | Expr::YieldFrom(_) => true,
+            // Expressions that can be _inferred_ not to be hashable.
+            Expr::Name(name) => {
+                let Some(id) = checker.semantic().resolve_name(name) else {
+                    return false;
+                };
+                let binding = checker.semantic().binding(id);
+                typing::is_list(binding, checker.semantic())
+                    || typing::is_dict(binding, checker.semantic())
+                    || typing::is_set(binding, checker.semantic())
+            }
+            _ => false,
+        })
+    {
         return;
     }
 

crates/ruff_linter/src/rules/pylint/rules/magic_value_comparison.rs

@@ -86,8 +86,10 @@ fn is_magic_value(literal_expr: LiteralExpressionRef, allowed_types: &[ConstantT
             !matches!(value.to_str(), "" | "__main__")
         }
         LiteralExpressionRef::NumberLiteral(ast::ExprNumberLiteral { value, .. }) => match value {
+            #[allow(clippy::float_cmp)]
+            ast::Number::Float(value) => !(*value == 0.0 || *value == 1.0),
             ast::Number::Int(value) => !matches!(*value, Int::ZERO | Int::ONE),
-            _ => true,
+            ast::Number::Complex { .. } => true,
         },
         LiteralExpressionRef::BytesLiteral(_) => true,
     }

crates/ruff_linter/src/rules/pylint/rules/non_ascii_name.rs

@@ -52,6 +52,7 @@ pub(crate) fn non_ascii_name(binding: &Binding, locator: &Locator) -> Option<Dia
         BindingKind::Assignment => Kind::Assignment,
         BindingKind::TypeParam => Kind::TypeParam,
         BindingKind::LoopVar => Kind::LoopVar,
+        BindingKind::ComprehensionVar => Kind::ComprenhensionVar,
         BindingKind::WithItemVar => Kind::WithItemVar,
         BindingKind::Global => Kind::Global,
         BindingKind::Nonlocal(_) => Kind::Nonlocal,
@@ -88,6 +89,7 @@ enum Kind {
     Assignment,
     TypeParam,
     LoopVar,
+    ComprenhensionVar,
     WithItemVar,
     Global,
     Nonlocal,
@@ -105,6 +107,7 @@ impl fmt::Display for Kind {
             Kind::Assignment => f.write_str("Variable"),
             Kind::TypeParam => f.write_str("Type parameter"),
             Kind::LoopVar => f.write_str("Variable"),
+            Kind::ComprenhensionVar => f.write_str("Variable"),
             Kind::WithItemVar => f.write_str("Variable"),
             Kind::Global => f.write_str("Global"),
             Kind::Nonlocal => f.write_str("Nonlocal"),

crates/ruff_linter/src/rules/pylint/snapshots/ruff_linter__rules__pylint__tests__PLR2004_magic_value_comparison.py.snap

@@ -10,49 +10,67 @@ magic_value_comparison.py:5:4: PLR2004 Magic value used in comparison, consider
 6 |     pass
   |
 
-magic_value_comparison.py:38:12: PLR2004 Magic value used in comparison, consider replacing `2` with a constant variable
+magic_value_comparison.py:47:12: PLR2004 Magic value used in comparison, consider replacing `2` with a constant variable
    |
-36 |     pass
-37 | 
-38 | if argc != 2:  # [magic-value-comparison]
-   |            ^ PLR2004
-39 |     pass
-   |
-
-magic_value_comparison.py:41:12: PLR2004 Magic value used in comparison, consider replacing `-2` with a constant variable
-   |
-39 |     pass
-40 | 
-41 | if argc != -2:  # [magic-value-comparison]
-   |            ^^ PLR2004
-42 |     pass
-   |
-
-magic_value_comparison.py:44:12: PLR2004 Magic value used in comparison, consider replacing `+2` with a constant variable
-   |
-42 |     pass
-43 | 
-44 | if argc != +2:  # [magic-value-comparison]
-   |            ^^ PLR2004
 45 |     pass
+46 | 
+47 | if argc != 2:  # [magic-value-comparison]
+   |            ^ PLR2004
+48 |     pass
    |
 
-magic_value_comparison.py:65:21: PLR2004 Magic value used in comparison, consider replacing `3.141592653589793238` with a constant variable
+magic_value_comparison.py:50:12: PLR2004 Magic value used in comparison, consider replacing `-2` with a constant variable
    |
-63 | pi_estimation = 3.14
-64 | 
-65 | if pi_estimation == 3.141592653589793238:  # [magic-value-comparison]
+48 |     pass
+49 | 
+50 | if argc != -2:  # [magic-value-comparison]
+   |            ^^ PLR2004
+51 |     pass
+   |
+
+magic_value_comparison.py:53:12: PLR2004 Magic value used in comparison, consider replacing `+2` with a constant variable
+   |
+51 |     pass
+52 | 
+53 | if argc != +2:  # [magic-value-comparison]
+   |            ^^ PLR2004
+54 |     pass
+   |
+
+magic_value_comparison.py:56:12: PLR2004 Magic value used in comparison, consider replacing `-2.0` with a constant variable
+   |
+54 |     pass
+55 | 
+56 | if argc != -2.0:  # [magic-value-comparison]
+   |            ^^^^ PLR2004
+57 |     pass
+   |
+
+magic_value_comparison.py:59:12: PLR2004 Magic value used in comparison, consider replacing `+2.0` with a constant variable
+   |
+57 |     pass
+58 | 
+59 | if argc != +2.0:  # [magic-value-comparison]
+   |            ^^^^ PLR2004
+60 |     pass
+   |
+
+magic_value_comparison.py:80:21: PLR2004 Magic value used in comparison, consider replacing `3.141592653589793238` with a constant variable
+   |
+78 | pi_estimation = 3.14
+79 | 
+80 | if pi_estimation == 3.141592653589793238:  # [magic-value-comparison]
    |                     ^^^^^^^^^^^^^^^^^^^^ PLR2004
-66 |     pass
+81 |     pass
    |
 
-magic_value_comparison.py:71:21: PLR2004 Magic value used in comparison, consider replacing `0x3` with a constant variable
+magic_value_comparison.py:86:21: PLR2004 Magic value used in comparison, consider replacing `0x3` with a constant variable
    |
-69 |     pass
-70 | 
-71 | if pi_estimation == 0x3:  # [magic-value-comparison]
+84 |     pass
+85 | 
+86 | if pi_estimation == 0x3:  # [magic-value-comparison]
    |                     ^^^ PLR2004
-72 |     pass
+87 |     pass
    |
 
 

crates/ruff_linter/src/rules/pylint/snapshots/ruff_linter__rules__pylint__tests__PLR6201_literal_membership.py.snap

@@ -48,8 +48,8 @@ literal_membership.py:4:6: PLR6201 [*] Use a `set` literal when testing for memb
 5 | |     1, 2, 3
 6 | | )
   | |_^ PLR6201
-7 |   
-8 |   # OK
+7 |   fruits = ["cherry", "grapes"]
+8 |   "cherry" in fruits
   |
   = help: Convert to `set`
 
@@ -62,8 +62,29 @@ literal_membership.py:4:6: PLR6201 [*] Use a `set` literal when testing for memb
 5 5 |     1, 2, 3
 6   |-)
   6 |+}
-7 7 | 
-8 8 | # OK
-9 9 | fruits = ["cherry", "grapes"]
+7 7 | fruits = ["cherry", "grapes"]
+8 8 | "cherry" in fruits
+9 9 | _ = {key: value for key, value in {"a": 1, "b": 2}.items() if key in ("a", "b")}
+
+literal_membership.py:9:70: PLR6201 [*] Use a `set` literal when testing for membership
+   |
+ 7 | fruits = ["cherry", "grapes"]
+ 8 | "cherry" in fruits
+ 9 | _ = {key: value for key, value in {"a": 1, "b": 2}.items() if key in ("a", "b")}
+   |                                                                      ^^^^^^^^^^ PLR6201
+10 | 
+11 | # OK
+   |
+   = help: Convert to `set`
+
+ℹ Unsafe fix
+6  6  | )
+7  7  | fruits = ["cherry", "grapes"]
+8  8  | "cherry" in fruits
+9     |-_ = {key: value for key, value in {"a": 1, "b": 2}.items() if key in ("a", "b")}
+   9  |+_ = {key: value for key, value in {"a": 1, "b": 2}.items() if key in {"a", "b"}}
+10 10 | 
+11 11 | # OK
+12 12 | fruits in [[1, 2, 3], [4, 5, 6]]
 
 

crates/ruff_linter/src/rules/pylint/snapshots/ruff_linter__rules__pylint__tests__allow_magic_value_types.snap

@@ -1,31 +1,49 @@
 ---
 source: crates/ruff_linter/src/rules/pylint/mod.rs
 ---
-magic_value_comparison.py:59:22: PLR2004 Magic value used in comparison, consider replacing `"Hunter2"` with a constant variable
+magic_value_comparison.py:56:12: PLR2004 Magic value used in comparison, consider replacing `-2.0` with a constant variable
+   |
+54 |     pass
+55 | 
+56 | if argc != -2.0:  # [magic-value-comparison]
+   |            ^^^^ PLR2004
+57 |     pass
+   |
+
+magic_value_comparison.py:59:12: PLR2004 Magic value used in comparison, consider replacing `+2.0` with a constant variable
    |
 57 |     pass
 58 | 
-59 | if input_password == "Hunter2":  # correct
-   |                      ^^^^^^^^^ PLR2004
+59 | if argc != +2.0:  # [magic-value-comparison]
+   |            ^^^^ PLR2004
 60 |     pass
    |
 
-magic_value_comparison.py:65:21: PLR2004 Magic value used in comparison, consider replacing `3.141592653589793238` with a constant variable
+magic_value_comparison.py:74:22: PLR2004 Magic value used in comparison, consider replacing `"Hunter2"` with a constant variable
    |
-63 | pi_estimation = 3.14
-64 | 
-65 | if pi_estimation == 3.141592653589793238:  # [magic-value-comparison]
+72 |     pass
+73 | 
+74 | if input_password == "Hunter2":  # correct
+   |                      ^^^^^^^^^ PLR2004
+75 |     pass
+   |
+
+magic_value_comparison.py:80:21: PLR2004 Magic value used in comparison, consider replacing `3.141592653589793238` with a constant variable
+   |
+78 | pi_estimation = 3.14
+79 | 
+80 | if pi_estimation == 3.141592653589793238:  # [magic-value-comparison]
    |                     ^^^^^^^^^^^^^^^^^^^^ PLR2004
-66 |     pass
+81 |     pass
    |
 
-magic_value_comparison.py:77:18: PLR2004 Magic value used in comparison, consider replacing `b"something"` with a constant variable
+magic_value_comparison.py:92:18: PLR2004 Magic value used in comparison, consider replacing `b"something"` with a constant variable
    |
-75 | user_input = b"Hello, There!"
-76 | 
-77 | if user_input == b"something":  # correct
+90 | user_input = b"Hello, There!"
+91 | 
+92 | if user_input == b"something":  # correct
    |                  ^^^^^^^^^^^^ PLR2004
-78 |     pass
+93 |     pass
    |
 
 

crates/ruff_linter/src/rules/refurb/mod.rs

@@ -17,6 +17,7 @@ mod tests {
     #[test_case(Rule::ReadWholeFile, Path::new("FURB101.py"))]
     #[test_case(Rule::RepeatedAppend, Path::new("FURB113.py"))]
     #[test_case(Rule::ReimplementedOperator, Path::new("FURB118.py"))]
+    #[test_case(Rule::ReadlinesInFor, Path::new("FURB129.py"))]
     #[test_case(Rule::DeleteFullSlice, Path::new("FURB131.py"))]
     #[test_case(Rule::CheckAndRemoveFromSet, Path::new("FURB132.py"))]
     #[test_case(Rule::IfExprMinMax, Path::new("FURB136.py"))]

crates/ruff_linter/src/rules/refurb/rules/mod.rs

@@ -9,6 +9,7 @@ pub(crate) use math_constant::*;
 pub(crate) use metaclass_abcmeta::*;
 pub(crate) use print_empty_string::*;
 pub(crate) use read_whole_file::*;
+pub(crate) use readlines_in_for::*;
 pub(crate) use redundant_log_base::*;
 pub(crate) use regex_flag_alias::*;
 pub(crate) use reimplemented_operator::*;
@@ -30,6 +31,7 @@ mod math_constant;
 mod metaclass_abcmeta;
 mod print_empty_string;
 mod read_whole_file;
+mod readlines_in_for;
 mod redundant_log_base;
 mod regex_flag_alias;
 mod reimplemented_operator;

crates/ruff_linter/src/rules/refurb/rules/readlines_in_for.rs

@@ -0,0 +1,92 @@
+use ruff_diagnostics::{AlwaysFixableViolation, Diagnostic, Edit, Fix};
+use ruff_macros::{derive_message_formats, violation};
+use ruff_python_ast::{Comprehension, Expr, StmtFor};
+use ruff_python_semantic::analyze::typing;
+use ruff_python_semantic::analyze::typing::is_io_base_expr;
+use ruff_text_size::Ranged;
+
+use crate::checkers::ast::Checker;
+
+/// ## What it does
+/// Checks for uses of `readlines()` when iterating over a file line-by-line.
+///
+/// ## Why is this bad?
+/// Rather than iterating over all lines in a file by calling `readlines()`,
+/// it's more convenient and performant to iterate over the file object
+/// directly.
+///
+/// ## Example
+/// ```python
+/// with open("file.txt") as fp:
+///     for line in fp.readlines():
+///         ...
+/// ```
+///
+/// Use instead:
+/// ```python
+/// with open("file.txt") as fp:
+///     for line in fp:
+///         ...
+/// ```
+///
+/// ## References
+/// - [Python documentation: `io.IOBase.readlines`](https://docs.python.org/3/library/io.html#io.IOBase.readlines)
+#[violation]
+pub(crate) struct ReadlinesInFor;
+
+impl AlwaysFixableViolation for ReadlinesInFor {
+    #[derive_message_formats]
+    fn message(&self) -> String {
+        format!("Instead of calling `readlines()`, iterate over file object directly")
+    }
+
+    fn fix_title(&self) -> String {
+        "Remove `readlines()`".into()
+    }
+}
+
+/// FURB129
+pub(crate) fn readlines_in_for(checker: &mut Checker, for_stmt: &StmtFor) {
+    readlines_in_iter(checker, for_stmt.iter.as_ref());
+}
+
+/// FURB129
+pub(crate) fn readlines_in_comprehension(checker: &mut Checker, comprehension: &Comprehension) {
+    readlines_in_iter(checker, &comprehension.iter);
+}
+
+fn readlines_in_iter(checker: &mut Checker, iter_expr: &Expr) {
+    let Expr::Call(expr_call) = iter_expr else {
+        return;
+    };
+
+    let Expr::Attribute(expr_attr) = expr_call.func.as_ref() else {
+        return;
+    };
+
+    if expr_attr.attr.as_str() != "readlines" || !expr_call.arguments.is_empty() {
+        return;
+    }
+
+    // Determine whether `fp` in `fp.readlines()` was bound to a file object.
+    if let Expr::Name(name) = expr_attr.value.as_ref() {
+        if !checker
+            .semantic()
+            .resolve_name(name)
+            .map(|id| checker.semantic().binding(id))
+            .is_some_and(|binding| typing::is_io_base(binding, checker.semantic()))
+        {
+            return;
+        }
+    } else {
+        if !is_io_base_expr(expr_attr.value.as_ref(), checker.semantic()) {
+            return;
+        }
+    }
+
+    let mut diagnostic = Diagnostic::new(ReadlinesInFor, expr_call.range());
+    diagnostic.set_fix(Fix::unsafe_edit(Edit::range_deletion(
+        expr_call.range().add_start(expr_attr.value.range().len()),
+    )));
+    checker.diagnostics.push(diagnostic);
+}

crates/ruff_linter/src/rules/refurb/snapshots/ruff_linter__rules__refurb__tests__FURB129_FURB129.py.snap

@@ -0,0 +1,207 @@
+---
+source: crates/ruff_linter/src/rules/refurb/mod.rs
+---
+FURB129.py:7:18: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+  |
+5 | # Errors
+6 | with open("FURB129.py") as f:
+7 |     for _line in f.readlines():
+  |                  ^^^^^^^^^^^^^ FURB129
+8 |         pass
+9 |     a = [line.lower() for line in f.readlines()]
+  |
+  = help: Remove `readlines()`
+
+ℹ Unsafe fix
+4 4 | 
+5 5 | # Errors
+6 6 | with open("FURB129.py") as f:
+7   |-    for _line in f.readlines():
+  7 |+    for _line in f:
+8 8 |         pass
+9 9 |     a = [line.lower() for line in f.readlines()]
+10 10 |     b = {line.upper() for line in f.readlines()}
+
+FURB129.py:9:35: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+   |
+ 7 |     for _line in f.readlines():
+ 8 |         pass
+ 9 |     a = [line.lower() for line in f.readlines()]
+   |                                   ^^^^^^^^^^^^^ FURB129
+10 |     b = {line.upper() for line in f.readlines()}
+11 |     c = {line.lower(): line.upper() for line in f.readlines()}
+   |
+   = help: Remove `readlines()`
+
+ℹ Unsafe fix
+6  6  | with open("FURB129.py") as f:
+7  7  |     for _line in f.readlines():
+8  8  |         pass
+9     |-    a = [line.lower() for line in f.readlines()]
+   9  |+    a = [line.lower() for line in f]
+10 10 |     b = {line.upper() for line in f.readlines()}
+11 11 |     c = {line.lower(): line.upper() for line in f.readlines()}
+12 12 | 
+
+FURB129.py:10:35: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+   |
+ 8 |         pass
+ 9 |     a = [line.lower() for line in f.readlines()]
+10 |     b = {line.upper() for line in f.readlines()}
+   |                                   ^^^^^^^^^^^^^ FURB129
+11 |     c = {line.lower(): line.upper() for line in f.readlines()}
+   |
+   = help: Remove `readlines()`
+
+ℹ Unsafe fix
+7  7  |     for _line in f.readlines():
+8  8  |         pass
+9  9  |     a = [line.lower() for line in f.readlines()]
+10    |-    b = {line.upper() for line in f.readlines()}
+   10 |+    b = {line.upper() for line in f}
+11 11 |     c = {line.lower(): line.upper() for line in f.readlines()}
+12 12 | 
+13 13 | with Path("FURB129.py").open() as f:
+
+FURB129.py:11:49: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+   |
+ 9 |     a = [line.lower() for line in f.readlines()]
+10 |     b = {line.upper() for line in f.readlines()}
+11 |     c = {line.lower(): line.upper() for line in f.readlines()}
+   |                                                 ^^^^^^^^^^^^^ FURB129
+12 | 
+13 | with Path("FURB129.py").open() as f:
+   |
+   = help: Remove `readlines()`
+
+ℹ Unsafe fix
+8  8  |         pass
+9  9  |     a = [line.lower() for line in f.readlines()]
+10 10 |     b = {line.upper() for line in f.readlines()}
+11    |-    c = {line.lower(): line.upper() for line in f.readlines()}
+   11 |+    c = {line.lower(): line.upper() for line in f}
+12 12 | 
+13 13 | with Path("FURB129.py").open() as f:
+14 14 |     for _line in f.readlines():
+
+FURB129.py:14:18: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+   |
+13 | with Path("FURB129.py").open() as f:
+14 |     for _line in f.readlines():
+   |                  ^^^^^^^^^^^^^ FURB129
+15 |         pass
+   |
+   = help: Remove `readlines()`
+
+ℹ Unsafe fix
+11 11 |     c = {line.lower(): line.upper() for line in f.readlines()}
+12 12 | 
+13 13 | with Path("FURB129.py").open() as f:
+14    |-    for _line in f.readlines():
+   14 |+    for _line in f:
+15 15 |         pass
+16 16 | 
+17 17 | for _line in open("FURB129.py").readlines():
+
+FURB129.py:17:14: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+   |
+15 |         pass
+16 | 
+17 | for _line in open("FURB129.py").readlines():
+   |              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ FURB129
+18 |     pass
+   |
+   = help: Remove `readlines()`
+
+ℹ Unsafe fix
+14 14 |     for _line in f.readlines():
+15 15 |         pass
+16 16 | 
+17    |-for _line in open("FURB129.py").readlines():
+   17 |+for _line in open("FURB129.py"):
+18 18 |     pass
+19 19 | 
+20 20 | for _line in Path("FURB129.py").open().readlines():
+
+FURB129.py:20:14: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+   |
+18 |     pass
+19 | 
+20 | for _line in Path("FURB129.py").open().readlines():
+   |              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ FURB129
+21 |     pass
+   |
+   = help: Remove `readlines()`
+
+ℹ Unsafe fix
+17 17 | for _line in open("FURB129.py").readlines():
+18 18 |     pass
+19 19 | 
+20    |-for _line in Path("FURB129.py").open().readlines():
+   20 |+for _line in Path("FURB129.py").open():
+21 21 |     pass
+22 22 | 
+23 23 | 
+
+FURB129.py:26:18: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+   |
+24 | def func():
+25 |     f = Path("FURB129.py").open()
+26 |     for _line in f.readlines():
+   |                  ^^^^^^^^^^^^^ FURB129
+27 |         pass
+28 |     f.close()
+   |
+   = help: Remove `readlines()`
+
+ℹ Unsafe fix
+23 23 | 
+24 24 | def func():
+25 25 |     f = Path("FURB129.py").open()
+26    |-    for _line in f.readlines():
+   26 |+    for _line in f:
+27 27 |         pass
+28 28 |     f.close()
+29 29 | 
+
+FURB129.py:32:18: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+   |
+31 | def func(f: io.BytesIO):
+32 |     for _line in f.readlines():
+   |                  ^^^^^^^^^^^^^ FURB129
+33 |         pass
+   |
+   = help: Remove `readlines()`
+
+ℹ Unsafe fix
+29 29 | 
+30 30 | 
+31 31 | def func(f: io.BytesIO):
+32    |-    for _line in f.readlines():
+   32 |+    for _line in f:
+33 33 |         pass
+34 34 | 
+35 35 | 
+
+FURB129.py:38:22: FURB129 [*] Instead of calling `readlines()`, iterate over file object directly
+   |
+36 | def func():
+37 |     with (open("FURB129.py") as f, foo as bar):
+38 |         for _line in f.readlines():
+   |                      ^^^^^^^^^^^^^ FURB129
+39 |             pass
+40 |         for _line in bar.readlines():
+   |
+   = help: Remove `readlines()`
+
+ℹ Unsafe fix
+35 35 | 
+36 36 | def func():
+37 37 |     with (open("FURB129.py") as f, foo as bar):
+38    |-        for _line in f.readlines():
+   38 |+        for _line in f:
+39 39 |             pass
+40 40 |         for _line in bar.readlines():
+41 41 |             pass
+
+

crates/ruff_linter/src/rules/ruff/mod.rs

@@ -48,6 +48,7 @@ mod tests {
     #[test_case(Rule::DefaultFactoryKwarg, Path::new("RUF026.py"))]
     #[test_case(Rule::MissingFStringSyntax, Path::new("RUF027_0.py"))]
     #[test_case(Rule::MissingFStringSyntax, Path::new("RUF027_1.py"))]
+    #[test_case(Rule::MissingFStringSyntax, Path::new("RUF027_2.py"))]
     fn rules(rule_code: Rule, path: &Path) -> Result<()> {
         let snapshot = format!("{}_{}", rule_code.noqa_code(), path.to_string_lossy());
         let diagnostics = test_path(

crates/ruff_linter/src/rules/ruff/rules/ambiguous_unicode_character.rs

@@ -4,9 +4,11 @@ use bitflags::bitflags;
 
 use ruff_diagnostics::{Diagnostic, DiagnosticKind, Violation};
 use ruff_macros::{derive_message_formats, violation};
+use ruff_python_ast::StringLike;
 use ruff_source_file::Locator;
-use ruff_text_size::{TextLen, TextRange, TextSize};
+use ruff_text_size::{Ranged, TextLen, TextRange, TextSize};
 
+use crate::checkers::ast::Checker;
 use crate::registry::AsRule;
 use crate::rules::ruff::rules::confusables::confusable;
 use crate::rules::ruff::rules::Context;
@@ -171,16 +173,59 @@ impl Violation for AmbiguousUnicodeCharacterComment {
     }
 }
 
-/// RUF001, RUF002, RUF003
-pub(crate) fn ambiguous_unicode_character(
+/// RUF003
+pub(crate) fn ambiguous_unicode_character_comment(
     diagnostics: &mut Vec<Diagnostic>,
     locator: &Locator,
     range: TextRange,
+    settings: &LinterSettings,
+) {
+    let text = locator.slice(range);
+    ambiguous_unicode_character(diagnostics, text, range, Context::Comment, settings);
+}
+
+/// RUF001, RUF002
+pub(crate) fn ambiguous_unicode_character_string(checker: &mut Checker, string_like: StringLike) {
+    let context = if checker.semantic().in_docstring() {
+        Context::Docstring
+    } else {
+        Context::String
+    };
+
+    match string_like {
+        StringLike::StringLiteral(string_literal) => {
+            for string in &string_literal.value {
+                let text = checker.locator().slice(string);
+                ambiguous_unicode_character(
+                    &mut checker.diagnostics,
+                    text,
+                    string.range(),
+                    context,
+                    checker.settings,
+                );
+            }
+        }
+        StringLike::FStringLiteral(f_string_literal) => {
+            let text = checker.locator().slice(f_string_literal);
+            ambiguous_unicode_character(
+                &mut checker.diagnostics,
+                text,
+                f_string_literal.range(),
+                context,
+                checker.settings,
+            );
+        }
+        StringLike::BytesLiteral(_) => (),
+    }
+}
+
+fn ambiguous_unicode_character(
+    diagnostics: &mut Vec<Diagnostic>,
+    text: &str,
+    range: TextRange,
     context: Context,
     settings: &LinterSettings,
 ) {
-    let text = locator.slice(range);
-
     // Most of the time, we don't need to check for ambiguous unicode characters at all.
     if text.is_ascii() {
         return;

crates/ruff_linter/src/rules/ruff/rules/asyncio_dangling_task.rs

@@ -52,14 +52,15 @@ use ruff_text_size::Ranged;
 /// - [The Python Standard Library](https://docs.python.org/3/library/asyncio-task.html#asyncio.create_task)
 #[violation]
 pub struct AsyncioDanglingTask {
+    expr: String,
     method: Method,
 }
 
 impl Violation for AsyncioDanglingTask {
     #[derive_message_formats]
     fn message(&self) -> String {
-        let AsyncioDanglingTask { method } = self;
-        format!("Store a reference to the return value of `asyncio.{method}`")
+        let AsyncioDanglingTask { expr, method } = self;
+        format!("Store a reference to the return value of `{expr}.{method}`")
     }
 }
 
@@ -80,23 +81,35 @@ pub(crate) fn asyncio_dangling_task(expr: &Expr, semantic: &SemanticModel) -> Op
             })
     {
         return Some(Diagnostic::new(
-            AsyncioDanglingTask { method },
+            AsyncioDanglingTask {
+                expr: "asyncio".to_string(),
+                method,
+            },
             expr.range(),
         ));
     }
 
-    // Ex) `loop = asyncio.get_running_loop(); loop.create_task(...)`
+    // Ex) `loop = ...; loop.create_task(...)`
     if let Expr::Attribute(ast::ExprAttribute { attr, value, .. }) = func.as_ref() {
         if attr == "create_task" {
-            if typing::resolve_assignment(value, semantic).is_some_and(|call_path| {
-                matches!(call_path.as_slice(), ["asyncio", "get_running_loop"])
-            }) {
-                return Some(Diagnostic::new(
-                    AsyncioDanglingTask {
-                        method: Method::CreateTask,
-                    },
-                    expr.range(),
-                ));
+            if let Expr::Name(name) = value.as_ref() {
+                if typing::resolve_assignment(value, semantic).is_some_and(|call_path| {
+                    matches!(
+                        call_path.as_slice(),
+                        [
+                            "asyncio",
+                            "get_event_loop" | "get_running_loop" | "new_event_loop"
+                        ]
+                    )
+                }) {
+                    return Some(Diagnostic::new(
+                        AsyncioDanglingTask {
+                            expr: name.id.to_string(),
+                            method: Method::CreateTask,
+                        },
+                        expr.range(),
+                    ));
+                }
             }
         }
     }

crates/ruff_linter/src/rules/ruff/rules/missing_fstring_syntax.rs

@@ -88,8 +88,10 @@ fn should_be_fstring(
         return false;
     }
 
-    let Ok(ast::Expr::FString(ast::ExprFString { value, .. })) =
-        parse_expression(&format!("f{}", locator.slice(literal.range())))
+    let fstring_expr = format!("f{}", locator.slice(literal));
+
+    // Note: Range offsets for `value` are based on `fstring_expr`
+    let Ok(ast::Expr::FString(ast::ExprFString { value, .. })) = parse_expression(&fstring_expr)
     else {
         return false;
     };
@@ -159,7 +161,7 @@ fn should_be_fstring(
                 has_name = true;
             }
             if let Some(spec) = &element.format_spec {
-                let spec = locator.slice(spec.range());
+                let spec = &fstring_expr[spec.range()];
                 if FormatSpec::parse(spec).is_err() {
                     return false;
                 }

crates/ruff_linter/src/rules/ruff/snapshots/ruff_linter__rules__ruff__tests__RUF006_RUF006.py.snap

@@ -25,7 +25,7 @@ RUF006.py:68:12: RUF006 Store a reference to the return value of `asyncio.create
    |            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ RUF006
    |
 
-RUF006.py:74:26: RUF006 Store a reference to the return value of `asyncio.create_task`
+RUF006.py:74:26: RUF006 Store a reference to the return value of `loop.create_task`
    |
 72 | def f():
 73 |     loop = asyncio.get_running_loop()
@@ -33,7 +33,7 @@ RUF006.py:74:26: RUF006 Store a reference to the return value of `asyncio.create
    |                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ RUF006
    |
 
-RUF006.py:97:5: RUF006 Store a reference to the return value of `asyncio.create_task`
+RUF006.py:97:5: RUF006 Store a reference to the return value of `loop.create_task`
    |
 95 | def f():
 96 |     loop = asyncio.get_running_loop()
@@ -41,4 +41,24 @@ RUF006.py:97:5: RUF006 Store a reference to the return value of `asyncio.create_
    |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ RUF006
    |
 
+RUF006.py:170:5: RUF006 Store a reference to the return value of `loop.create_task`
+    |
+168 | def f():
+169 |     loop = asyncio.new_event_loop()
+170 |     loop.create_task(main()) # Error
+    |     ^^^^^^^^^^^^^^^^^^^^^^^^ RUF006
+171 | 
+172 | # Error
+    |
+
+RUF006.py:175:5: RUF006 Store a reference to the return value of `loop.create_task`
+    |
+173 | def f():
+174 |     loop = asyncio.get_event_loop()
+175 |     loop.create_task(main()) # Error
+    |     ^^^^^^^^^^^^^^^^^^^^^^^^ RUF006
+176 | 
+177 | # OK
+    |
+
 

crates/ruff_linter/src/rules/ruff/snapshots/ruff_linter__rules__ruff__tests__RUF027_RUF027_0.py.snap

@@ -285,6 +285,8 @@ RUF027_0.py:70:18: RUF027 [*] Possible f-string without an `f` prefix
 69 |     last = "Appleseed"
 70 |     value.method("{first} {last}")  # RUF027
    |                  ^^^^^^^^^^^^^^^^ RUF027
+71 | 
+72 | def format_specifiers():
    |
    = help: Add `f` prefix
 
@@ -294,5 +296,24 @@ RUF027_0.py:70:18: RUF027 [*] Possible f-string without an `f` prefix
 69 69 |     last = "Appleseed"
 70    |-    value.method("{first} {last}")  # RUF027
    70 |+    value.method(f"{first} {last}")  # RUF027
+71 71 | 
+72 72 | def format_specifiers():
+73 73 |     a = 4
+
+RUF027_0.py:74:9: RUF027 [*] Possible f-string without an `f` prefix
+   |
+72 | def format_specifiers():
+73 |     a = 4
+74 |     b = "{a:b} {a:^5}"
+   |         ^^^^^^^^^^^^^^ RUF027
+   |
+   = help: Add `f` prefix
+
+ℹ Unsafe fix
+71 71 | 
+72 72 | def format_specifiers():
+73 73 |     a = 4
+74    |-    b = "{a:b} {a:^5}"
+   74 |+    b = f"{a:b} {a:^5}"
 
 

crates/ruff_linter/src/rules/ruff/snapshots/ruff_linter__rules__ruff__tests__RUF027_RUF027_2.py.snap

@@ -0,0 +1,4 @@
+---
+source: crates/ruff_linter/src/rules/ruff/mod.rs
+---
+

crates/ruff_macros/src/lib.rs

@@ -92,7 +92,9 @@ pub fn derive_message_formats(_attr: TokenStream, item: TokenStream) -> TokenStr
 ///
 /// Good:
 ///
-/// ```rust
+/// ```ignroe
+/// use ruff_macros::newtype_index;
+///
 /// #[newtype_index]
 /// #[derive(Ord, PartialOrd)]
 /// struct MyIndex;

crates/ruff_notebook/Cargo.toml

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

crates/ruff_python_ast/src/helpers.rs

@@ -935,7 +935,7 @@ where
     }
 }
 
-/// A [`StatementVisitor`] that collects all `return` statements in a function or method.
+/// A [`Visitor`] that collects all `return` statements in a function or method.
 #[derive(Default)]
 pub struct ReturnStatementVisitor<'a> {
     pub returns: Vec<&'a ast::StmtReturn>,

crates/ruff_python_ast/src/nodes.rs

@@ -1011,7 +1011,7 @@ pub struct DebugText {
 #[derive(Clone, Debug, PartialEq)]
 pub struct ExprFString {
     pub range: TextRange,
-    pub value: FStringValue,
+    inner: ExprFStringInner,
 }
 
 impl From<ExprFString> for Expr {
@@ -1020,17 +1020,12 @@ impl From<ExprFString> for Expr {
     }
 }
 
-/// The value representing an [`ExprFString`].
-#[derive(Clone, Debug, PartialEq)]
-pub struct FStringValue {
-    inner: FStringValueInner,
-}
-
-impl FStringValue {
+impl ExprFString {
     /// Creates a new f-string with the given value.
     pub fn single(value: FString) -> Self {
         Self {
-            inner: FStringValueInner::Single(FStringPart::FString(value)),
+            range: value.range,
+            inner: ExprFStringInner::Single(FStringPart::FString(value)),
         }
     }
 
@@ -1040,31 +1035,32 @@ impl FStringValue {
     /// # Panics
     ///
     /// Panics if `values` is less than 2. Use [`FStringValue::single`] instead.
-    pub fn concatenated(values: Vec<FStringPart>) -> Self {
+    pub fn concatenated(values: Vec<FStringPart>, range: TextRange) -> Self {
         assert!(values.len() > 1);
         Self {
-            inner: FStringValueInner::Concatenated(values),
+            range,
+            inner: ExprFStringInner::Concatenated(values),
         }
     }
 
     /// Returns `true` if the f-string is implicitly concatenated, `false` otherwise.
     pub fn is_implicit_concatenated(&self) -> bool {
-        matches!(self.inner, FStringValueInner::Concatenated(_))
+        matches!(self.inner, ExprFStringInner::Concatenated(_))
     }
 
     /// Returns a slice of all the [`FStringPart`]s contained in this value.
     pub fn as_slice(&self) -> &[FStringPart] {
         match &self.inner {
-            FStringValueInner::Single(part) => std::slice::from_ref(part),
-            FStringValueInner::Concatenated(parts) => parts,
+            ExprFStringInner::Single(part) => std::slice::from_ref(part),
+            ExprFStringInner::Concatenated(parts) => parts,
         }
     }
 
     /// Returns a mutable slice of all the [`FStringPart`]s contained in this value.
     fn as_mut_slice(&mut self) -> &mut [FStringPart] {
         match &mut self.inner {
-            FStringValueInner::Single(part) => std::slice::from_mut(part),
-            FStringValueInner::Concatenated(parts) => parts,
+            ExprFStringInner::Single(part) => std::slice::from_mut(part),
+            ExprFStringInner::Concatenated(parts) => parts,
         }
     }
 
@@ -1121,7 +1117,7 @@ impl FStringValue {
     }
 }
 
-impl<'a> IntoIterator for &'a FStringValue {
+impl<'a> IntoIterator for &'a ExprFString {
     type Item = &'a FStringPart;
     type IntoIter = Iter<'a, FStringPart>;
 
@@ -1130,9 +1126,10 @@ impl<'a> IntoIterator for &'a FStringValue {
     }
 }
 
-impl<'a> IntoIterator for &'a mut FStringValue {
+impl<'a> IntoIterator for &'a mut ExprFString {
     type Item = &'a mut FStringPart;
     type IntoIter = IterMut<'a, FStringPart>;
+
     fn into_iter(self) -> Self::IntoIter {
         self.iter_mut()
     }
@@ -1140,7 +1137,7 @@ impl<'a> IntoIterator for &'a mut FStringValue {
 
 /// An internal representation of [`FStringValue`].
 #[derive(Clone, Debug, PartialEq)]
-enum FStringValueInner {
+enum ExprFStringInner {
     /// A single f-string i.e., `f"foo"`.
     ///
     /// This is always going to be `FStringPart::FString` variant which is
@@ -1182,11 +1179,7 @@ impl Ranged for FString {
 
 impl From<FString> for Expr {
     fn from(payload: FString) -> Self {
-        ExprFString {
-            range: payload.range,
-            value: FStringValue::single(payload),
-        }
-        .into()
+        Expr::FString(ExprFString::single(payload))
     }
 }
 
@@ -1210,7 +1203,7 @@ impl Ranged for FStringElement {
 #[derive(Clone, Debug, Default, PartialEq)]
 pub struct ExprStringLiteral {
     pub range: TextRange,
-    pub value: StringLiteralValue,
+    inner: ExprStringLiteralInner,
 }
 
 impl From<ExprStringLiteral> for Expr {
@@ -1225,17 +1218,12 @@ impl Ranged for ExprStringLiteral {
     }
 }
 
-/// The value representing a [`ExprStringLiteral`].
-#[derive(Clone, Debug, Default, PartialEq)]
-pub struct StringLiteralValue {
-    inner: StringLiteralValueInner,
-}
-
-impl StringLiteralValue {
+impl ExprStringLiteral {
     /// Creates a new single string literal with the given value.
     pub fn single(string: StringLiteral) -> Self {
         Self {
-            inner: StringLiteralValueInner::Single(string),
+            range: string.range,
+            inner: ExprStringLiteralInner::Single(string),
         }
     }
 
@@ -1246,10 +1234,11 @@ impl StringLiteralValue {
     ///
     /// Panics if `strings` is less than 2. Use [`StringLiteralValue::single`]
     /// instead.
-    pub fn concatenated(strings: Vec<StringLiteral>) -> Self {
+    pub fn concatenated(strings: Vec<StringLiteral>, range: TextRange) -> Self {
         assert!(strings.len() > 1);
         Self {
-            inner: StringLiteralValueInner::Concatenated(ConcatenatedStringLiteral {
+            range,
+            inner: ExprStringLiteralInner::Concatenated(ConcatenatedStringLiteral {
                 strings,
                 value: OnceCell::new(),
             }),
@@ -1258,7 +1247,7 @@ impl StringLiteralValue {
 
     /// Returns `true` if the string literal is implicitly concatenated.
     pub const fn is_implicit_concatenated(&self) -> bool {
-        matches!(self.inner, StringLiteralValueInner::Concatenated(_))
+        matches!(self.inner, ExprStringLiteralInner::Concatenated(_))
     }
 
     /// Returns `true` if the string literal is a unicode string.
@@ -1272,16 +1261,16 @@ impl StringLiteralValue {
     /// Returns a slice of all the [`StringLiteral`] parts contained in this value.
     pub fn as_slice(&self) -> &[StringLiteral] {
         match &self.inner {
-            StringLiteralValueInner::Single(value) => std::slice::from_ref(value),
-            StringLiteralValueInner::Concatenated(value) => value.strings.as_slice(),
+            ExprStringLiteralInner::Single(value) => std::slice::from_ref(value),
+            ExprStringLiteralInner::Concatenated(value) => value.strings.as_slice(),
         }
     }
 
     /// Returns a mutable slice of all the [`StringLiteral`] parts contained in this value.
     fn as_mut_slice(&mut self) -> &mut [StringLiteral] {
         match &mut self.inner {
-            StringLiteralValueInner::Single(value) => std::slice::from_mut(value),
-            StringLiteralValueInner::Concatenated(value) => value.strings.as_mut_slice(),
+            ExprStringLiteralInner::Single(value) => std::slice::from_mut(value),
+            ExprStringLiteralInner::Concatenated(value) => value.strings.as_mut_slice(),
         }
     }
 
@@ -1318,13 +1307,13 @@ impl StringLiteralValue {
     /// string value is implicitly concatenated.
     pub fn to_str(&self) -> &str {
         match &self.inner {
-            StringLiteralValueInner::Single(value) => value.as_str(),
-            StringLiteralValueInner::Concatenated(value) => value.to_str(),
+            ExprStringLiteralInner::Single(value) => value.as_str(),
+            ExprStringLiteralInner::Concatenated(value) => value.to_str(),
         }
     }
 }
 
-impl<'a> IntoIterator for &'a StringLiteralValue {
+impl<'a> IntoIterator for &'a ExprStringLiteral {
     type Item = &'a StringLiteral;
     type IntoIter = Iter<'a, StringLiteral>;
 
@@ -1333,15 +1322,16 @@ impl<'a> IntoIterator for &'a StringLiteralValue {
     }
 }
 
-impl<'a> IntoIterator for &'a mut StringLiteralValue {
+impl<'a> IntoIterator for &'a mut ExprStringLiteral {
     type Item = &'a mut StringLiteral;
     type IntoIter = IterMut<'a, StringLiteral>;
+
     fn into_iter(self) -> Self::IntoIter {
         self.iter_mut()
     }
 }
 
-impl PartialEq<str> for StringLiteralValue {
+impl PartialEq<str> for ExprStringLiteral {
     fn eq(&self, other: &str) -> bool {
         if self.len() != other.len() {
             return false;
@@ -1351,13 +1341,13 @@ impl PartialEq<str> for StringLiteralValue {
     }
 }
 
-impl PartialEq<String> for StringLiteralValue {
+impl PartialEq<String> for ExprStringLiteral {
     fn eq(&self, other: &String) -> bool {
         self == other.as_str()
     }
 }
 
-impl fmt::Display for StringLiteralValue {
+impl fmt::Display for ExprStringLiteral {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         f.write_str(self.to_str())
     }
@@ -1365,7 +1355,7 @@ impl fmt::Display for StringLiteralValue {
 
 /// An internal representation of [`StringLiteralValue`].
 #[derive(Clone, Debug, PartialEq)]
-enum StringLiteralValueInner {
+enum ExprStringLiteralInner {
     /// A single string literal i.e., `"foo"`.
     Single(StringLiteral),
 
@@ -1373,7 +1363,7 @@ enum StringLiteralValueInner {
     Concatenated(ConcatenatedStringLiteral),
 }
 
-impl Default for StringLiteralValueInner {
+impl Default for ExprStringLiteralInner {
     fn default() -> Self {
         Self::Single(StringLiteral::default())
     }
@@ -1411,11 +1401,7 @@ impl StringLiteral {
 
 impl From<StringLiteral> for Expr {
     fn from(payload: StringLiteral) -> Self {
-        ExprStringLiteral {
-            range: payload.range,
-            value: StringLiteralValue::single(payload),
-        }
-        .into()
+        Expr::StringLiteral(ExprStringLiteral::single(payload))
     }
 }
 
@@ -1466,7 +1452,7 @@ impl Debug for ConcatenatedStringLiteral {
 #[derive(Clone, Debug, Default, PartialEq)]
 pub struct ExprBytesLiteral {
     pub range: TextRange,
-    pub value: BytesLiteralValue,
+    inner: ExprBytesLiteralInner,
 }
 
 impl From<ExprBytesLiteral> for Expr {
@@ -1481,17 +1467,12 @@ impl Ranged for ExprBytesLiteral {
     }
 }
 
-/// The value representing a [`ExprBytesLiteral`].
-#[derive(Clone, Debug, Default, PartialEq)]
-pub struct BytesLiteralValue {
-    inner: BytesLiteralValueInner,
-}
-
-impl BytesLiteralValue {
+impl ExprBytesLiteral {
     /// Creates a new single bytes literal with the given value.
     pub fn single(value: BytesLiteral) -> Self {
         Self {
-            inner: BytesLiteralValueInner::Single(value),
+            range: value.range,
+            inner: ExprBytesLiteralInner::Single(value),
         }
     }
 
@@ -1502,31 +1483,32 @@ impl BytesLiteralValue {
     ///
     /// Panics if `values` is less than 2. Use [`BytesLiteralValue::single`]
     /// instead.
-    pub fn concatenated(values: Vec<BytesLiteral>) -> Self {
+    pub fn concatenated(values: Vec<BytesLiteral>, range: TextRange) -> Self {
         assert!(values.len() > 1);
         Self {
-            inner: BytesLiteralValueInner::Concatenated(values),
+            range,
+            inner: ExprBytesLiteralInner::Concatenated(values),
         }
     }
 
     /// Returns `true` if the bytes literal is implicitly concatenated.
     pub const fn is_implicit_concatenated(&self) -> bool {
-        matches!(self.inner, BytesLiteralValueInner::Concatenated(_))
+        matches!(self.inner, ExprBytesLiteralInner::Concatenated(_))
     }
 
     /// Returns a slice of all the [`BytesLiteral`] parts contained in this value.
     pub fn as_slice(&self) -> &[BytesLiteral] {
         match &self.inner {
-            BytesLiteralValueInner::Single(value) => std::slice::from_ref(value),
-            BytesLiteralValueInner::Concatenated(value) => value.as_slice(),
+            ExprBytesLiteralInner::Single(value) => std::slice::from_ref(value),
+            ExprBytesLiteralInner::Concatenated(value) => value.as_slice(),
         }
     }
 
     /// Returns a mutable slice of all the [`BytesLiteral`] parts contained in this value.
     fn as_mut_slice(&mut self) -> &mut [BytesLiteral] {
         match &mut self.inner {
-            BytesLiteralValueInner::Single(value) => std::slice::from_mut(value),
-            BytesLiteralValueInner::Concatenated(value) => value.as_mut_slice(),
+            ExprBytesLiteralInner::Single(value) => std::slice::from_mut(value),
+            ExprBytesLiteralInner::Concatenated(value) => value.as_mut_slice(),
         }
     }
 
@@ -1557,7 +1539,7 @@ impl BytesLiteralValue {
     }
 }
 
-impl<'a> IntoIterator for &'a BytesLiteralValue {
+impl<'a> IntoIterator for &'a ExprBytesLiteral {
     type Item = &'a BytesLiteral;
     type IntoIter = Iter<'a, BytesLiteral>;
 
@@ -1566,15 +1548,16 @@ impl<'a> IntoIterator for &'a BytesLiteralValue {
     }
 }
 
-impl<'a> IntoIterator for &'a mut BytesLiteralValue {
+impl<'a> IntoIterator for &'a mut ExprBytesLiteral {
     type Item = &'a mut BytesLiteral;
     type IntoIter = IterMut<'a, BytesLiteral>;
+
     fn into_iter(self) -> Self::IntoIter {
         self.iter_mut()
     }
 }
 
-impl PartialEq<[u8]> for BytesLiteralValue {
+impl PartialEq<[u8]> for ExprBytesLiteral {
     fn eq(&self, other: &[u8]) -> bool {
         if self.len() != other.len() {
             return false;
@@ -1588,7 +1571,7 @@ impl PartialEq<[u8]> for BytesLiteralValue {
 
 /// An internal representation of [`BytesLiteralValue`].
 #[derive(Clone, Debug, PartialEq)]
-enum BytesLiteralValueInner {
+enum ExprBytesLiteralInner {
     /// A single bytes literal i.e., `b"foo"`.
     Single(BytesLiteral),
 
@@ -1596,7 +1579,7 @@ enum BytesLiteralValueInner {
     Concatenated(Vec<BytesLiteral>),
 }
 
-impl Default for BytesLiteralValueInner {
+impl Default for ExprBytesLiteralInner {
     fn default() -> Self {
         Self::Single(BytesLiteral::default())
     }
@@ -1633,11 +1616,7 @@ impl BytesLiteral {
 
 impl From<BytesLiteral> for Expr {
     fn from(payload: BytesLiteral) -> Self {
-        ExprBytesLiteral {
-            range: payload.range,
-            value: BytesLiteralValue::single(payload),
-        }
-        .into()
+        Expr::BytesLiteral(ExprBytesLiteral::single(payload))
     }
 }
 

crates/ruff_python_codegen/Cargo.toml

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

crates/ruff_python_codegen/src/generator.rs

@@ -1705,6 +1705,7 @@ class Foo:
         assert_round_trip!(r#"f"{ chr(65)  =   !s}""#);
         assert_round_trip!(r#"f"{ chr(65)  =   !r}""#);
         assert_round_trip!(r#"f"{ chr(65)  =   :#x}""#);
+        assert_round_trip!(r#"f"{  ( chr(65)  ) = }""#);
         assert_round_trip!(r#"f"{a=!r:0.05f}""#);
     }
 

crates/ruff_python_formatter/Cargo.toml

@@ -10,6 +10,9 @@ documentation = { workspace = true }
 repository = { workspace = true }
 license = { workspace = true }
 
+[lib]
+doctest= false
+
 [dependencies]
 ruff_cache = { path = "../ruff_cache" }
 ruff_formatter = { path = "../ruff_formatter" }

crates/ruff_python_formatter/resources/test/fixtures/ruff/.editorconfig

@@ -4,4 +4,8 @@ ij_formatter_enabled = false
 
 ["range_formatting/*.py"]
 generated_code = true
+ij_formatter_enabled = false
+
+[docstring_tab_indentation.py]
+generated_code = true
 ij_formatter_enabled = false

crates/ruff_python_formatter/resources/test/fixtures/ruff/docstring_tab_indentation.options.json

@@ -0,0 +1,10 @@
+[
+  {
+    "indent_style": "tab",
+    "indent_width": 4
+  },
+  {
+    "indent_style": "tab",
+    "indent_width": 8
+  }
+]

crates/ruff_python_formatter/resources/test/fixtures/ruff/docstring_tab_indentation.py

@@ -0,0 +1,72 @@
+# Tests the behavior of the formatter when it comes to tabs inside docstrings
+# when using `indent_style="tab`
+
+# The example below uses tabs exclusively. The formatter should preserve the tab indentation
+# of `arg1`.
+def tab_argument(arg1: str) -> None:
+	"""
+	Arguments:
+		arg1: super duper arg with 2 tabs in front
+	"""
+
+# The `arg1` is intended with spaces. The formatter should not change the spaces to a tab
+# because it must assume that the spaces are used for alignment and not indentation.
+def space_argument(arg1: str) -> None:
+	"""
+	Arguments:
+	        arg1: super duper arg with a tab and a space in front
+	"""
+
+def under_indented(arg1: str) -> None:
+	"""
+	Arguments:
+	        arg1: super duper arg with a tab and a space in front
+arg2: Not properly indented
+	"""
+
+def under_indented_tabs(arg1: str) -> None:
+	"""
+	Arguments:
+		arg1: super duper arg with a tab and a space in front
+arg2: Not properly indented
+	"""
+
+def spaces_tabs_over_indent(arg1: str) -> None:
+    """
+    Arguments:
+      	arg1: super duper arg with a tab and a space in front
+    """
+
+# The docstring itself is indented with spaces but the argument is indented by a tab.
+# Keep the tab indentation of the argument, convert th docstring indent to tabs.
+def space_indented_docstring_containing_tabs(arg1: str) -> None:
+    """
+    Arguments:
+    	arg1: super duper arg
+    """
+
+
+# The docstring uses tabs, spaces, tabs indentation.
+# Fallback to use space indentation
+def mixed_indentation(arg1: str) -> None:
+	"""
+	Arguments:
+	        	arg1: super duper arg with a tab and a space in front
+	"""
+
+
+# The example shows an ascii art. The formatter should not change the spaces
+# to tabs because it breaks the ASCII art when inspecting the docstring with `inspect.cleandoc(ascii_art.__doc__)`
+# when using an indent width other than 8.
+def ascii_art():
+	r"""
+	Look at this beautiful tree.
+
+	    a
+	   / \
+	  b   c
+	 / \
+	d   e
+	"""
+
+

crates/ruff_python_formatter/resources/test/fixtures/ruff/expression/fstring.options.json

@@ -0,0 +1,8 @@
+[
+  {
+    "preview": "enabled"
+  },
+  {
+    "preview": "disabled"
+  }
+]

crates/ruff_python_formatter/resources/test/fixtures/ruff/expression/fstring.py

@@ -30,22 +30,22 @@ result_f = (
 # an expression inside a formatted value
 (
     f'{1}'
-    # comment
+    # comment 1
     ''
 )
 
 (
-    f'{1}'  # comment
+    f'{1}'  # comment 2
     f'{2}'
 )
 
 (
     f'{1}'
-    f'{2}'  # comment
+    f'{2}'  # comment 3
 )
 
 (
-    1, (  # comment
+    1, (  # comment 4
         f'{2}'
     )
 )
@@ -53,7 +53,7 @@ result_f = (
 (
     (
         f'{1}'
-        # comment
+        # comment 5
     ),
     2
 )
@@ -62,3 +62,221 @@ result_f = (
 x = f'''a{""}b'''
 y = f'''c{1}d"""e'''
 z = f'''a{""}b''' f'''c{1}d"""e'''
+
+# F-String formatting test cases (Preview)
+
+# Simple expression with a mix of debug expression and comments.
+x = f"{a}"
+x = f"{
+    a = }"
+x = f"{ # comment 6
+    a }"
+x = f"{   # comment 7
+    a = }"
+
+# Remove the parentheses as adding them doesn't make then fit within the line length limit.
+# This is similar to how we format it before f-string formatting.
+aaaaaaaaaaa = (
+    f"asaaaaaaaaaaaaaaaa { aaaaaaaaaaaa + bbbbbbbbbbbb + ccccccccccccccc + dddddddd } cccccccccc"
+)
+# Here, we would use the best fit layout to put the f-string indented on the next line
+# similar to the next example.
+aaaaaaaaaaa = f"asaaaaaaaaaaaaaaaa { aaaaaaaaaaaa + bbbbbbbbbbbb + ccccccccccccccc } cccccccccc"
+aaaaaaaaaaa = (
+    f"asaaaaaaaaaaaaaaaa { aaaaaaaaaaaa + bbbbbbbbbbbb + ccccccccccccccc } cccccccccc"
+)
+
+# This should never add the optional parentheses because even after adding them, the
+# f-string exceeds the line length limit.
+x = f"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa { "bbbbbbbbbbbbbbbbbbbbbbbbbbbbb" } ccccccccccccccc"
+x = f"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa { "bbbbbbbbbbbbbbbbbbbbbbbbbbbbb" = } ccccccccccccccc"
+x = f"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa { # comment 8
+                                             "bbbbbbbbbbbbbbbbbbbbbbbbbbbbb" } ccccccccccccccc"
+x = f"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa { # comment 9
+                                             "bbbbbbbbbbbbbbbbbbbbbbbbbbbbb" = } ccccccccccccccc"
+
+# Multiple larger expressions which exceeds the line length limit. Here, we need to decide
+# whether to split at the first or second expression. This should work similarly to the
+# assignment statement formatting where we split from right to left in preview mode.
+x = f"aaaaaaaaaaaa { bbbbbbbbbbbbbb } cccccccccccccccccccc { ddddddddddddddd } eeeeeeeeeeeeee"
+
+# The above example won't split but when we start introducing line breaks:
+x = f"aaaaaaaaaaaa {
+        bbbbbbbbbbbbbb } cccccccccccccccccccc { ddddddddddddddd } eeeeeeeeeeeeee"
+x = f"aaaaaaaaaaaa { bbbbbbbbbbbbbb
+                    } cccccccccccccccccccc { ddddddddddddddd } eeeeeeeeeeeeee"
+x = f"aaaaaaaaaaaa { bbbbbbbbbbbbbb } cccccccccccccccccccc {
+        ddddddddddddddd } eeeeeeeeeeeeee"
+x = f"aaaaaaaaaaaa { bbbbbbbbbbbbbb } cccccccccccccccccccc { ddddddddddddddd
+                                                            } eeeeeeeeeeeeee"
+
+# But, in case comments are present, we would split at the expression containing the
+# comments:
+x = f"aaaaaaaaaaaa { bbbbbbbbbbbbbb # comment 10
+                    } cccccccccccccccccccc { ddddddddddddddd } eeeeeeeeeeeeee"
+x = f"aaaaaaaaaaaa { bbbbbbbbbbbbbb
+                    } cccccccccccccccccccc { # comment 11
+                                            ddddddddddddddd } eeeeeeeeeeeeee"
+
+# Here, the expression part itself starts with a curly brace so we need to add an extra
+# space between the opening curly brace and the expression.
+x = f"{ {'x': 1, 'y': 2} }"
+# Although the extra space isn't required before the ending curly brace, we add it for
+# consistency.
+x = f"{ {'x': 1, 'y': 2}}"
+x = f"{ {'x': 1, 'y': 2} = }"
+x = f"{  # comment 12
+    {'x': 1, 'y': 2} }"
+x = f"{    # comment 13
+    {'x': 1, 'y': 2} = }"
+
+# But, in this case, we would split the expression itself because it exceeds the line
+# length limit so we need not add the extra space.
+xxxxxxx = f"{
+    {'aaaaaaaaaaaaaaaaaaa', 'bbbbbbbbbbbbbbbbbbbbbb', 'ccccccccccccccccccccc'}
+}"
+# And, split the expression itself because it exceeds the line length.
+xxxxxxx = f"{
+    {'aaaaaaaaaaaaaaaaaaaaaaaaa', 'bbbbbbbbbbbbbbbbbbbbbbbbbbb', 'cccccccccccccccccccccccccc'}
+}"
+
+# Quotes
+f"foo 'bar' {x}"
+f"foo \"bar\" {x}"
+f'foo "bar" {x}'
+f'foo \'bar\' {x}'
+f"foo {"bar"}"
+f"foo {'\'bar\''}"
+
+# Here, the formatter will remove the escapes which is correct because they aren't allowed
+# pre 3.12. This means we can assume that the f-string is used in the context of 3.12.
+f"foo {'\"bar\"'}"
+
+
+# Triple-quoted strings
+# It's ok to use the same quote char for the inner string if it's single-quoted.
+f"""test {'inner'}"""
+f"""test {"inner"}"""
+# But if the inner string is also triple-quoted then we should preserve the existing quotes.
+f"""test {'''inner'''}"""
+
+# Magic trailing comma
+#
+# The expression formatting will result in breaking it across multiple lines with a
+# trailing comma but as the expression isn't already broken, we will remove all the line
+# breaks which results in the trailing comma being present. This test case makes sure
+# that the trailing comma is removed as well.
+f"aaaaaaa {['aaaaaaaaaaaaaaa', 'bbbbbbbbbbbbb', 'ccccccccccccccccc', 'ddddddddddddddd', 'eeeeeeeeeeeeee']} aaaaaaa"
+
+# And, if the trailing comma is already present, we still need to remove it.
+f"aaaaaaa {['aaaaaaaaaaaaaaa', 'bbbbbbbbbbbbb', 'ccccccccccccccccc', 'ddddddddddddddd', 'eeeeeeeeeeeeee',]} aaaaaaa"
+
+# Keep this Multiline by breaking it at the square brackets.
+f"""aaaaaa {[
+    xxxxxxxx,
+    yyyyyyyy,
+]} ccc"""
+
+# Add the magic trailing comma because the elements don't fit within the line length limit
+# when collapsed.
+f"aaaaaa {[
+    xxxxxxxxxxxx,
+    xxxxxxxxxxxx,
+    xxxxxxxxxxxx,
+    xxxxxxxxxxxx,
+    xxxxxxxxxxxx,
+    xxxxxxxxxxxx,
+    yyyyyyyyyyyy
+]} ccccccc"
+
+# Remove the parenthese because they aren't required
+xxxxxxxxxxxxxxx = (
+    f"aaaaaaaaaaaaaaaa bbbbbbbbbbbbbbb {
+        xxxxxxxxxxx  # comment 14
+        + yyyyyyyyyy
+    } dddddddddd"
+)
+
+# Comments
+
+# No comments should be dropped!
+f"{ # comment 15
+    # comment 16
+    foo # comment 17
+    # comment 18
+}"  # comment 19
+# comment 20
+
+# Conversion flags
+#
+# This is not a valid Python code because of the additional whitespace between the `!`
+# and conversion type. But, our parser isn't strict about this. This should probably be
+# removed once we have a strict parser.
+x = f"aaaaaaaaa { x !  r }"
+
+# Even in the case of debug expresions, we only need to preserve the whitespace within
+# the expression part of the replacement field.
+x = f"aaaaaaaaa { x   = !  r  }"
+
+# Combine conversion flags with format specifiers
+x = f"{x   =   !  s
+         :>0
+
+         }"
+# This is interesting. There can be a comment after the format specifier but only if it's
+# on it's own line. Refer to https://github.com/astral-sh/ruff/pull/7787 for more details.
+# We'll format is as trailing comments.
+x = f"{x  !s
+         :>0
+         # comment 21
+         }"
+
+x = f"""
+{              # comment 22
+ x =   :.0{y # comment 23
+           }f}"""
+
+# Here, the debug expression is in a nested f-string so we should start preserving
+# whitespaces from that point onwards. This means we should format the outer f-string.
+x = f"""{"foo " +    # comment 24
+    f"{   x =
+
+       }"    # comment 25
+ }
+        """
+
+# Mix of various features.
+f"{  # comment 26
+    foo # after foo
+   :>{
+          x # after x
+          }
+    # comment 27
+    # comment 28
+} woah {x}"
+
+# Indentation
+
+# What should be the indentation?
+# https://github.com/astral-sh/ruff/discussions/9785#discussioncomment-8470590
+if indent0:
+    if indent1:
+        if indent2:
+            foo = f"""hello world
+hello {
+          f"aaaaaaa {
+              [
+                  'aaaaaaaaaaaaaaaaaaaaa',
+                  'bbbbbbbbbbbbbbbbbbbbb',
+                  'ccccccccccccccccccccc',
+                  'ddddddddddddddddddddd'
+              ]
+          } bbbbbbbb" +
+          [
+              'aaaaaaaaaaaaaaaaaaaaa',
+              'bbbbbbbbbbbbbbbbbbbbb',
+              'ccccccccccccccccccccc',
+              'ddddddddddddddddddddd'
+          ]
+      } --------
+"""

crates/ruff_python_formatter/resources/test/fixtures/ruff/expression/fstring_py312.options.json

@@ -0,0 +1,5 @@
+[
+  {
+    "target_version": "py312"
+  }
+]

crates/ruff_python_formatter/resources/test/fixtures/ruff/expression/fstring_py312.py

@@ -0,0 +1,6 @@
+# This file contains test cases only for cases where the logic tests for whether
+# the target version is 3.12 or later. A user can have 3.12 syntax even if the target
+# version isn't set.
+
+# Quotes re-use
+f"{'a'}"

crates/ruff_python_formatter/resources/test/fixtures/ruff/notebook_docstring.options.json

@@ -0,0 +1,8 @@
+[
+  {
+    "source_type": "Ipynb"
+  },
+  {
+    "source_type": "Python"
+  }
+]

crates/ruff_python_formatter/resources/test/fixtures/ruff/notebook_docstring.py

@@ -0,0 +1,6 @@
+"""
+    This looks like a docstring but is not in a notebook because notebooks can't be imported as a module.
+    Ruff should leave it as is
+""";
+
+"another normal string"

crates/ruff_python_formatter/src/builders.rs

@@ -1,7 +1,7 @@
 use ruff_formatter::{write, Argument, Arguments};
 use ruff_text_size::{Ranged, TextRange, TextSize};
 
-use crate::context::{NodeLevel, WithNodeLevel};
+use crate::context::{FStringState, NodeLevel, WithNodeLevel};
 use crate::other::commas::has_magic_trailing_comma;
 use crate::prelude::*;
 
@@ -206,6 +206,16 @@ impl<'fmt, 'ast, 'buf> JoinCommaSeparatedBuilder<'fmt, 'ast, 'buf> {
 
     pub(crate) fn finish(&mut self) -> FormatResult<()> {
         self.result.and_then(|()| {
+            // If the formatter is inside an f-string expression element, and the layout
+            // is flat, then we don't need to add a trailing comma.
+            if let FStringState::InsideExpressionElement(context) =
+                self.fmt.context().f_string_state()
+            {
+                if context.layout().is_flat() {
+                    return Ok(());
+                }
+            }
+
             if let Some(last_end) = self.entries.position() {
                 let magic_trailing_comma = has_magic_trailing_comma(
                     TextRange::new(last_end, self.sequence_end),

crates/ruff_python_formatter/src/comments/placement.rs

@@ -289,6 +289,28 @@ fn handle_enclosed_comment<'a>(
             }
         }
         AnyNodeRef::FString(fstring) => CommentPlacement::dangling(fstring, comment),
+        AnyNodeRef::FStringExpressionElement(_) => {
+            // Handle comments after the format specifier (should be rare):
+            //
+            // ```python
+            // f"literal {
+            //     expr:.3f
+            //     # comment
+            // }"
+            // ```
+            //
+            // This is a valid comment placement.
+            if matches!(
+                comment.preceding_node(),
+                Some(
+                    AnyNodeRef::FStringExpressionElement(_) | AnyNodeRef::FStringLiteralElement(_)
+                )
+            ) {
+                CommentPlacement::trailing(comment.enclosing_node(), comment)
+            } else {
+                handle_bracketed_end_of_line_comment(comment, locator)
+            }
+        }
         AnyNodeRef::ExprList(_)
         | AnyNodeRef::ExprSet(_)
         | AnyNodeRef::ExprListComp(_)

crates/ruff_python_formatter/src/context.rs

@@ -1,4 +1,5 @@
 use crate::comments::Comments;
+use crate::other::f_string::FStringContext;
 use crate::string::QuoteChar;
 use crate::PyFormatOptions;
 use ruff_formatter::{Buffer, FormatContext, GroupId, IndentWidth, SourceCode};
@@ -22,6 +23,8 @@ pub struct PyFormatContext<'a> {
     /// quote style that is inverted from the one here in order to ensure that
     /// the formatted Python code will be valid.
     docstring: Option<QuoteChar>,
+    /// The state of the formatter with respect to f-strings.
+    f_string_state: FStringState,
 }
 
 impl<'a> PyFormatContext<'a> {
@@ -33,6 +36,7 @@ impl<'a> PyFormatContext<'a> {
             node_level: NodeLevel::TopLevel(TopLevelStatementPosition::Other),
             indent_level: IndentLevel::new(0),
             docstring: None,
+            f_string_state: FStringState::Outside,
         }
     }
 
@@ -86,6 +90,14 @@ impl<'a> PyFormatContext<'a> {
         }
     }
 
+    pub(crate) fn f_string_state(&self) -> FStringState {
+        self.f_string_state
+    }
+
+    pub(crate) fn set_f_string_state(&mut self, f_string_state: FStringState) {
+        self.f_string_state = f_string_state;
+    }
+
     /// Returns `true` if preview mode is enabled.
     pub(crate) const fn is_preview(&self) -> bool {
         self.options.preview().is_enabled()
@@ -115,6 +127,18 @@ impl Debug for PyFormatContext<'_> {
     }
 }
 
+#[derive(Copy, Clone, Debug, Default)]
+pub(crate) enum FStringState {
+    /// The formatter is inside an f-string expression element i.e., between the
+    /// curly brace in `f"foo {x}"`.
+    ///
+    /// The containing `FStringContext` is the surrounding f-string context.
+    InsideExpressionElement(FStringContext),
+    /// The formatter is outside an f-string.
+    #[default]
+    Outside,
+}
+
 /// The position of a top-level statement in the module.
 #[derive(Copy, Clone, Debug, Eq, PartialEq, Default)]
 pub(crate) enum TopLevelStatementPosition {
@@ -332,3 +356,65 @@ where
             .set_indent_level(self.saved_level);
     }
 }
+
+pub(crate) struct WithFStringState<'a, B, D>
+where
+    D: DerefMut<Target = B>,
+    B: Buffer<Context = PyFormatContext<'a>>,
+{
+    buffer: D,
+    saved_location: FStringState,
+}
+
+impl<'a, B, D> WithFStringState<'a, B, D>
+where
+    D: DerefMut<Target = B>,
+    B: Buffer<Context = PyFormatContext<'a>>,
+{
+    pub(crate) fn new(expr_location: FStringState, mut buffer: D) -> Self {
+        let context = buffer.state_mut().context_mut();
+        let saved_location = context.f_string_state();
+
+        context.set_f_string_state(expr_location);
+
+        Self {
+            buffer,
+            saved_location,
+        }
+    }
+}
+
+impl<'a, B, D> Deref for WithFStringState<'a, B, D>
+where
+    D: DerefMut<Target = B>,
+    B: Buffer<Context = PyFormatContext<'a>>,
+{
+    type Target = B;
+
+    fn deref(&self) -> &Self::Target {
+        &self.buffer
+    }
+}
+
+impl<'a, B, D> DerefMut for WithFStringState<'a, B, D>
+where
+    D: DerefMut<Target = B>,
+    B: Buffer<Context = PyFormatContext<'a>>,
+{
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.buffer
+    }
+}
+
+impl<'a, B, D> Drop for WithFStringState<'a, B, D>
+where
+    D: DerefMut<Target = B>,
+    B: Buffer<Context = PyFormatContext<'a>>,
+{
+    fn drop(&mut self) {
+        self.buffer
+            .state_mut()
+            .context_mut()
+            .set_f_string_state(self.saved_location);
+    }
+}

crates/ruff_python_formatter/src/expression/expr_f_string.rs

@@ -48,6 +48,24 @@ impl NeedsParentheses for ExprFString {
     ) -> OptionalParentheses {
         if self.value.is_implicit_concatenated() {
             OptionalParentheses::Multiline
+        // TODO(dhruvmanila): Ideally what we want here is a new variant which
+        // is something like:
+        // - If the expression fits by just adding the parentheses, then add them and
+        //   avoid breaking the f-string expression. So,
+        //   ```
+        //   xxxxxxxxx = (
+        //       f"aaaaaaaaaaaa { xxxxxxx + yyyyyyyy } bbbbbbbbbbbbb"
+        //   )
+        //   ```
+        // - But, if the expression is too long to fit even with parentheses, then
+        //   don't add the parentheses and instead break the expression at `soft_line_break`.
+        //   ```
+        //   xxxxxxxxx = f"aaaaaaaaaaaa {
+        //       xxxxxxxxx + yyyyyyyyyy
+        //   } bbbbbbbbbbbbb"
+        //   ```
+        // This isn't decided yet, refer to the relevant discussion:
+        // https://github.com/astral-sh/ruff/discussions/9785
         } else if AnyString::FString(self).is_multiline(context.source()) {
             OptionalParentheses::Never
         } else {

crates/ruff_python_formatter/src/options.rs

@@ -248,6 +248,12 @@ pub enum QuoteStyle {
     Preserve,
 }
 
+impl QuoteStyle {
+    pub const fn is_preserve(self) -> bool {
+        matches!(self, QuoteStyle::Preserve)
+    }
+}
+
 impl fmt::Display for QuoteStyle {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         match self {
@@ -460,3 +466,12 @@ pub enum PythonVersion {
     Py311,
     Py312,
 }
+
+impl PythonVersion {
+    /// Return `true` if the current version supports [PEP 701].
+    ///
+    /// [PEP 701]: https://peps.python.org/pep-0701/
+    pub fn supports_pep_701(self) -> bool {
+        self >= Self::Py312
+    }
+}

crates/ruff_python_formatter/src/other/bytes_literal.rs

@@ -2,8 +2,7 @@ use ruff_python_ast::BytesLiteral;
 use ruff_text_size::Ranged;
 
 use crate::prelude::*;
-use crate::preview::is_hex_codes_in_unicode_sequences_enabled;
-use crate::string::{Quoting, StringPart};
+use crate::string::{StringNormalizer, StringPart};
 
 #[derive(Default)]
 pub struct FormatBytesLiteral;
@@ -12,14 +11,9 @@ impl FormatNodeRule<BytesLiteral> for FormatBytesLiteral {
     fn fmt_fields(&self, item: &BytesLiteral, f: &mut PyFormatter) -> FormatResult<()> {
         let locator = f.context().locator();
 
-        StringPart::from_source(item.range(), &locator)
-            .normalize(
-                Quoting::CanChange,
-                &locator,
-                f.options().quote_style(),
-                f.context().docstring(),
-                is_hex_codes_in_unicode_sequences_enabled(f.context()),
-            )
+        StringNormalizer::from_context(f.context())
+            .with_preferred_quote_style(f.options().quote_style())
+            .normalize(&StringPart::from_source(item.range(), &locator), &locator)
             .fmt(f)
     }
 }

crates/ruff_python_formatter/src/other/f_string.rs

@@ -1,9 +1,13 @@
+use ruff_formatter::write;
 use ruff_python_ast::FString;
+use ruff_source_file::Locator;
 use ruff_text_size::Ranged;
 
 use crate::prelude::*;
-use crate::preview::is_hex_codes_in_unicode_sequences_enabled;
-use crate::string::{Quoting, StringPart};
+use crate::preview::is_f_string_formatting_enabled;
+use crate::string::{Quoting, StringNormalizer, StringPart, StringPrefix, StringQuotes};
+
+use super::f_string_element::FormatFStringElement;
 
 /// Formats an f-string which is part of a larger f-string expression.
 ///
@@ -26,26 +30,126 @@ impl Format<PyFormatContext<'_>> for FormatFString<'_> {
     fn fmt(&self, f: &mut PyFormatter) -> FormatResult<()> {
         let locator = f.context().locator();
 
-        let result = StringPart::from_source(self.value.range(), &locator)
-            .normalize(
-                self.quoting,
-                &locator,
-                f.options().quote_style(),
-                f.context().docstring(),
-                is_hex_codes_in_unicode_sequences_enabled(f.context()),
+        let string = StringPart::from_source(self.value.range(), &locator);
+
+        let normalizer = StringNormalizer::from_context(f.context())
+            .with_quoting(self.quoting)
+            .with_preferred_quote_style(f.options().quote_style());
+
+        // If f-string formatting is disabled (not in preview), then we will
+        // fall back to the previous behavior of normalizing the f-string.
+        if !is_f_string_formatting_enabled(f.context()) {
+            let result = normalizer.normalize(&string, &locator).fmt(f);
+            let comments = f.context().comments();
+            self.value.elements.iter().for_each(|value| {
+                comments.mark_verbatim_node_comments_formatted(value.into());
+                // Above method doesn't mark the trailing comments of the f-string elements
+                // as formatted, so we need to do it manually. For example,
+                //
+                // ```python
+                // f"""foo {
+                //     x:.3f
+                //     # comment
+                // }"""
+                // ```
+                for trailing_comment in comments.trailing(value) {
+                    trailing_comment.mark_formatted();
+                }
+            });
+            return result;
+        }
+
+        let quotes = normalizer.choose_quotes(&string, &locator);
+
+        let context = FStringContext::new(
+            string.prefix(),
+            quotes,
+            FStringLayout::from_f_string(self.value, &locator),
+        );
+
+        // Starting prefix and quote
+        write!(f, [string.prefix(), quotes])?;
+
+        f.join()
+            .entries(
+                self.value
+                    .elements
+                    .iter()
+                    .map(|element| FormatFStringElement::new(element, context)),
             )
-            .fmt(f);
+            .finish()?;
 
-        // TODO(dhruvmanila): With PEP 701, comments can be inside f-strings.
-        // This is to mark all of those comments as formatted but we need to
-        // figure out how to handle them. Note that this needs to be done only
-        // after the f-string is formatted, so only for all the non-formatted
-        // comments.
-        let comments = f.context().comments();
-        self.value.elements.iter().for_each(|value| {
-            comments.mark_verbatim_node_comments_formatted(value.into());
-        });
-
-        result
+        // Ending quote
+        quotes.fmt(f)
+    }
+}
+
+#[derive(Clone, Copy, Debug)]
+pub(crate) struct FStringContext {
+    prefix: StringPrefix,
+    quotes: StringQuotes,
+    layout: FStringLayout,
+}
+
+impl FStringContext {
+    const fn new(prefix: StringPrefix, quotes: StringQuotes, layout: FStringLayout) -> Self {
+        Self {
+            prefix,
+            quotes,
+            layout,
+        }
+    }
+
+    pub(crate) const fn quotes(self) -> StringQuotes {
+        self.quotes
+    }
+
+    pub(crate) const fn prefix(self) -> StringPrefix {
+        self.prefix
+    }
+
+    pub(crate) const fn layout(self) -> FStringLayout {
+        self.layout
+    }
+}
+
+#[derive(Copy, Clone, Debug)]
+pub(crate) enum FStringLayout {
+    /// Original f-string is flat.
+    /// Don't break expressions to keep the string flat.
+    Flat,
+    /// Original f-string has multiline expressions in the replacement fields.
+    /// Allow breaking expressions across multiple lines.
+    Multiline,
+}
+
+impl FStringLayout {
+    fn from_f_string(f_string: &FString, locator: &Locator) -> Self {
+        // Heuristic: Allow breaking the f-string expressions across multiple lines
+        // only if there already is at least one multiline expression. This puts the
+        // control in the hands of the user to decide if they want to break the
+        // f-string expressions across multiple lines or not. This is similar to
+        // how Prettier does it for template literals in JavaScript.
+        //
+        // If it's single quoted f-string and it contains a multiline expression, then we
+        // assume that the target version of Python supports it (3.12+). If there are comments
+        // used in any of the expression of the f-string, then it's always going to be multiline
+        // and we assume that the target version of Python supports it (3.12+).
+        //
+        // Reference: https://prettier.io/docs/en/next/rationale.html#template-literals
+        if f_string
+            .elements
+            .iter()
+            .filter_map(|element| element.as_expression())
+            .any(|expr| memchr::memchr2(b'\n', b'\r', locator.slice(expr).as_bytes()).is_some())
+        {
+            Self::Multiline
+        } else {
+            Self::Flat
+        }
+    }
+
+    pub(crate) const fn is_flat(self) -> bool {
+        matches!(self, Self::Flat)
     }
 }

crates/ruff_python_formatter/src/other/f_string_element.rs

@@ -0,0 +1,244 @@
+use std::borrow::Cow;
+
+use ruff_formatter::{format_args, write, Buffer, RemoveSoftLinesBuffer};
+use ruff_python_ast::{
+    ConversionFlag, Expr, FStringElement, FStringExpressionElement, FStringLiteralElement,
+};
+use ruff_text_size::Ranged;
+
+use crate::comments::{dangling_open_parenthesis_comments, trailing_comments};
+use crate::context::{FStringState, NodeLevel, WithFStringState, WithNodeLevel};
+use crate::prelude::*;
+use crate::preview::is_hex_codes_in_unicode_sequences_enabled;
+use crate::string::normalize_string;
+use crate::verbatim::verbatim_text;
+
+use super::f_string::FStringContext;
+
+/// Formats an f-string element which is either a literal or a formatted expression.
+///
+/// This delegates the actual formatting to the appropriate formatter.
+pub(crate) struct FormatFStringElement<'a> {
+    element: &'a FStringElement,
+    context: FStringContext,
+}
+
+impl<'a> FormatFStringElement<'a> {
+    pub(crate) fn new(element: &'a FStringElement, context: FStringContext) -> Self {
+        Self { element, context }
+    }
+}
+
+impl Format<PyFormatContext<'_>> for FormatFStringElement<'_> {
+    fn fmt(&self, f: &mut PyFormatter) -> FormatResult<()> {
+        match self.element {
+            FStringElement::Literal(string_literal) => {
+                FormatFStringLiteralElement::new(string_literal, self.context).fmt(f)
+            }
+            FStringElement::Expression(expression) => {
+                FormatFStringExpressionElement::new(expression, self.context).fmt(f)
+            }
+        }
+    }
+}
+
+/// Formats an f-string literal element.
+pub(crate) struct FormatFStringLiteralElement<'a> {
+    element: &'a FStringLiteralElement,
+    context: FStringContext,
+}
+
+impl<'a> FormatFStringLiteralElement<'a> {
+    pub(crate) fn new(element: &'a FStringLiteralElement, context: FStringContext) -> Self {
+        Self { element, context }
+    }
+}
+
+impl Format<PyFormatContext<'_>> for FormatFStringLiteralElement<'_> {
+    fn fmt(&self, f: &mut PyFormatter) -> FormatResult<()> {
+        let literal_content = f.context().locator().slice(self.element.range());
+        let normalized = normalize_string(
+            literal_content,
+            self.context.quotes(),
+            self.context.prefix(),
+            is_hex_codes_in_unicode_sequences_enabled(f.context()),
+        );
+        match &normalized {
+            Cow::Borrowed(_) => source_text_slice(self.element.range()).fmt(f),
+            Cow::Owned(normalized) => text(normalized).fmt(f),
+        }
+    }
+}
+
+/// Formats an f-string expression element.
+pub(crate) struct FormatFStringExpressionElement<'a> {
+    element: &'a FStringExpressionElement,
+    context: FStringContext,
+}
+
+impl<'a> FormatFStringExpressionElement<'a> {
+    pub(crate) fn new(element: &'a FStringExpressionElement, context: FStringContext) -> Self {
+        Self { element, context }
+    }
+}
+
+impl Format<PyFormatContext<'_>> for FormatFStringExpressionElement<'_> {
+    fn fmt(&self, f: &mut PyFormatter) -> FormatResult<()> {
+        let FStringExpressionElement {
+            expression,
+            debug_text,
+            conversion,
+            format_spec,
+            ..
+        } = self.element;
+
+        if let Some(debug_text) = debug_text {
+            token("{").fmt(f)?;
+
+            let comments = f.context().comments();
+
+            // If the element has a debug text, preserve the same formatting as
+            // in the source code (`verbatim`). This requires us to mark all of
+            // the surrounding comments as formatted.
+            comments.mark_verbatim_node_comments_formatted(self.element.into());
+
+            // Above method doesn't mark the leading and trailing comments of the element.
+            // There can't be any leading comments for an expression element, but there
+            // can be trailing comments. For example,
+            //
+            // ```python
+            // f"""foo {
+            //     x:.3f
+            //     # trailing comment
+            // }"""
+            // ```
+            for trailing_comment in comments.trailing(self.element) {
+                trailing_comment.mark_formatted();
+            }
+
+            write!(
+                f,
+                [
+                    text(&debug_text.leading),
+                    verbatim_text(&**expression),
+                    text(&debug_text.trailing),
+                ]
+            )?;
+
+            // Even if debug text is present, any whitespace between the
+            // conversion flag and the format spec doesn't need to be preserved.
+            match conversion {
+                ConversionFlag::Str => text("!s").fmt(f)?,
+                ConversionFlag::Ascii => text("!a").fmt(f)?,
+                ConversionFlag::Repr => text("!r").fmt(f)?,
+                ConversionFlag::None => (),
+            }
+
+            if let Some(format_spec) = format_spec.as_deref() {
+                write!(f, [token(":"), verbatim_text(format_spec)])?;
+            }
+
+            token("}").fmt(f)
+        } else {
+            let comments = f.context().comments().clone();
+            let dangling_item_comments = comments.dangling(self.element);
+
+            let item = format_with(|f| {
+                let bracket_spacing = match expression.as_ref() {
+                    // If an expression starts with a `{`, we need to add a space before the
+                    // curly brace to avoid turning it into a literal curly with `{{`.
+                    //
+                    // For example,
+                    // ```python
+                    // f"{ {'x': 1, 'y': 2} }"
+                    // #  ^                ^
+                    // ```
+                    //
+                    // We need to preserve the space highlighted by `^`. The whitespace
+                    // before the closing curly brace is not strictly necessary, but it's
+                    // added to maintain consistency.
+                    Expr::Dict(_) | Expr::DictComp(_) | Expr::Set(_) | Expr::SetComp(_) => {
+                        Some(format_with(|f| {
+                            if self.context.layout().is_flat() {
+                                space().fmt(f)
+                            } else {
+                                soft_line_break_or_space().fmt(f)
+                            }
+                        }))
+                    }
+                    _ => None,
+                };
+
+                // Update the context to be inside the f-string expression element.
+                let f = &mut WithFStringState::new(
+                    FStringState::InsideExpressionElement(self.context),
+                    f,
+                );
+
+                write!(f, [bracket_spacing, expression.format()])?;
+
+                // Conversion comes first, then the format spec.
+                match conversion {
+                    ConversionFlag::Str => text("!s").fmt(f)?,
+                    ConversionFlag::Ascii => text("!a").fmt(f)?,
+                    ConversionFlag::Repr => text("!r").fmt(f)?,
+                    ConversionFlag::None => (),
+                }
+
+                if let Some(format_spec) = format_spec.as_deref() {
+                    token(":").fmt(f)?;
+
+                    f.join()
+                        .entries(
+                            format_spec
+                                .elements
+                                .iter()
+                                .map(|element| FormatFStringElement::new(element, self.context)),
+                        )
+                        .finish()?;
+
+                    // These trailing comments can only occur if the format specifier is
+                    // present. For example,
+                    //
+                    // ```python
+                    // f"{
+                    //    x:.3f
+                    //    # comment
+                    // }"
+                    // ```
+                    //
+                    // Any other trailing comments are attached to the expression itself.
+                    trailing_comments(comments.trailing(self.element)).fmt(f)?;
+                }
+
+                bracket_spacing.fmt(f)
+            });
+
+            let open_parenthesis_comments = if dangling_item_comments.is_empty() {
+                None
+            } else {
+                Some(dangling_open_parenthesis_comments(dangling_item_comments))
+            };
+
+            token("{").fmt(f)?;
+
+            {
+                let mut f = WithNodeLevel::new(NodeLevel::ParenthesizedExpression, f);
+
+                if self.context.layout().is_flat() {
+                    let mut buffer = RemoveSoftLinesBuffer::new(&mut *f);
+
+                    write!(buffer, [open_parenthesis_comments, item])?;
+                } else {
+                    group(&format_args![
+                        open_parenthesis_comments,
+                        soft_block_indent(&item)
+                    ])
+                    .fmt(&mut f)?;
+                }
+            }
+
+            token("}").fmt(f)
+        }
+    }
+}

crates/ruff_python_formatter/src/other/mod.rs

@@ -7,6 +7,7 @@ pub(crate) mod decorator;
 pub(crate) mod elif_else_clause;
 pub(crate) mod except_handler_except_handler;
 pub(crate) mod f_string;
+pub(crate) mod f_string_element;
 pub(crate) mod f_string_part;
 pub(crate) mod identifier;
 pub(crate) mod keyword;

crates/ruff_python_formatter/src/other/string_literal.rs

@@ -2,8 +2,7 @@ use ruff_python_ast::StringLiteral;
 use ruff_text_size::Ranged;
 
 use crate::prelude::*;
-use crate::preview::is_hex_codes_in_unicode_sequences_enabled;
-use crate::string::{docstring, Quoting, StringPart};
+use crate::string::{docstring, Quoting, StringNormalizer, StringPart};
 use crate::QuoteStyle;
 
 pub(crate) struct FormatStringLiteral<'a> {
@@ -50,20 +49,22 @@ impl Format<PyFormatContext<'_>> for FormatStringLiteral<'_> {
     fn fmt(&self, f: &mut PyFormatter) -> FormatResult<()> {
         let locator = f.context().locator();
 
-        let quote_style = if self.layout.is_docstring() {
-            // Per PEP 8 and PEP 257, always prefer double quotes for docstrings
+        let quote_style = f.options().quote_style();
+        let quote_style = if self.layout.is_docstring() && !quote_style.is_preserve() {
+            // Per PEP 8 and PEP 257, always prefer double quotes for docstrings,
+            // except when using quote-style=preserve
             QuoteStyle::Double
         } else {
-            f.options().quote_style()
+            quote_style
         };
 
-        let normalized = StringPart::from_source(self.value.range(), &locator).normalize(
-            self.layout.quoting(),
-            &locator,
-            quote_style,
-            f.context().docstring(),
-            is_hex_codes_in_unicode_sequences_enabled(f.context()),
-        );
+        let normalized = StringNormalizer::from_context(f.context())
+            .with_quoting(self.layout.quoting())
+            .with_preferred_quote_style(quote_style)
+            .normalize(
+                &StringPart::from_source(self.value.range(), &locator),
+                &locator,
+            );
 
         if self.layout.is_docstring() {
             docstring::format(&normalized, f)

crates/ruff_python_formatter/src/preview.rs

@@ -81,3 +81,8 @@ pub(crate) const fn is_multiline_string_handling_enabled(context: &PyFormatConte
 pub(crate) const fn is_format_module_docstring_enabled(context: &PyFormatContext) -> bool {
     context.is_preview()
 }
+
+/// Returns `true` if the [`f-string formatting`](https://github.com/astral-sh/ruff/issues/7594) preview style is enabled.
+pub(crate) fn is_f_string_formatting_enabled(context: &PyFormatContext) -> bool {
+    context.is_preview()
+}

crates/ruff_python_formatter/src/range.rs

@@ -214,9 +214,9 @@ impl<'ast> PreorderVisitor<'ast> for FindEnclosingNode<'_, 'ast> {
         // Don't pick potential docstrings as the closest enclosing node because `suite.rs` than fails to identify them as
         // docstrings and docstring formatting won't kick in.
         // Format the enclosing node instead and slice the formatted docstring from the result.
-        let is_maybe_docstring = node
-            .as_stmt_expr()
-            .is_some_and(|stmt| DocstringStmt::is_docstring_statement(stmt));
+        let is_maybe_docstring = node.as_stmt_expr().is_some_and(|stmt| {
+            DocstringStmt::is_docstring_statement(stmt, self.context.options().source_type())
+        });
 
         if is_maybe_docstring {
             return TraversalSignal::Skip;

crates/ruff_python_formatter/src/statement/suite.rs

@@ -103,7 +103,9 @@ impl FormatRule<Suite, PyFormatContext<'_>> for FormatSuite {
             }
 
             SuiteKind::Function => {
-                if let Some(docstring) = DocstringStmt::try_from_statement(first, self.kind) {
+                if let Some(docstring) =
+                    DocstringStmt::try_from_statement(first, self.kind, source_type)
+                {
                     SuiteChildStatement::Docstring(docstring)
                 } else {
                     SuiteChildStatement::Other(first)
@@ -111,7 +113,9 @@ impl FormatRule<Suite, PyFormatContext<'_>> for FormatSuite {
             }
 
             SuiteKind::Class => {
-                if let Some(docstring) = DocstringStmt::try_from_statement(first, self.kind) {
+                if let Some(docstring) =
+                    DocstringStmt::try_from_statement(first, self.kind, source_type)
+                {
                     if !comments.has_leading(first)
                         && lines_before(first.start(), source) > 1
                         && !source_type.is_stub()
@@ -143,7 +147,9 @@ impl FormatRule<Suite, PyFormatContext<'_>> for FormatSuite {
             }
             SuiteKind::TopLevel => {
                 if is_format_module_docstring_enabled(f.context()) {
-                    if let Some(docstring) = DocstringStmt::try_from_statement(first, self.kind) {
+                    if let Some(docstring) =
+                        DocstringStmt::try_from_statement(first, self.kind, source_type)
+                    {
                         SuiteChildStatement::Docstring(docstring)
                     } else {
                         SuiteChildStatement::Other(first)
@@ -184,7 +190,8 @@ impl FormatRule<Suite, PyFormatContext<'_>> for FormatSuite {
                 true
             } else if is_module_docstring_newlines_enabled(f.context())
                 && self.kind == SuiteKind::TopLevel
-                && DocstringStmt::try_from_statement(first.statement(), self.kind).is_some()
+                && DocstringStmt::try_from_statement(first.statement(), self.kind, source_type)
+                    .is_some()
             {
                 // Only in preview mode, insert a newline after a module level docstring, but treat
                 // it as a docstring otherwise. See: https://github.com/psf/black/pull/3932.
@@ -734,7 +741,16 @@ pub(crate) struct DocstringStmt<'a> {
 
 impl<'a> DocstringStmt<'a> {
     /// Checks if the statement is a simple string that can be formatted as a docstring
-    fn try_from_statement(stmt: &'a Stmt, suite_kind: SuiteKind) -> Option<DocstringStmt<'a>> {
+    fn try_from_statement(
+        stmt: &'a Stmt,
+        suite_kind: SuiteKind,
+        source_type: PySourceType,
+    ) -> Option<DocstringStmt<'a>> {
+        // Notebooks don't have a concept of modules, therefore, don't recognise the first string as the module docstring.
+        if source_type.is_ipynb() && suite_kind == SuiteKind::TopLevel {
+            return None;
+        }
+
         let Stmt::Expr(ast::StmtExpr { value, .. }) = stmt else {
             return None;
         };
@@ -752,7 +768,11 @@ impl<'a> DocstringStmt<'a> {
         }
     }
 
-    pub(crate) fn is_docstring_statement(stmt: &StmtExpr) -> bool {
+    pub(crate) fn is_docstring_statement(stmt: &StmtExpr, source_type: PySourceType) -> bool {
+        if source_type.is_ipynb() {
+            return false;
+        }
+
         if let Expr::StringLiteral(ast::ExprStringLiteral { value, .. }) = stmt.value.as_ref() {
             !value.is_implicit_concatenated()
         } else {

crates/ruff_python_formatter/src/string/any.rs

@@ -0,0 +1,212 @@
+use std::iter::FusedIterator;
+
+use memchr::memchr2;
+
+use ruff_python_ast::{
+    self as ast, AnyNodeRef, Expr, ExprBytesLiteral, ExprFString, ExprStringLiteral, ExpressionRef,
+    StringLiteral,
+};
+use ruff_source_file::Locator;
+use ruff_text_size::{Ranged, TextLen, TextRange};
+
+use crate::expression::expr_f_string::f_string_quoting;
+use crate::other::f_string::FormatFString;
+use crate::other::string_literal::{FormatStringLiteral, StringLiteralKind};
+use crate::prelude::*;
+use crate::string::{Quoting, StringPrefix, StringQuotes};
+
+/// Represents any kind of string expression. This could be either a string,
+/// bytes or f-string.
+#[derive(Copy, Clone, Debug)]
+pub(crate) enum AnyString<'a> {
+    String(&'a ExprStringLiteral),
+    Bytes(&'a ExprBytesLiteral),
+    FString(&'a ExprFString),
+}
+
+impl<'a> AnyString<'a> {
+    /// Creates a new [`AnyString`] from the given [`Expr`].
+    ///
+    /// Returns `None` if the expression is not either a string, bytes or f-string.
+    pub(crate) fn from_expression(expression: &'a Expr) -> Option<AnyString<'a>> {
+        match expression {
+            Expr::StringLiteral(string) => Some(AnyString::String(string)),
+            Expr::BytesLiteral(bytes) => Some(AnyString::Bytes(bytes)),
+            Expr::FString(fstring) => Some(AnyString::FString(fstring)),
+            _ => None,
+        }
+    }
+
+    /// Returns `true` if the string is implicitly concatenated.
+    pub(crate) fn is_implicit_concatenated(self) -> bool {
+        match self {
+            Self::String(ExprStringLiteral { value, .. }) => value.is_implicit_concatenated(),
+            Self::Bytes(ExprBytesLiteral { value, .. }) => value.is_implicit_concatenated(),
+            Self::FString(ExprFString { value, .. }) => value.is_implicit_concatenated(),
+        }
+    }
+
+    /// Returns the quoting to be used for this string.
+    pub(super) fn quoting(self, locator: &Locator<'_>) -> Quoting {
+        match self {
+            Self::String(_) | Self::Bytes(_) => Quoting::CanChange,
+            Self::FString(f_string) => f_string_quoting(f_string, locator),
+        }
+    }
+
+    /// Returns a vector of all the [`AnyStringPart`] of this string.
+    pub(super) fn parts(self, quoting: Quoting) -> AnyStringPartsIter<'a> {
+        match self {
+            Self::String(ExprStringLiteral { value, .. }) => {
+                AnyStringPartsIter::String(value.iter())
+            }
+            Self::Bytes(ExprBytesLiteral { value, .. }) => AnyStringPartsIter::Bytes(value.iter()),
+            Self::FString(ExprFString { value, .. }) => {
+                AnyStringPartsIter::FString(value.iter(), quoting)
+            }
+        }
+    }
+
+    pub(crate) fn is_multiline(self, source: &str) -> bool {
+        match self {
+            AnyString::String(_) | AnyString::Bytes(_) => {
+                let contents = &source[self.range()];
+                let prefix = StringPrefix::parse(contents);
+                let quotes = StringQuotes::parse(
+                    &contents[TextRange::new(prefix.text_len(), contents.text_len())],
+                );
+
+                quotes.is_some_and(StringQuotes::is_triple)
+                    && memchr2(b'\n', b'\r', contents.as_bytes()).is_some()
+            }
+            AnyString::FString(fstring) => {
+                memchr2(b'\n', b'\r', source[fstring.range].as_bytes()).is_some()
+            }
+        }
+    }
+}
+
+impl Ranged for AnyString<'_> {
+    fn range(&self) -> TextRange {
+        match self {
+            Self::String(expr) => expr.range(),
+            Self::Bytes(expr) => expr.range(),
+            Self::FString(expr) => expr.range(),
+        }
+    }
+}
+
+impl<'a> From<&AnyString<'a>> for AnyNodeRef<'a> {
+    fn from(value: &AnyString<'a>) -> Self {
+        match value {
+            AnyString::String(expr) => AnyNodeRef::ExprStringLiteral(expr),
+            AnyString::Bytes(expr) => AnyNodeRef::ExprBytesLiteral(expr),
+            AnyString::FString(expr) => AnyNodeRef::ExprFString(expr),
+        }
+    }
+}
+
+impl<'a> From<AnyString<'a>> for AnyNodeRef<'a> {
+    fn from(value: AnyString<'a>) -> Self {
+        AnyNodeRef::from(&value)
+    }
+}
+
+impl<'a> From<&AnyString<'a>> for ExpressionRef<'a> {
+    fn from(value: &AnyString<'a>) -> Self {
+        match value {
+            AnyString::String(expr) => ExpressionRef::StringLiteral(expr),
+            AnyString::Bytes(expr) => ExpressionRef::BytesLiteral(expr),
+            AnyString::FString(expr) => ExpressionRef::FString(expr),
+        }
+    }
+}
+
+pub(super) enum AnyStringPartsIter<'a> {
+    String(std::slice::Iter<'a, StringLiteral>),
+    Bytes(std::slice::Iter<'a, ast::BytesLiteral>),
+    FString(std::slice::Iter<'a, ast::FStringPart>, Quoting),
+}
+
+impl<'a> Iterator for AnyStringPartsIter<'a> {
+    type Item = AnyStringPart<'a>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        let part = match self {
+            Self::String(inner) => {
+                let part = inner.next()?;
+                AnyStringPart::String {
+                    part,
+                    layout: StringLiteralKind::String,
+                }
+            }
+            Self::Bytes(inner) => AnyStringPart::Bytes(inner.next()?),
+            Self::FString(inner, quoting) => {
+                let part = inner.next()?;
+                match part {
+                    ast::FStringPart::Literal(string_literal) => AnyStringPart::String {
+                        part: string_literal,
+                        layout: StringLiteralKind::InImplicitlyConcatenatedFString(*quoting),
+                    },
+                    ast::FStringPart::FString(f_string) => AnyStringPart::FString {
+                        part: f_string,
+                        quoting: *quoting,
+                    },
+                }
+            }
+        };
+
+        Some(part)
+    }
+}
+
+impl FusedIterator for AnyStringPartsIter<'_> {}
+
+/// Represents any kind of string which is part of an implicitly concatenated
+/// string. This could be either a string, bytes or f-string.
+///
+/// This is constructed from the [`AnyString::parts`] method on [`AnyString`].
+#[derive(Clone, Debug)]
+pub(super) enum AnyStringPart<'a> {
+    String {
+        part: &'a ast::StringLiteral,
+        layout: StringLiteralKind,
+    },
+    Bytes(&'a ast::BytesLiteral),
+    FString {
+        part: &'a ast::FString,
+        quoting: Quoting,
+    },
+}
+
+impl<'a> From<&AnyStringPart<'a>> for AnyNodeRef<'a> {
+    fn from(value: &AnyStringPart<'a>) -> Self {
+        match value {
+            AnyStringPart::String { part, .. } => AnyNodeRef::StringLiteral(part),
+            AnyStringPart::Bytes(part) => AnyNodeRef::BytesLiteral(part),
+            AnyStringPart::FString { part, .. } => AnyNodeRef::FString(part),
+        }
+    }
+}
+
+impl Ranged for AnyStringPart<'_> {
+    fn range(&self) -> TextRange {
+        match self {
+            Self::String { part, .. } => part.range(),
+            Self::Bytes(part) => part.range(),
+            Self::FString { part, .. } => part.range(),
+        }
+    }
+}
+
+impl Format<PyFormatContext<'_>> for AnyStringPart<'_> {
+    fn fmt(&self, f: &mut PyFormatter) -> FormatResult<()> {
+        match self {
+            AnyStringPart::String { part, layout } => {
+                FormatStringLiteral::new(part, *layout).fmt(f)
+            }
+            AnyStringPart::Bytes(bytes_literal) => bytes_literal.format().fmt(f),
+            AnyStringPart::FString { part, quoting } => FormatFString::new(part, *quoting).fmt(f),
+        }
+    }
+}

crates/ruff_python_formatter/src/string/docstring.rs

@@ -2,11 +2,13 @@
 // "reStructuredText."
 #![allow(clippy::doc_markdown)]
 
+use std::cmp::Ordering;
 use std::{borrow::Cow, collections::VecDeque};
 
+use itertools::Itertools;
+
 use ruff_formatter::printer::SourceMapGeneration;
 use ruff_python_parser::ParseError;
-
 use {once_cell::sync::Lazy, regex::Regex};
 use {
     ruff_formatter::{write, FormatOptions, IndentStyle, LineWidth, Printed},
@@ -80,9 +82,7 @@ use super::{NormalizedString, QuoteChar};
 /// ```
 ///
 /// Tabs are counted by padding them to the next multiple of 8 according to
-/// [`str.expandtabs`](https://docs.python.org/3/library/stdtypes.html#str.expandtabs). When
-/// we see indentation that contains a tab or any other none ascii-space whitespace we rewrite the
-/// string.
+/// [`str.expandtabs`](https://docs.python.org/3/library/stdtypes.html#str.expandtabs).
 ///
 /// Additionally, if any line in the docstring has less indentation than the docstring
 /// (effectively a negative indentation wrt. to the current level), we pad all lines to the
@@ -104,8 +104,12 @@ use super::{NormalizedString, QuoteChar};
 ///         line c
 ///    """
 /// ```
+/// The indentation is rewritten to all-spaces when using [`IndentStyle::Space`].
+/// The formatter preserves tab-indentations when using [`IndentStyle::Tab`], but doesn't convert
+/// `indent-width * spaces` to tabs because doing so could break ASCII art and other docstrings
+/// that use spaces for alignment.
 pub(crate) fn format(normalized: &NormalizedString, f: &mut PyFormatter) -> FormatResult<()> {
-    let docstring = &normalized.text;
+    let docstring = &normalized.text();
 
     // Black doesn't change the indentation of docstrings that contain an escaped newline
     if contains_unescaped_newline(docstring) {
@@ -121,7 +125,7 @@ pub(crate) fn format(normalized: &NormalizedString, f: &mut PyFormatter) -> Form
     let mut lines = docstring.split('\n').peekable();
 
     // Start the string
-    write!(f, [normalized.prefix, normalized.quotes])?;
+    write!(f, [normalized.prefix(), normalized.quotes()])?;
     // We track where in the source docstring we are (in source code byte offsets)
     let mut offset = normalized.start();
 
@@ -137,7 +141,7 @@ pub(crate) fn format(normalized: &NormalizedString, f: &mut PyFormatter) -> Form
 
     // Edge case: The first line is `""" "content`, so we need to insert chaperone space that keep
     // inner quotes and closing quotes from getting to close to avoid `""""content`
-    if trim_both.starts_with(normalized.quotes.quote_char.as_char()) {
+    if trim_both.starts_with(normalized.quotes().quote_char.as_char()) {
         space().fmt(f)?;
     }
 
@@ -164,7 +168,7 @@ pub(crate) fn format(normalized: &NormalizedString, f: &mut PyFormatter) -> Form
         {
             space().fmt(f)?;
         }
-        normalized.quotes.fmt(f)?;
+        normalized.quotes().fmt(f)?;
         return Ok(());
     }
 
@@ -176,21 +180,21 @@ pub(crate) fn format(normalized: &NormalizedString, f: &mut PyFormatter) -> Form
     // align it with the docstring statement. Conversely, if all lines are over-indented, we strip
     // the extra indentation. We call this stripped indentation since it's relative to the block
     // indent printer-made indentation.
-    let stripped_indentation_length = lines
+    let stripped_indentation = lines
         .clone()
         // We don't want to count whitespace-only lines as miss-indented
         .filter(|line| !line.trim().is_empty())
-        .map(indentation_length)
-        .min()
+        .map(Indentation::from_str)
+        .min_by_key(|indentation| indentation.width())
         .unwrap_or_default();
 
     DocstringLinePrinter {
         f,
         action_queue: VecDeque::new(),
         offset,
-        stripped_indentation_length,
+        stripped_indentation,
         already_normalized,
-        quote_char: normalized.quotes.quote_char,
+        quote_char: normalized.quotes().quote_char,
         code_example: CodeExample::default(),
     }
     .add_iter(lines)?;
@@ -203,7 +207,7 @@ pub(crate) fn format(normalized: &NormalizedString, f: &mut PyFormatter) -> Form
         space().fmt(f)?;
     }
 
-    write!(f, [normalized.quotes])
+    write!(f, [normalized.quotes()])
 }
 
 fn contains_unescaped_newline(haystack: &str) -> bool {
@@ -240,9 +244,9 @@ struct DocstringLinePrinter<'ast, 'buf, 'fmt, 'src> {
     /// printed.
     offset: TextSize,
 
-    /// Indentation alignment (in columns) based on the least indented line in the
+    /// Indentation alignment based on the least indented line in the
     /// docstring.
-    stripped_indentation_length: usize,
+    stripped_indentation: Indentation,
 
     /// Whether the docstring is overall already considered normalized. When it
     /// is, the formatter can take a fast path.
@@ -345,7 +349,7 @@ impl<'ast, 'buf, 'fmt, 'src> DocstringLinePrinter<'ast, 'buf, 'fmt, 'src> {
                             };
                             // This looks suspicious, but it's consistent with the whitespace
                             // normalization that will occur anyway.
-                            let indent = " ".repeat(min_indent);
+                            let indent = " ".repeat(min_indent.width());
                             for docline in formatted_lines {
                                 self.print_one(
                                     &docline.map(|line| std::format!("{indent}{line}")),
@@ -355,7 +359,7 @@ impl<'ast, 'buf, 'fmt, 'src> DocstringLinePrinter<'ast, 'buf, 'fmt, 'src> {
                         CodeExampleKind::Markdown(fenced) => {
                             // This looks suspicious, but it's consistent with the whitespace
                             // normalization that will occur anyway.
-                            let indent = " ".repeat(fenced.opening_fence_indent);
+                            let indent = " ".repeat(fenced.opening_fence_indent.width());
                             for docline in formatted_lines {
                                 self.print_one(
                                     &docline.map(|line| std::format!("{indent}{line}")),
@@ -387,12 +391,58 @@ impl<'ast, 'buf, 'fmt, 'src> DocstringLinePrinter<'ast, 'buf, 'fmt, 'src> {
             };
         }
 
-        let tab_or_non_ascii_space = trim_end
-            .chars()
-            .take_while(|c| c.is_whitespace())
-            .any(|c| c != ' ');
+        let indent_offset = match self.f.options().indent_style() {
+            // Normalize all indent to spaces.
+            IndentStyle::Space => {
+                let tab_or_non_ascii_space = trim_end
+                    .chars()
+                    .take_while(|c| c.is_whitespace())
+                    .any(|c| c != ' ');
 
-        if tab_or_non_ascii_space {
+                if tab_or_non_ascii_space {
+                    None
+                } else {
+                    // It's guaranteed that the `indent` is all spaces because `tab_or_non_ascii_space` is
+                    // `false` (indent contains neither tabs nor non-space whitespace).
+                    let stripped_indentation_len = self.stripped_indentation.text_len();
+
+                    // Take the string with the trailing whitespace removed, then also
+                    // skip the leading whitespace.
+                    Some(stripped_indentation_len)
+                }
+            }
+            IndentStyle::Tab => {
+                let line_indent = Indentation::from_str(trim_end);
+
+                let non_ascii_whitespace = trim_end
+                    .chars()
+                    .take_while(|c| c.is_whitespace())
+                    .any(|c| !matches!(c, ' ' | '\t'));
+
+                let trimmed = line_indent.trim_start(self.stripped_indentation);
+
+                // Preserve tabs that are used for indentation, but only if the indent isn't
+                // * a mix of tabs and spaces
+                // * the `stripped_indentation` is a prefix of the line's indent
+                // * the trimmed indent isn't spaces followed by tabs because that would result in a
+                //   mixed tab, spaces, tab indentation, resulting in instabilities.
+                let preserve_indent = !non_ascii_whitespace
+                    && trimmed.is_some_and(|trimmed| !trimmed.is_spaces_tabs());
+                preserve_indent.then_some(self.stripped_indentation.text_len())
+            }
+        };
+
+        if let Some(indent_offset) = indent_offset {
+            // Take the string with the trailing whitespace removed, then also
+            // skip the leading whitespace.
+            if self.already_normalized {
+                let trimmed_line_range =
+                    TextRange::at(line.offset, trim_end.text_len()).add_start(indent_offset);
+                source_text_slice(trimmed_line_range).fmt(self.f)?;
+            } else {
+                text(&trim_end[indent_offset.to_usize()..]).fmt(self.f)?;
+            }
+        } else {
             // We strip the indentation that is shared with the docstring
             // statement, unless a line was indented less than the docstring
             // statement, in which case we strip only this much indentation to
@@ -400,24 +450,11 @@ impl<'ast, 'buf, 'fmt, 'src> DocstringLinePrinter<'ast, 'buf, 'fmt, 'src> {
             // overindented, in which case we strip the additional whitespace
             // (see example in [`format_docstring`] doc comment). We then
             // prepend the in-docstring indentation to the string.
-            let indent_len = indentation_length(trim_end) - self.stripped_indentation_length;
+            let indent_len =
+                Indentation::from_str(trim_end).width() - self.stripped_indentation.width();
             let in_docstring_indent = " ".repeat(indent_len) + trim_end.trim_start();
             text(&in_docstring_indent).fmt(self.f)?;
-        } else {
-            // It's guaranteed that the `indent` is all spaces because `tab_or_non_ascii_space` is
-            // `false` (indent contains neither tabs nor non-space whitespace).
-
-            // Take the string with the trailing whitespace removed, then also
-            // skip the leading whitespace.
-            let trimmed_line_range = TextRange::at(line.offset, trim_end.text_len())
-                .add_start(TextSize::try_from(self.stripped_indentation_length).unwrap());
-            if self.already_normalized {
-                source_text_slice(trimmed_line_range).fmt(self.f)?;
-            } else {
-                // All indents are ascii spaces, so the slicing is correct.
-                text(&trim_end[self.stripped_indentation_length..]).fmt(self.f)?;
-            }
-        }
+        };
 
         // We handled the case that the closing quotes are on their own line
         // above (the last line is empty except for whitespace). If they are on
@@ -898,8 +935,7 @@ struct CodeExampleRst<'src> {
     /// The lines that have been seen so far that make up the block.
     lines: Vec<CodeExampleLine<'src>>,
 
-    /// The indent of the line "opening" this block measured via
-    /// `indentation_length` (in columns).
+    /// The indent of the line "opening" this block in columns.
     ///
     /// It can either be the indent of a line ending with `::` (for a literal
     /// block) or the indent of a line starting with `.. ` (a directive).
@@ -907,9 +943,9 @@ struct CodeExampleRst<'src> {
     /// The content body of a block needs to be indented more than the line
     /// opening the block, so we use this indentation to look for indentation
     /// that is "more than" it.
-    opening_indent: usize,
+    opening_indent: Indentation,
 
-    /// The minimum indent of the block measured via `indentation_length`.
+    /// The minimum indent of the block in columns.
     ///
     /// This is `None` until the first such line is seen. If no such line is
     /// found, then we consider it an invalid block and bail out of trying to
@@ -926,7 +962,7 @@ struct CodeExampleRst<'src> {
     /// When the code snippet has been extracted, it is re-built before being
     /// reformatted. The minimum indent is stripped from each line when it is
     /// re-built.
-    min_indent: Option<usize>,
+    min_indent: Option<Indentation>,
 
     /// Whether this is a directive block or not. When not a directive, this is
     /// a literal block. The main difference between them is that they start
@@ -975,7 +1011,7 @@ impl<'src> CodeExampleRst<'src> {
         }
         Some(CodeExampleRst {
             lines: vec![],
-            opening_indent: indentation_length(opening_indent),
+            opening_indent: Indentation::from_str(opening_indent),
             min_indent: None,
             is_directive: false,
         })
@@ -1013,7 +1049,7 @@ impl<'src> CodeExampleRst<'src> {
         }
         Some(CodeExampleRst {
             lines: vec![],
-            opening_indent: indentation_length(original.line),
+            opening_indent: Indentation::from_str(original.line),
             min_indent: None,
             is_directive: true,
         })
@@ -1033,7 +1069,7 @@ impl<'src> CodeExampleRst<'src> {
             line.code = if line.original.line.trim().is_empty() {
                 ""
             } else {
-                indentation_trim(min_indent, line.original.line)
+                min_indent.trim_start_str(line.original.line)
             };
         }
         &self.lines
@@ -1070,7 +1106,9 @@ impl<'src> CodeExampleRst<'src> {
             // an empty line followed by an unindented non-empty line.
             if let Some(next) = original.next {
                 let (next_indent, next_rest) = indent_with_suffix(next);
-                if !next_rest.is_empty() && indentation_length(next_indent) <= self.opening_indent {
+                if !next_rest.is_empty()
+                    && Indentation::from_str(next_indent) <= self.opening_indent
+                {
                     self.push_format_action(queue);
                     return None;
                 }
@@ -1082,7 +1120,7 @@ impl<'src> CodeExampleRst<'src> {
             queue.push_back(CodeExampleAddAction::Kept);
             return Some(self);
         }
-        let indent_len = indentation_length(indent);
+        let indent_len = Indentation::from_str(indent);
         if indent_len <= self.opening_indent {
             // If we find an unindented non-empty line at the same (or less)
             // indentation of the opening line at this point, then we know it
@@ -1144,7 +1182,7 @@ impl<'src> CodeExampleRst<'src> {
             queue.push_back(CodeExampleAddAction::Print { original });
             return Some(self);
         }
-        let min_indent = indentation_length(indent);
+        let min_indent = Indentation::from_str(indent);
         // At this point, we found a non-empty line. The only thing we require
         // is that its indentation is strictly greater than the indentation of
         // the line containing the `::`. Otherwise, we treat this as an invalid
@@ -1218,12 +1256,11 @@ struct CodeExampleMarkdown<'src> {
     /// The lines that have been seen so far that make up the block.
     lines: Vec<CodeExampleLine<'src>>,
 
-    /// The indent of the line "opening" fence of this block measured via
-    /// `indentation_length` (in columns).
+    /// The indent of the line "opening" fence of this block in columns.
     ///
     /// This indentation is trimmed from the indentation of every line in the
     /// body of the code block,
-    opening_fence_indent: usize,
+    opening_fence_indent: Indentation,
 
     /// The kind of fence, backticks or tildes, used for this block. We need to
     /// keep track of which kind was used to open the block in order to look
@@ -1292,7 +1329,7 @@ impl<'src> CodeExampleMarkdown<'src> {
         };
         Some(CodeExampleMarkdown {
             lines: vec![],
-            opening_fence_indent: indentation_length(opening_fence_indent),
+            opening_fence_indent: Indentation::from_str(opening_fence_indent),
             fence_kind,
             fence_len,
         })
@@ -1325,7 +1362,7 @@ impl<'src> CodeExampleMarkdown<'src> {
         // its indent normalized. And, at the time of writing, a subsequent
         // formatting run undoes this indentation, thus violating idempotency.
         if !original.line.trim_whitespace().is_empty()
-            && indentation_length(original.line) < self.opening_fence_indent
+            && Indentation::from_str(original.line) < self.opening_fence_indent
         {
             queue.push_back(self.into_reset_action());
             queue.push_back(CodeExampleAddAction::Print { original });
@@ -1371,7 +1408,7 @@ impl<'src> CodeExampleMarkdown<'src> {
         // Unlike reStructuredText blocks, for Markdown fenced code blocks, the
         // indentation that we want to strip from each line is known when the
         // block is opened. So we can strip it as we collect lines.
-        let code = indentation_trim(self.opening_fence_indent, original.line);
+        let code = self.opening_fence_indent.trim_start_str(original.line);
         self.lines.push(CodeExampleLine { original, code });
     }
 
@@ -1486,7 +1523,6 @@ enum CodeExampleAddAction<'src> {
     /// results in that code example becoming invalid. In this case,
     /// we don't want to treat it as a code example, but instead write
     /// back the lines to the docstring unchanged.
-    #[allow(dead_code)] // FIXME: remove when reStructuredText support is added
     Reset {
         /// The lines of code that we collected but should be printed back to
         /// the docstring as-is and not formatted.
@@ -1533,57 +1569,245 @@ fn docstring_format_source(
 /// that avoids `content""""` and `content\"""`. This does only applies to un-escaped backslashes,
 /// so `content\\ """` doesn't need a space while `content\\\ """` does.
 fn needs_chaperone_space(normalized: &NormalizedString, trim_end: &str) -> bool {
-    trim_end.ends_with(normalized.quotes.quote_char.as_char())
+    trim_end.ends_with(normalized.quotes().quote_char.as_char())
         || trim_end.chars().rev().take_while(|c| *c == '\\').count() % 2 == 1
 }
 
-/// Returns the indentation's visual width in columns/spaces.
-///
-/// For docstring indentation, black counts spaces as 1 and tabs by increasing the indentation up
-/// to the next multiple of 8. This is effectively a port of
-/// [`str.expandtabs`](https://docs.python.org/3/library/stdtypes.html#str.expandtabs),
-/// which black [calls with the default tab width of 8](https://github.com/psf/black/blob/c36e468794f9256d5e922c399240d49782ba04f1/src/black/strings.py#L61).
-fn indentation_length(line: &str) -> usize {
-    let mut indentation = 0usize;
-    for char in line.chars() {
-        if char == '\t' {
-            // Pad to the next multiple of tab_width
-            indentation += 8 - (indentation.rem_euclid(8));
-        } else if char.is_whitespace() {
-            indentation += char.len_utf8();
-        } else {
-            break;
-        }
-    }
-    indentation
+#[derive(Copy, Clone, Debug)]
+enum Indentation {
+    /// Space only indentation or an empty indentation.
+    ///
+    /// The value is the number of spaces.
+    Spaces(usize),
+
+    /// Tabs only indentation.
+    Tabs(usize),
+
+    /// Indentation that uses tabs followed by spaces.
+    /// Also known as smart tabs where tabs are used for indents, and spaces for alignment.
+    TabSpaces { tabs: usize, spaces: usize },
+
+    /// Indentation that uses spaces followed by tabs.
+    SpacesTabs { spaces: usize, tabs: usize },
+
+    /// Mixed indentation of tabs and spaces.
+    Mixed {
+        /// The visual width of the indentation in columns.
+        width: usize,
+
+        /// The length of the indentation in bytes
+        len: TextSize,
+    },
 }
 
-/// Trims at most `indent_len` indentation from the beginning of `line`.
-///
-/// This treats indentation in precisely the same way as `indentation_length`.
-/// As such, it is expected that `indent_len` is computed from
-/// `indentation_length`. This is useful when one needs to trim some minimum
-/// level of indentation from a code snippet collected from a docstring before
-/// attempting to reformat it.
-fn indentation_trim(indent_len: usize, line: &str) -> &str {
-    let mut seen_indent_len = 0;
-    let mut trimmed = line;
-    for char in line.chars() {
-        if seen_indent_len >= indent_len {
-            return trimmed;
+impl Indentation {
+    const TAB_INDENT_WIDTH: usize = 8;
+
+    fn from_str(s: &str) -> Self {
+        let mut iter = s.chars().peekable();
+
+        let spaces = iter.peeking_take_while(|c| *c == ' ').count();
+        let tabs = iter.peeking_take_while(|c| *c == '\t').count();
+
+        if tabs == 0 {
+            // No indent, or spaces only indent
+            return Self::Spaces(spaces);
         }
-        if char == '\t' {
-            // Pad to the next multiple of tab_width
-            seen_indent_len += 8 - (seen_indent_len.rem_euclid(8));
-            trimmed = &trimmed[1..];
-        } else if char.is_whitespace() {
-            seen_indent_len += char.len_utf8();
-            trimmed = &trimmed[char.len_utf8()..];
-        } else {
-            break;
+
+        let align_spaces = iter.peeking_take_while(|c| *c == ' ').count();
+
+        if spaces == 0 {
+            if align_spaces == 0 {
+                return Self::Tabs(tabs);
+            }
+
+            // At this point it's either a smart tab (tabs followed by spaces) or a wild mix of tabs and spaces.
+            if iter.peek().copied() != Some('\t') {
+                return Self::TabSpaces {
+                    tabs,
+                    spaces: align_spaces,
+                };
+            }
+        } else if align_spaces == 0 {
+            return Self::SpacesTabs { spaces, tabs };
+        }
+
+        // Sequence of spaces.. tabs, spaces, tabs...
+        let mut width = spaces + tabs * Self::TAB_INDENT_WIDTH + align_spaces;
+        // SAFETY: Safe because Ruff doesn't support files larger than 4GB.
+        let mut len = TextSize::try_from(spaces + tabs + align_spaces).unwrap();
+
+        for char in iter {
+            if char == '\t' {
+                // Pad to the next multiple of tab_width
+                width += Self::TAB_INDENT_WIDTH - (width.rem_euclid(Self::TAB_INDENT_WIDTH));
+                len += '\t'.text_len();
+            } else if char.is_whitespace() {
+                width += char.len_utf8();
+                len += char.text_len();
+            } else {
+                break;
+            }
+        }
+
+        // Mixed tabs and spaces
+        Self::Mixed { width, len }
+    }
+
+    /// Returns the indentation's visual width in columns/spaces.
+    ///
+    /// For docstring indentation, black counts spaces as 1 and tabs by increasing the indentation up
+    /// to the next multiple of 8. This is effectively a port of
+    /// [`str.expandtabs`](https://docs.python.org/3/library/stdtypes.html#str.expandtabs),
+    /// which black [calls with the default tab width of 8](https://github.com/psf/black/blob/c36e468794f9256d5e922c399240d49782ba04f1/src/black/strings.py#L61).
+    const fn width(self) -> usize {
+        match self {
+            Self::Spaces(count) => count,
+            Self::Tabs(count) => count * Self::TAB_INDENT_WIDTH,
+            Self::TabSpaces { tabs, spaces } => tabs * Self::TAB_INDENT_WIDTH + spaces,
+            Self::SpacesTabs { spaces, tabs } => {
+                let mut indent = spaces;
+                indent += Self::TAB_INDENT_WIDTH - indent.rem_euclid(Self::TAB_INDENT_WIDTH);
+                indent + (tabs - 1) * Self::TAB_INDENT_WIDTH
+            }
+            Self::Mixed { width, .. } => width,
         }
     }
-    trimmed
+
+    /// Returns the length of the indentation in bytes.
+    ///
+    /// # Panics
+    /// If the indentation is longer than 4GB.
+    fn text_len(self) -> TextSize {
+        let len = match self {
+            Self::Spaces(count) => count,
+            Self::Tabs(count) => count,
+            Self::TabSpaces { tabs, spaces } => tabs + spaces,
+            Self::SpacesTabs { spaces, tabs } => spaces + tabs,
+            Self::Mixed { len, .. } => return len,
+        };
+
+        TextSize::try_from(len).unwrap()
+    }
+
+    /// Trims the indent of `rhs` by `self`.
+    ///
+    /// Returns `None` if `self` is not a prefix of `rhs` or either `self` or `rhs` use mixed indentation.
+    fn trim_start(self, rhs: Self) -> Option<Self> {
+        let (left_tabs, left_spaces) = match self {
+            Self::Spaces(spaces) => (0usize, spaces),
+            Self::Tabs(tabs) => (tabs, 0usize),
+            Self::TabSpaces { tabs, spaces } => (tabs, spaces),
+            // Handle spaces here because it is the only indent where the spaces come before the tabs.
+            Self::SpacesTabs {
+                spaces: left_spaces,
+                tabs: left_tabs,
+            } => {
+                return match rhs {
+                    Self::Spaces(right_spaces) => {
+                        left_spaces.checked_sub(right_spaces).map(|spaces| {
+                            if spaces == 0 {
+                                Self::Tabs(left_tabs)
+                            } else {
+                                Self::SpacesTabs {
+                                    tabs: left_tabs,
+                                    spaces,
+                                }
+                            }
+                        })
+                    }
+                    Self::SpacesTabs {
+                        spaces: right_spaces,
+                        tabs: right_tabs,
+                    } => left_spaces.checked_sub(right_spaces).and_then(|spaces| {
+                        let tabs = left_tabs.checked_sub(right_tabs)?;
+
+                        Some(if spaces == 0 {
+                            if tabs == 0 {
+                                Self::Spaces(0)
+                            } else {
+                                Self::Tabs(tabs)
+                            }
+                        } else {
+                            Self::SpacesTabs { spaces, tabs }
+                        })
+                    }),
+
+                    _ => None,
+                }
+            }
+            Self::Mixed { .. } => return None,
+        };
+
+        let (right_tabs, right_spaces) = match rhs {
+            Self::Spaces(spaces) => (0usize, spaces),
+            Self::Tabs(tabs) => (tabs, 0usize),
+            Self::TabSpaces { tabs, spaces } => (tabs, spaces),
+            Self::SpacesTabs { .. } | Self::Mixed { .. } => return None,
+        };
+
+        let tabs = left_tabs.checked_sub(right_tabs)?;
+        let spaces = left_spaces.checked_sub(right_spaces)?;
+
+        Some(if tabs == 0 {
+            Self::Spaces(spaces)
+        } else if spaces == 0 {
+            Self::Tabs(tabs)
+        } else {
+            Self::TabSpaces { tabs, spaces }
+        })
+    }
+
+    /// Trims at most `indent_len` indentation from the beginning of `line`.
+    ///
+    /// This is useful when one needs to trim some minimum
+    /// level of indentation from a code snippet collected from a docstring before
+    /// attempting to reformat it.
+    fn trim_start_str(self, line: &str) -> &str {
+        let mut seen_indent_len = 0;
+        let mut trimmed = line;
+        let indent_len = self.width();
+
+        for char in line.chars() {
+            if seen_indent_len >= indent_len {
+                return trimmed;
+            }
+            if char == '\t' {
+                // Pad to the next multiple of tab_width
+                seen_indent_len +=
+                    Self::TAB_INDENT_WIDTH - (seen_indent_len.rem_euclid(Self::TAB_INDENT_WIDTH));
+                trimmed = &trimmed[1..];
+            } else if char.is_whitespace() {
+                seen_indent_len += char.len_utf8();
+                trimmed = &trimmed[char.len_utf8()..];
+            } else {
+                break;
+            }
+        }
+        trimmed
+    }
+
+    const fn is_spaces_tabs(self) -> bool {
+        matches!(self, Self::SpacesTabs { .. })
+    }
+}
+
+impl PartialOrd for Indentation {
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        Some(self.width().cmp(&other.width()))
+    }
+}
+
+impl PartialEq for Indentation {
+    fn eq(&self, other: &Self) -> bool {
+        self.width() == other.width()
+    }
+}
+
+impl Default for Indentation {
+    fn default() -> Self {
+        Self::Spaces(0)
+    }
 }
 
 /// Returns the indentation of the given line and everything following it.
@@ -1613,14 +1837,13 @@ fn is_rst_option(line: &str) -> bool {
 
 #[cfg(test)]
 mod tests {
-
-    use super::indentation_length;
+    use crate::string::docstring::Indentation;
 
     #[test]
     fn test_indentation_like_black() {
-        assert_eq!(indentation_length("\t \t  \t"), 24);
-        assert_eq!(indentation_length("\t        \t"), 24);
-        assert_eq!(indentation_length("\t\t\t"), 24);
-        assert_eq!(indentation_length("    "), 4);
+        assert_eq!(Indentation::from_str("\t \t  \t").width(), 24);
+        assert_eq!(Indentation::from_str("\t        \t").width(), 24);
+        assert_eq!(Indentation::from_str("\t\t\t").width(), 24);
+        assert_eq!(Indentation::from_str("    ").width(), 4);
     }
 }

crates/ruff_python_formatter/src/string/mod.rs

@@ -1,26 +1,19 @@
-use std::borrow::Cow;
-use std::iter::FusedIterator;
-
 use bitflags::bitflags;
-use memchr::memchr2;
 
-use ruff_formatter::{format_args, write};
-use ruff_python_ast::{
-    self as ast, Expr, ExprBytesLiteral, ExprFString, ExprStringLiteral, ExpressionRef,
-};
-use ruff_python_ast::{AnyNodeRef, StringLiteral};
+pub(crate) use any::AnyString;
+pub(crate) use normalize::{normalize_string, NormalizedString, StringNormalizer};
+use ruff_formatter::format_args;
 use ruff_source_file::Locator;
-use ruff_text_size::{Ranged, TextLen, TextRange, TextSize};
+use ruff_text_size::{TextLen, TextRange, TextSize};
 
 use crate::comments::{leading_comments, trailing_comments};
-use crate::expression::expr_f_string::f_string_quoting;
 use crate::expression::parentheses::in_parentheses_only_soft_line_break_or_space;
-use crate::other::f_string::FormatFString;
-use crate::other::string_literal::{FormatStringLiteral, StringLiteralKind};
 use crate::prelude::*;
 use crate::QuoteStyle;
 
+mod any;
 pub(crate) mod docstring;
+mod normalize;
 
 #[derive(Copy, Clone, Debug, Default)]
 pub(crate) enum Quoting {
@@ -29,202 +22,6 @@ pub(crate) enum Quoting {
     Preserve,
 }
 
-/// Represents any kind of string expression. This could be either a string,
-/// bytes or f-string.
-#[derive(Copy, Clone, Debug)]
-pub(crate) enum AnyString<'a> {
-    String(&'a ExprStringLiteral),
-    Bytes(&'a ExprBytesLiteral),
-    FString(&'a ExprFString),
-}
-
-impl<'a> AnyString<'a> {
-    /// Creates a new [`AnyString`] from the given [`Expr`].
-    ///
-    /// Returns `None` if the expression is not either a string, bytes or f-string.
-    pub(crate) fn from_expression(expression: &'a Expr) -> Option<AnyString<'a>> {
-        match expression {
-            Expr::StringLiteral(string) => Some(AnyString::String(string)),
-            Expr::BytesLiteral(bytes) => Some(AnyString::Bytes(bytes)),
-            Expr::FString(fstring) => Some(AnyString::FString(fstring)),
-            _ => None,
-        }
-    }
-
-    /// Returns `true` if the string is implicitly concatenated.
-    pub(crate) fn is_implicit_concatenated(self) -> bool {
-        match self {
-            Self::String(ExprStringLiteral { value, .. }) => value.is_implicit_concatenated(),
-            Self::Bytes(ExprBytesLiteral { value, .. }) => value.is_implicit_concatenated(),
-            Self::FString(ExprFString { value, .. }) => value.is_implicit_concatenated(),
-        }
-    }
-
-    /// Returns the quoting to be used for this string.
-    fn quoting(self, locator: &Locator<'_>) -> Quoting {
-        match self {
-            Self::String(_) | Self::Bytes(_) => Quoting::CanChange,
-            Self::FString(f_string) => f_string_quoting(f_string, locator),
-        }
-    }
-
-    /// Returns a vector of all the [`AnyStringPart`] of this string.
-    fn parts(self, quoting: Quoting) -> AnyStringPartsIter<'a> {
-        match self {
-            Self::String(ExprStringLiteral { value, .. }) => {
-                AnyStringPartsIter::String(value.iter())
-            }
-            Self::Bytes(ExprBytesLiteral { value, .. }) => AnyStringPartsIter::Bytes(value.iter()),
-            Self::FString(ExprFString { value, .. }) => {
-                AnyStringPartsIter::FString(value.iter(), quoting)
-            }
-        }
-    }
-
-    pub(crate) fn is_multiline(self, source: &str) -> bool {
-        match self {
-            AnyString::String(_) | AnyString::Bytes(_) => {
-                let contents = &source[self.range()];
-                let prefix = StringPrefix::parse(contents);
-                let quotes = StringQuotes::parse(
-                    &contents[TextRange::new(prefix.text_len(), contents.text_len())],
-                );
-
-                quotes.is_some_and(StringQuotes::is_triple)
-                    && memchr2(b'\n', b'\r', contents.as_bytes()).is_some()
-            }
-            AnyString::FString(fstring) => {
-                memchr2(b'\n', b'\r', source[fstring.range].as_bytes()).is_some()
-            }
-        }
-    }
-}
-
-impl Ranged for AnyString<'_> {
-    fn range(&self) -> TextRange {
-        match self {
-            Self::String(expr) => expr.range(),
-            Self::Bytes(expr) => expr.range(),
-            Self::FString(expr) => expr.range(),
-        }
-    }
-}
-
-impl<'a> From<&AnyString<'a>> for AnyNodeRef<'a> {
-    fn from(value: &AnyString<'a>) -> Self {
-        match value {
-            AnyString::String(expr) => AnyNodeRef::ExprStringLiteral(expr),
-            AnyString::Bytes(expr) => AnyNodeRef::ExprBytesLiteral(expr),
-            AnyString::FString(expr) => AnyNodeRef::ExprFString(expr),
-        }
-    }
-}
-
-impl<'a> From<AnyString<'a>> for AnyNodeRef<'a> {
-    fn from(value: AnyString<'a>) -> Self {
-        AnyNodeRef::from(&value)
-    }
-}
-
-impl<'a> From<&AnyString<'a>> for ExpressionRef<'a> {
-    fn from(value: &AnyString<'a>) -> Self {
-        match value {
-            AnyString::String(expr) => ExpressionRef::StringLiteral(expr),
-            AnyString::Bytes(expr) => ExpressionRef::BytesLiteral(expr),
-            AnyString::FString(expr) => ExpressionRef::FString(expr),
-        }
-    }
-}
-
-enum AnyStringPartsIter<'a> {
-    String(std::slice::Iter<'a, StringLiteral>),
-    Bytes(std::slice::Iter<'a, ast::BytesLiteral>),
-    FString(std::slice::Iter<'a, ast::FStringPart>, Quoting),
-}
-
-impl<'a> Iterator for AnyStringPartsIter<'a> {
-    type Item = AnyStringPart<'a>;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        let part = match self {
-            Self::String(inner) => {
-                let part = inner.next()?;
-                AnyStringPart::String {
-                    part,
-                    layout: StringLiteralKind::String,
-                }
-            }
-            Self::Bytes(inner) => AnyStringPart::Bytes(inner.next()?),
-            Self::FString(inner, quoting) => {
-                let part = inner.next()?;
-                match part {
-                    ast::FStringPart::Literal(string_literal) => AnyStringPart::String {
-                        part: string_literal,
-                        layout: StringLiteralKind::InImplicitlyConcatenatedFString(*quoting),
-                    },
-                    ast::FStringPart::FString(f_string) => AnyStringPart::FString {
-                        part: f_string,
-                        quoting: *quoting,
-                    },
-                }
-            }
-        };
-
-        Some(part)
-    }
-}
-
-impl FusedIterator for AnyStringPartsIter<'_> {}
-
-/// Represents any kind of string which is part of an implicitly concatenated
-/// string. This could be either a string, bytes or f-string.
-///
-/// This is constructed from the [`AnyString::parts`] method on [`AnyString`].
-#[derive(Clone, Debug)]
-enum AnyStringPart<'a> {
-    String {
-        part: &'a ast::StringLiteral,
-        layout: StringLiteralKind,
-    },
-    Bytes(&'a ast::BytesLiteral),
-    FString {
-        part: &'a ast::FString,
-        quoting: Quoting,
-    },
-}
-
-impl<'a> From<&AnyStringPart<'a>> for AnyNodeRef<'a> {
-    fn from(value: &AnyStringPart<'a>) -> Self {
-        match value {
-            AnyStringPart::String { part, .. } => AnyNodeRef::StringLiteral(part),
-            AnyStringPart::Bytes(part) => AnyNodeRef::BytesLiteral(part),
-            AnyStringPart::FString { part, .. } => AnyNodeRef::FString(part),
-        }
-    }
-}
-
-impl Ranged for AnyStringPart<'_> {
-    fn range(&self) -> TextRange {
-        match self {
-            Self::String { part, .. } => part.range(),
-            Self::Bytes(part) => part.range(),
-            Self::FString { part, .. } => part.range(),
-        }
-    }
-}
-
-impl Format<PyFormatContext<'_>> for AnyStringPart<'_> {
-    fn fmt(&self, f: &mut PyFormatter) -> FormatResult<()> {
-        match self {
-            AnyStringPart::String { part, layout } => {
-                FormatStringLiteral::new(part, *layout).fmt(f)
-            }
-            AnyStringPart::Bytes(bytes_literal) => bytes_literal.format().fmt(f),
-            AnyStringPart::FString { part, quoting } => FormatFString::new(part, *quoting).fmt(f),
-        }
-    }
-}
-
 /// Formats any implicitly concatenated string. This could be any valid combination
 /// of string, bytes or f-string literals.
 pub(crate) struct FormatStringContinuation<'a> {
@@ -291,139 +88,22 @@ impl StringPart {
         }
     }
 
-    /// Computes the strings preferred quotes and normalizes its content.
-    ///
-    /// The parent docstring quote style should be set when formatting a code
-    /// snippet within the docstring. The quote style should correspond to the
-    /// style of quotes used by said docstring. Normalization will ensure the
-    /// quoting styles don't conflict.
-    pub(crate) fn normalize<'a>(
-        self,
-        quoting: Quoting,
-        locator: &'a Locator,
-        configured_style: QuoteStyle,
-        parent_docstring_quote_char: Option<QuoteChar>,
-        normalize_hex: bool,
-    ) -> NormalizedString<'a> {
-        // Per PEP 8, always prefer double quotes for triple-quoted strings.
-        let preferred_style = if self.quotes.triple {
-            // ... unless we're formatting a code snippet inside a docstring,
-            // then we specifically want to invert our quote style to avoid
-            // writing out invalid Python.
-            //
-            // It's worth pointing out that we can actually wind up being
-            // somewhat out of sync with PEP8 in this case. Consider this
-            // example:
-            //
-            //     def foo():
-            //         '''
-            //         Something.
-            //
-            //         >>> """tricksy"""
-            //         '''
-            //         pass
-            //
-            // Ideally, this would be reformatted as:
-            //
-            //     def foo():
-            //         """
-            //         Something.
-            //
-            //         >>> '''tricksy'''
-            //         """
-            //         pass
-            //
-            // But the logic here results in the original quoting being
-            // preserved. This is because the quoting style of the outer
-            // docstring is determined, in part, by looking at its contents. In
-            // this case, it notices that it contains a `"""` and thus infers
-            // that using `'''` would overall read better because it avoids
-            // the need to escape the interior `"""`. Except... in this case,
-            // the `"""` is actually part of a code snippet that could get
-            // reformatted to using a different quoting style itself.
-            //
-            // Fixing this would, I believe, require some fairly seismic
-            // changes to how formatting strings works. Namely, we would need
-            // to look for code snippets before normalizing the docstring, and
-            // then figure out the quoting style more holistically by looking
-            // at the various kinds of quotes used in the code snippets and
-            // what reformatting them might look like.
-            //
-            // Overall this is a bit of a corner case and just inverting the
-            // style from what the parent ultimately decided upon works, even
-            // if it doesn't have perfect alignment with PEP8.
-            if let Some(quote) = parent_docstring_quote_char {
-                QuoteStyle::from(quote.invert())
-            } else {
-                QuoteStyle::Double
-            }
-        } else {
-            configured_style
-        };
-
-        let raw_content = &locator.slice(self.content_range);
-
-        let quotes = match quoting {
-            Quoting::Preserve => self.quotes,
-            Quoting::CanChange => {
-                if let Some(preferred_quote) = QuoteChar::from_style(preferred_style) {
-                    if self.prefix.is_raw_string() {
-                        choose_quotes_raw(raw_content, self.quotes, preferred_quote)
-                    } else {
-                        choose_quotes(raw_content, self.quotes, preferred_quote)
-                    }
-                } else {
-                    self.quotes
-                }
-            }
-        };
-
-        let normalized = normalize_string(raw_content, quotes, self.prefix, normalize_hex);
-
-        NormalizedString {
-            prefix: self.prefix,
-            content_range: self.content_range,
-            text: normalized,
-            quotes,
-        }
+    /// Returns the prefix of the string part.
+    pub(crate) const fn prefix(&self) -> StringPrefix {
+        self.prefix
     }
-}
 
-#[derive(Debug)]
-pub(crate) struct NormalizedString<'a> {
-    prefix: StringPrefix,
+    /// Returns the surrounding quotes of the string part.
+    pub(crate) const fn quotes(&self) -> StringQuotes {
+        self.quotes
+    }
 
-    /// The quotes of the normalized string (preferred quotes)
-    quotes: StringQuotes,
-
-    /// The range of the string's content in the source (minus prefix and quotes).
-    content_range: TextRange,
-
-    /// The normalized text
-    text: Cow<'a, str>,
-}
-
-impl Ranged for NormalizedString<'_> {
-    fn range(&self) -> TextRange {
+    /// Returns the range of the string's content in the source (minus prefix and quotes).
+    pub(crate) const fn content_range(&self) -> TextRange {
         self.content_range
     }
 }
 
-impl Format<PyFormatContext<'_>> for NormalizedString<'_> {
-    fn fmt(&self, f: &mut Formatter<PyFormatContext<'_>>) -> FormatResult<()> {
-        write!(f, [self.prefix, self.quotes])?;
-        match &self.text {
-            Cow::Borrowed(_) => {
-                source_text_slice(self.range()).fmt(f)?;
-            }
-            Cow::Owned(normalized) => {
-                text(normalized).fmt(f)?;
-            }
-        }
-        self.quotes.fmt(f)
-    }
-}
-
 bitflags! {
     #[derive(Copy, Clone, Debug, PartialEq, Eq)]
     pub(crate) struct StringPrefix: u8 {
@@ -504,171 +184,6 @@ impl Format<PyFormatContext<'_>> for StringPrefix {
     }
 }
 
-/// Choose the appropriate quote style for a raw string.
-///
-/// The preferred quote style is chosen unless the string contains unescaped quotes of the
-/// preferred style. For example, `r"foo"` is chosen over `r'foo'` if the preferred quote
-/// style is double quotes.
-fn choose_quotes_raw(
-    input: &str,
-    quotes: StringQuotes,
-    preferred_quote: QuoteChar,
-) -> StringQuotes {
-    let preferred_quote_char = preferred_quote.as_char();
-    let mut chars = input.chars().peekable();
-    let contains_unescaped_configured_quotes = loop {
-        match chars.next() {
-            Some('\\') => {
-                // Ignore escaped characters
-                chars.next();
-            }
-            // `"` or `'`
-            Some(c) if c == preferred_quote_char => {
-                if !quotes.triple {
-                    break true;
-                }
-
-                match chars.peek() {
-                    // We can't turn `r'''\""'''` into `r"""\"""""`, this would confuse the parser
-                    // about where the closing triple quotes start
-                    None => break true,
-                    Some(next) if *next == preferred_quote_char => {
-                        // `""` or `''`
-                        chars.next();
-
-                        // We can't turn `r'''""'''` into `r""""""""`, nor can we have
-                        // `"""` or `'''` respectively inside the string
-                        if chars.peek().is_none() || chars.peek() == Some(&preferred_quote_char) {
-                            break true;
-                        }
-                    }
-                    _ => {}
-                }
-            }
-            Some(_) => continue,
-            None => break false,
-        }
-    };
-
-    StringQuotes {
-        triple: quotes.triple,
-        quote_char: if contains_unescaped_configured_quotes {
-            quotes.quote_char
-        } else {
-            preferred_quote
-        },
-    }
-}
-
-/// Choose the appropriate quote style for a string.
-///
-/// For single quoted strings, the preferred quote style is used, unless the alternative quote style
-/// would require fewer escapes.
-///
-/// For triple quoted strings, the preferred quote style is always used, unless the string contains
-/// a triplet of the quote character (e.g., if double quotes are preferred, double quotes will be
-/// used unless the string contains `"""`).
-fn choose_quotes(input: &str, quotes: StringQuotes, preferred_quote: QuoteChar) -> StringQuotes {
-    let quote = if quotes.triple {
-        // True if the string contains a triple quote sequence of the configured quote style.
-        let mut uses_triple_quotes = false;
-        let mut chars = input.chars().peekable();
-
-        while let Some(c) = chars.next() {
-            let preferred_quote_char = preferred_quote.as_char();
-            match c {
-                '\\' => {
-                    if matches!(chars.peek(), Some('"' | '\\')) {
-                        chars.next();
-                    }
-                }
-                // `"` or `'`
-                c if c == preferred_quote_char => {
-                    match chars.peek().copied() {
-                        Some(c) if c == preferred_quote_char => {
-                            // `""` or `''`
-                            chars.next();
-
-                            match chars.peek().copied() {
-                                Some(c) if c == preferred_quote_char => {
-                                    // `"""` or `'''`
-                                    chars.next();
-                                    uses_triple_quotes = true;
-                                    break;
-                                }
-                                Some(_) => {}
-                                None => {
-                                    // Handle `''' ""'''`. At this point we have consumed both
-                                    // double quotes, so on the next iteration the iterator is empty
-                                    // and we'd miss the string ending with a preferred quote
-                                    uses_triple_quotes = true;
-                                    break;
-                                }
-                            }
-                        }
-                        Some(_) => {
-                            // A single quote char, this is ok
-                        }
-                        None => {
-                            // Trailing quote at the end of the comment
-                            uses_triple_quotes = true;
-                            break;
-                        }
-                    }
-                }
-                _ => continue,
-            }
-        }
-
-        if uses_triple_quotes {
-            // String contains a triple quote sequence of the configured quote style.
-            // Keep the existing quote style.
-            quotes.quote_char
-        } else {
-            preferred_quote
-        }
-    } else {
-        let mut single_quotes = 0u32;
-        let mut double_quotes = 0u32;
-
-        for c in input.chars() {
-            match c {
-                '\'' => {
-                    single_quotes += 1;
-                }
-
-                '"' => {
-                    double_quotes += 1;
-                }
-
-                _ => continue,
-            }
-        }
-
-        match preferred_quote {
-            QuoteChar::Single => {
-                if single_quotes > double_quotes {
-                    QuoteChar::Double
-                } else {
-                    QuoteChar::Single
-                }
-            }
-            QuoteChar::Double => {
-                if double_quotes > single_quotes {
-                    QuoteChar::Single
-                } else {
-                    QuoteChar::Double
-                }
-            }
-        }
-    };
-
-    StringQuotes {
-        triple: quotes.triple,
-        quote_char: quote,
-    }
-}
-
 #[derive(Copy, Clone, Debug)]
 pub(crate) struct StringQuotes {
     triple: bool,
@@ -772,269 +287,3 @@ impl TryFrom<char> for QuoteChar {
         }
     }
 }
-
-/// Adds the necessary quote escapes and removes unnecessary escape sequences when quoting `input`
-/// with the provided [`StringQuotes`] style.
-///
-/// Returns the normalized string and whether it contains new lines.
-fn normalize_string(
-    input: &str,
-    quotes: StringQuotes,
-    prefix: StringPrefix,
-    normalize_hex: bool,
-) -> Cow<str> {
-    // The normalized string if `input` is not yet normalized.
-    // `output` must remain empty if `input` is already normalized.
-    let mut output = String::new();
-    // Tracks the last index of `input` that has been written to `output`.
-    // If `last_index` is `0` at the end, then the input is already normalized and can be returned as is.
-    let mut last_index = 0;
-
-    let quote = quotes.quote_char;
-    let preferred_quote = quote.as_char();
-    let opposite_quote = quote.invert().as_char();
-
-    let mut chars = input.char_indices().peekable();
-
-    let is_raw = prefix.is_raw_string();
-    let is_fstring = prefix.is_fstring();
-    let mut formatted_value_nesting = 0u32;
-
-    while let Some((index, c)) = chars.next() {
-        if is_fstring && matches!(c, '{' | '}') {
-            if chars.peek().copied().is_some_and(|(_, next)| next == c) {
-                // Skip over the second character of the double braces
-                chars.next();
-            } else if c == '{' {
-                formatted_value_nesting += 1;
-            } else {
-                // Safe to assume that `c == '}'` here because of the matched pattern above
-                formatted_value_nesting = formatted_value_nesting.saturating_sub(1);
-            }
-            continue;
-        }
-        if c == '\r' {
-            output.push_str(&input[last_index..index]);
-
-            // Skip over the '\r' character, keep the `\n`
-            if chars.peek().copied().is_some_and(|(_, next)| next == '\n') {
-                chars.next();
-            }
-            // Replace the `\r` with a `\n`
-            else {
-                output.push('\n');
-            }
-
-            last_index = index + '\r'.len_utf8();
-        } else if !is_raw {
-            if c == '\\' {
-                if let Some((_, next)) = chars.clone().next() {
-                    if next == '\\' {
-                        // Skip over escaped backslashes
-                        chars.next();
-                    } else if normalize_hex {
-                        if let Some(normalised) = UnicodeEscape::new(next, !prefix.is_byte())
-                            .and_then(|escape| {
-                                escape.normalize(&input[index + c.len_utf8() + next.len_utf8()..])
-                            })
-                        {
-                            // Length of the `\` plus the length of the escape sequence character (`u` | `U` | `x`)
-                            let escape_start_len = '\\'.len_utf8() + next.len_utf8();
-                            let escape_start_offset = index + escape_start_len;
-                            if let Cow::Owned(normalised) = &normalised {
-                                output.push_str(&input[last_index..escape_start_offset]);
-                                output.push_str(normalised);
-                                last_index = escape_start_offset + normalised.len();
-                            };
-
-                            // Move the `chars` iterator passed the escape sequence.
-                            // Simply reassigning `chars` doesn't work because the indices` would
-                            // then be off.
-                            for _ in 0..next.len_utf8() + normalised.len() {
-                                chars.next();
-                            }
-                        }
-                    }
-
-                    if !quotes.triple {
-                        #[allow(clippy::if_same_then_else)]
-                        if next == opposite_quote && formatted_value_nesting == 0 {
-                            // Remove the escape by ending before the backslash and starting again with the quote
-                            chars.next();
-                            output.push_str(&input[last_index..index]);
-                            last_index = index + '\\'.len_utf8();
-                        } else if next == preferred_quote {
-                            // Quote is already escaped, skip over it.
-                            chars.next();
-                        }
-                    }
-                }
-            } else if !quotes.triple && c == preferred_quote && formatted_value_nesting == 0 {
-                // Escape the quote
-                output.push_str(&input[last_index..index]);
-                output.push('\\');
-                output.push(c);
-                last_index = index + preferred_quote.len_utf8();
-            }
-        }
-    }
-
-    let normalized = if last_index == 0 {
-        Cow::Borrowed(input)
-    } else {
-        output.push_str(&input[last_index..]);
-        Cow::Owned(output)
-    };
-
-    normalized
-}
-
-#[derive(Copy, Clone, Debug, PartialEq, Eq)]
-enum UnicodeEscape {
-    /// A hex escape sequence of either 2 (`\x`), 4 (`\u`) or 8 (`\U`) hex characters.
-    Hex(usize),
-
-    /// An escaped unicode name (`\N{name}`)
-    CharacterName,
-}
-
-impl UnicodeEscape {
-    fn new(first: char, allow_unicode: bool) -> Option<UnicodeEscape> {
-        Some(match first {
-            'x' => UnicodeEscape::Hex(2),
-            'u' if allow_unicode => UnicodeEscape::Hex(4),
-            'U' if allow_unicode => UnicodeEscape::Hex(8),
-            'N' if allow_unicode => UnicodeEscape::CharacterName,
-            _ => return None,
-        })
-    }
-
-    /// Normalises `\u..`, `\U..`, `\x..` and `\N{..}` escape sequences to:
-    ///
-    /// * `\u`, `\U'` and `\x`: To use lower case for the characters `a-f`.
-    /// * `\N`: To use uppercase letters
-    fn normalize(self, input: &str) -> Option<Cow<str>> {
-        let mut normalised = String::new();
-
-        let len = match self {
-            UnicodeEscape::Hex(len) => {
-                // It's not a valid escape sequence if the input string has fewer characters
-                // left than required by the escape sequence.
-                if input.len() < len {
-                    return None;
-                }
-
-                for (index, c) in input.char_indices().take(len) {
-                    match c {
-                        '0'..='9' | 'a'..='f' => {
-                            if !normalised.is_empty() {
-                                normalised.push(c);
-                            }
-                        }
-                        'A'..='F' => {
-                            if normalised.is_empty() {
-                                normalised.reserve(len);
-                                normalised.push_str(&input[..index]);
-                                normalised.push(c.to_ascii_lowercase());
-                            } else {
-                                normalised.push(c.to_ascii_lowercase());
-                            }
-                        }
-                        _ => {
-                            // not a valid escape sequence
-                            return None;
-                        }
-                    }
-                }
-
-                len
-            }
-            UnicodeEscape::CharacterName => {
-                let mut char_indices = input.char_indices();
-
-                if !matches!(char_indices.next(), Some((_, '{'))) {
-                    return None;
-                }
-
-                loop {
-                    if let Some((index, c)) = char_indices.next() {
-                        match c {
-                            '}' => {
-                                if !normalised.is_empty() {
-                                    normalised.push('}');
-                                }
-
-                                // Name must be at least two characters long.
-                                if index < 3 {
-                                    return None;
-                                }
-
-                                break index + '}'.len_utf8();
-                            }
-                            '0'..='9' | 'A'..='Z' | ' ' | '-' => {
-                                if !normalised.is_empty() {
-                                    normalised.push(c);
-                                }
-                            }
-                            'a'..='z' => {
-                                if normalised.is_empty() {
-                                    normalised.reserve(c.len_utf8() + '}'.len_utf8());
-                                    normalised.push_str(&input[..index]);
-                                    normalised.push(c.to_ascii_uppercase());
-                                } else {
-                                    normalised.push(c.to_ascii_uppercase());
-                                }
-                            }
-                            _ => {
-                                // Seems like an invalid escape sequence, don't normalise it.
-                                return None;
-                            }
-                        }
-                    } else {
-                        // Unterminated escape sequence, don't normalise it.
-                        return None;
-                    }
-                }
-            }
-        };
-
-        Some(if normalised.is_empty() {
-            Cow::Borrowed(&input[..len])
-        } else {
-            Cow::Owned(normalised)
-        })
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use crate::string::{normalize_string, QuoteChar, StringPrefix, StringQuotes, UnicodeEscape};
-    use std::borrow::Cow;
-
-    #[test]
-    fn normalize_32_escape() {
-        let escape_sequence = UnicodeEscape::new('U', true).unwrap();
-
-        assert_eq!(
-            Some(Cow::Owned("0001f60e".to_string())),
-            escape_sequence.normalize("0001F60E")
-        );
-    }
-
-    #[test]
-    fn normalize_hex_in_byte_string() {
-        let input = r"\x89\x50\x4E\x47\x0D\x0A\x1A\x0A";
-
-        let normalized = normalize_string(
-            input,
-            StringQuotes {
-                triple: false,
-                quote_char: QuoteChar::Double,
-            },
-            StringPrefix::BYTE,
-            true,
-        );
-
-        assert_eq!(r"\x89\x50\x4e\x47\x0d\x0a\x1a\x0a", &normalized);
-    }
-}

crates/ruff_python_formatter/src/string/normalize.rs

@@ -0,0 +1,655 @@
+use std::borrow::Cow;
+
+use ruff_formatter::FormatContext;
+use ruff_source_file::Locator;
+use ruff_text_size::{Ranged, TextRange};
+
+use crate::context::FStringState;
+use crate::options::PythonVersion;
+use crate::prelude::*;
+use crate::preview::is_hex_codes_in_unicode_sequences_enabled;
+use crate::string::{QuoteChar, Quoting, StringPart, StringPrefix, StringQuotes};
+use crate::QuoteStyle;
+
+pub(crate) struct StringNormalizer {
+    quoting: Quoting,
+    preferred_quote_style: QuoteStyle,
+    parent_docstring_quote_char: Option<QuoteChar>,
+    f_string_state: FStringState,
+    target_version: PythonVersion,
+    normalize_hex: bool,
+}
+
+impl StringNormalizer {
+    pub(crate) fn from_context(context: &PyFormatContext<'_>) -> Self {
+        Self {
+            quoting: Quoting::default(),
+            preferred_quote_style: QuoteStyle::default(),
+            parent_docstring_quote_char: context.docstring(),
+            f_string_state: context.f_string_state(),
+            target_version: context.options().target_version(),
+            normalize_hex: is_hex_codes_in_unicode_sequences_enabled(context),
+        }
+    }
+
+    pub(crate) fn with_preferred_quote_style(mut self, quote_style: QuoteStyle) -> Self {
+        self.preferred_quote_style = quote_style;
+        self
+    }
+
+    pub(crate) fn with_quoting(mut self, quoting: Quoting) -> Self {
+        self.quoting = quoting;
+        self
+    }
+
+    /// Computes the strings preferred quotes.
+    pub(crate) fn choose_quotes(&self, string: &StringPart, locator: &Locator) -> StringQuotes {
+        // Per PEP 8, always prefer double quotes for triple-quoted strings.
+        // Except when using quote-style-preserve.
+        let preferred_style = if string.quotes().triple {
+            // ... unless we're formatting a code snippet inside a docstring,
+            // then we specifically want to invert our quote style to avoid
+            // writing out invalid Python.
+            //
+            // It's worth pointing out that we can actually wind up being
+            // somewhat out of sync with PEP8 in this case. Consider this
+            // example:
+            //
+            //     def foo():
+            //         '''
+            //         Something.
+            //
+            //         >>> """tricksy"""
+            //         '''
+            //         pass
+            //
+            // Ideally, this would be reformatted as:
+            //
+            //     def foo():
+            //         """
+            //         Something.
+            //
+            //         >>> '''tricksy'''
+            //         """
+            //         pass
+            //
+            // But the logic here results in the original quoting being
+            // preserved. This is because the quoting style of the outer
+            // docstring is determined, in part, by looking at its contents. In
+            // this case, it notices that it contains a `"""` and thus infers
+            // that using `'''` would overall read better because it avoids
+            // the need to escape the interior `"""`. Except... in this case,
+            // the `"""` is actually part of a code snippet that could get
+            // reformatted to using a different quoting style itself.
+            //
+            // Fixing this would, I believe, require some fairly seismic
+            // changes to how formatting strings works. Namely, we would need
+            // to look for code snippets before normalizing the docstring, and
+            // then figure out the quoting style more holistically by looking
+            // at the various kinds of quotes used in the code snippets and
+            // what reformatting them might look like.
+            //
+            // Overall this is a bit of a corner case and just inverting the
+            // style from what the parent ultimately decided upon works, even
+            // if it doesn't have perfect alignment with PEP8.
+            if let Some(quote) = self.parent_docstring_quote_char {
+                QuoteStyle::from(quote.invert())
+            } else if self.preferred_quote_style.is_preserve() {
+                QuoteStyle::Preserve
+            } else {
+                QuoteStyle::Double
+            }
+        } else {
+            self.preferred_quote_style
+        };
+
+        let quoting = if let FStringState::InsideExpressionElement(context) = self.f_string_state {
+            // If we're inside an f-string, we need to make sure to preserve the
+            // existing quotes unless we're inside a triple-quoted f-string and
+            // the inner string itself isn't triple-quoted. For example:
+            //
+            // ```python
+            // f"""outer {"inner"}"""  # Valid
+            // f"""outer {"""inner"""}"""  # Invalid
+            // ```
+            //
+            // Or, if the target version supports PEP 701.
+            //
+            // The reason to preserve the quotes is based on the assumption that
+            // the original f-string is valid in terms of quoting, and we don't
+            // want to change that to make it invalid.
+            if (context.quotes().is_triple() && !string.quotes().is_triple())
+                || self.target_version.supports_pep_701()
+            {
+                self.quoting
+            } else {
+                Quoting::Preserve
+            }
+        } else {
+            self.quoting
+        };
+
+        match quoting {
+            Quoting::Preserve => string.quotes(),
+            Quoting::CanChange => {
+                if let Some(preferred_quote) = QuoteChar::from_style(preferred_style) {
+                    let raw_content = locator.slice(string.content_range());
+                    if string.prefix().is_raw_string() {
+                        choose_quotes_for_raw_string(raw_content, string.quotes(), preferred_quote)
+                    } else {
+                        choose_quotes_impl(raw_content, string.quotes(), preferred_quote)
+                    }
+                } else {
+                    string.quotes()
+                }
+            }
+        }
+    }
+
+    /// Computes the strings preferred quotes and normalizes its content.
+    pub(crate) fn normalize<'a>(
+        &self,
+        string: &StringPart,
+        locator: &'a Locator,
+    ) -> NormalizedString<'a> {
+        let raw_content = locator.slice(string.content_range());
+
+        let quotes = self.choose_quotes(string, locator);
+
+        let normalized = normalize_string(raw_content, quotes, string.prefix(), self.normalize_hex);
+
+        NormalizedString {
+            prefix: string.prefix(),
+            content_range: string.content_range(),
+            text: normalized,
+            quotes,
+        }
+    }
+}
+
+#[derive(Debug)]
+pub(crate) struct NormalizedString<'a> {
+    prefix: crate::string::StringPrefix,
+
+    /// The quotes of the normalized string (preferred quotes)
+    quotes: StringQuotes,
+
+    /// The range of the string's content in the source (minus prefix and quotes).
+    content_range: TextRange,
+
+    /// The normalized text
+    text: Cow<'a, str>,
+}
+
+impl<'a> NormalizedString<'a> {
+    pub(crate) fn text(&self) -> &Cow<'a, str> {
+        &self.text
+    }
+
+    pub(crate) fn quotes(&self) -> StringQuotes {
+        self.quotes
+    }
+
+    pub(crate) fn prefix(&self) -> StringPrefix {
+        self.prefix
+    }
+}
+
+impl Ranged for NormalizedString<'_> {
+    fn range(&self) -> TextRange {
+        self.content_range
+    }
+}
+
+impl Format<PyFormatContext<'_>> for NormalizedString<'_> {
+    fn fmt(&self, f: &mut Formatter<PyFormatContext<'_>>) -> FormatResult<()> {
+        ruff_formatter::write!(f, [self.prefix, self.quotes])?;
+        match &self.text {
+            Cow::Borrowed(_) => {
+                source_text_slice(self.range()).fmt(f)?;
+            }
+            Cow::Owned(normalized) => {
+                text(normalized).fmt(f)?;
+            }
+        }
+        self.quotes.fmt(f)
+    }
+}
+
+/// Choose the appropriate quote style for a raw string.
+///
+/// The preferred quote style is chosen unless the string contains unescaped quotes of the
+/// preferred style. For example, `r"foo"` is chosen over `r'foo'` if the preferred quote
+/// style is double quotes.
+fn choose_quotes_for_raw_string(
+    input: &str,
+    quotes: StringQuotes,
+    preferred_quote: QuoteChar,
+) -> StringQuotes {
+    let preferred_quote_char = preferred_quote.as_char();
+    let mut chars = input.chars().peekable();
+    let contains_unescaped_configured_quotes = loop {
+        match chars.next() {
+            Some('\\') => {
+                // Ignore escaped characters
+                chars.next();
+            }
+            // `"` or `'`
+            Some(c) if c == preferred_quote_char => {
+                if !quotes.triple {
+                    break true;
+                }
+
+                match chars.peek() {
+                    // We can't turn `r'''\""'''` into `r"""\"""""`, this would confuse the parser
+                    // about where the closing triple quotes start
+                    None => break true,
+                    Some(next) if *next == preferred_quote_char => {
+                        // `""` or `''`
+                        chars.next();
+
+                        // We can't turn `r'''""'''` into `r""""""""`, nor can we have
+                        // `"""` or `'''` respectively inside the string
+                        if chars.peek().is_none() || chars.peek() == Some(&preferred_quote_char) {
+                            break true;
+                        }
+                    }
+                    _ => {}
+                }
+            }
+            Some(_) => continue,
+            None => break false,
+        }
+    };
+
+    StringQuotes {
+        triple: quotes.triple,
+        quote_char: if contains_unescaped_configured_quotes {
+            quotes.quote_char
+        } else {
+            preferred_quote
+        },
+    }
+}
+
+/// Choose the appropriate quote style for a string.
+///
+/// For single quoted strings, the preferred quote style is used, unless the alternative quote style
+/// would require fewer escapes.
+///
+/// For triple quoted strings, the preferred quote style is always used, unless the string contains
+/// a triplet of the quote character (e.g., if double quotes are preferred, double quotes will be
+/// used unless the string contains `"""`).
+fn choose_quotes_impl(
+    input: &str,
+    quotes: StringQuotes,
+    preferred_quote: QuoteChar,
+) -> StringQuotes {
+    let quote = if quotes.triple {
+        // True if the string contains a triple quote sequence of the configured quote style.
+        let mut uses_triple_quotes = false;
+        let mut chars = input.chars().peekable();
+
+        while let Some(c) = chars.next() {
+            let preferred_quote_char = preferred_quote.as_char();
+            match c {
+                '\\' => {
+                    if matches!(chars.peek(), Some('"' | '\\')) {
+                        chars.next();
+                    }
+                }
+                // `"` or `'`
+                c if c == preferred_quote_char => {
+                    match chars.peek().copied() {
+                        Some(c) if c == preferred_quote_char => {
+                            // `""` or `''`
+                            chars.next();
+
+                            match chars.peek().copied() {
+                                Some(c) if c == preferred_quote_char => {
+                                    // `"""` or `'''`
+                                    chars.next();
+                                    uses_triple_quotes = true;
+                                    break;
+                                }
+                                Some(_) => {}
+                                None => {
+                                    // Handle `''' ""'''`. At this point we have consumed both
+                                    // double quotes, so on the next iteration the iterator is empty
+                                    // and we'd miss the string ending with a preferred quote
+                                    uses_triple_quotes = true;
+                                    break;
+                                }
+                            }
+                        }
+                        Some(_) => {
+                            // A single quote char, this is ok
+                        }
+                        None => {
+                            // Trailing quote at the end of the comment
+                            uses_triple_quotes = true;
+                            break;
+                        }
+                    }
+                }
+                _ => continue,
+            }
+        }
+
+        if uses_triple_quotes {
+            // String contains a triple quote sequence of the configured quote style.
+            // Keep the existing quote style.
+            quotes.quote_char
+        } else {
+            preferred_quote
+        }
+    } else {
+        let mut single_quotes = 0u32;
+        let mut double_quotes = 0u32;
+
+        for c in input.chars() {
+            match c {
+                '\'' => {
+                    single_quotes += 1;
+                }
+
+                '"' => {
+                    double_quotes += 1;
+                }
+
+                _ => continue,
+            }
+        }
+
+        match preferred_quote {
+            QuoteChar::Single => {
+                if single_quotes > double_quotes {
+                    QuoteChar::Double
+                } else {
+                    QuoteChar::Single
+                }
+            }
+            QuoteChar::Double => {
+                if double_quotes > single_quotes {
+                    QuoteChar::Single
+                } else {
+                    QuoteChar::Double
+                }
+            }
+        }
+    };
+
+    StringQuotes {
+        triple: quotes.triple,
+        quote_char: quote,
+    }
+}
+
+/// Adds the necessary quote escapes and removes unnecessary escape sequences when quoting `input`
+/// with the provided [`StringQuotes`] style.
+///
+/// Returns the normalized string and whether it contains new lines.
+pub(crate) fn normalize_string(
+    input: &str,
+    quotes: StringQuotes,
+    prefix: StringPrefix,
+    normalize_hex: bool,
+) -> Cow<str> {
+    // The normalized string if `input` is not yet normalized.
+    // `output` must remain empty if `input` is already normalized.
+    let mut output = String::new();
+    // Tracks the last index of `input` that has been written to `output`.
+    // If `last_index` is `0` at the end, then the input is already normalized and can be returned as is.
+    let mut last_index = 0;
+
+    let quote = quotes.quote_char;
+    let preferred_quote = quote.as_char();
+    let opposite_quote = quote.invert().as_char();
+
+    let mut chars = input.char_indices().peekable();
+
+    let is_raw = prefix.is_raw_string();
+    let is_fstring = prefix.is_fstring();
+    let mut formatted_value_nesting = 0u32;
+
+    while let Some((index, c)) = chars.next() {
+        if is_fstring && matches!(c, '{' | '}') {
+            if chars.peek().copied().is_some_and(|(_, next)| next == c) {
+                // Skip over the second character of the double braces
+                chars.next();
+            } else if c == '{' {
+                formatted_value_nesting += 1;
+            } else {
+                // Safe to assume that `c == '}'` here because of the matched pattern above
+                formatted_value_nesting = formatted_value_nesting.saturating_sub(1);
+            }
+            continue;
+        }
+        if c == '\r' {
+            output.push_str(&input[last_index..index]);
+
+            // Skip over the '\r' character, keep the `\n`
+            if chars.peek().copied().is_some_and(|(_, next)| next == '\n') {
+                chars.next();
+            }
+            // Replace the `\r` with a `\n`
+            else {
+                output.push('\n');
+            }
+
+            last_index = index + '\r'.len_utf8();
+        } else if !is_raw {
+            if c == '\\' {
+                if let Some((_, next)) = chars.clone().next() {
+                    if next == '\\' {
+                        // Skip over escaped backslashes
+                        chars.next();
+                    } else if normalize_hex {
+                        if let Some(normalised) = UnicodeEscape::new(next, !prefix.is_byte())
+                            .and_then(|escape| {
+                                escape.normalize(&input[index + c.len_utf8() + next.len_utf8()..])
+                            })
+                        {
+                            // Length of the `\` plus the length of the escape sequence character (`u` | `U` | `x`)
+                            let escape_start_len = '\\'.len_utf8() + next.len_utf8();
+                            let escape_start_offset = index + escape_start_len;
+                            if let Cow::Owned(normalised) = &normalised {
+                                output.push_str(&input[last_index..escape_start_offset]);
+                                output.push_str(normalised);
+                                last_index = escape_start_offset + normalised.len();
+                            };
+
+                            // Move the `chars` iterator passed the escape sequence.
+                            // Simply reassigning `chars` doesn't work because the indices` would
+                            // then be off.
+                            for _ in 0..next.len_utf8() + normalised.len() {
+                                chars.next();
+                            }
+                        }
+                    }
+
+                    if !quotes.triple {
+                        #[allow(clippy::if_same_then_else)]
+                        if next == opposite_quote && formatted_value_nesting == 0 {
+                            // Remove the escape by ending before the backslash and starting again with the quote
+                            chars.next();
+                            output.push_str(&input[last_index..index]);
+                            last_index = index + '\\'.len_utf8();
+                        } else if next == preferred_quote {
+                            // Quote is already escaped, skip over it.
+                            chars.next();
+                        }
+                    }
+                }
+            } else if !quotes.triple && c == preferred_quote && formatted_value_nesting == 0 {
+                // Escape the quote
+                output.push_str(&input[last_index..index]);
+                output.push('\\');
+                output.push(c);
+                last_index = index + preferred_quote.len_utf8();
+            }
+        }
+    }
+
+    let normalized = if last_index == 0 {
+        Cow::Borrowed(input)
+    } else {
+        output.push_str(&input[last_index..]);
+        Cow::Owned(output)
+    };
+
+    normalized
+}
+
+#[derive(Copy, Clone, Debug, PartialEq, Eq)]
+enum UnicodeEscape {
+    /// A hex escape sequence of either 2 (`\x`), 4 (`\u`) or 8 (`\U`) hex characters.
+    Hex(usize),
+
+    /// An escaped unicode name (`\N{name}`)
+    CharacterName,
+}
+
+impl UnicodeEscape {
+    fn new(first: char, allow_unicode: bool) -> Option<UnicodeEscape> {
+        Some(match first {
+            'x' => UnicodeEscape::Hex(2),
+            'u' if allow_unicode => UnicodeEscape::Hex(4),
+            'U' if allow_unicode => UnicodeEscape::Hex(8),
+            'N' if allow_unicode => UnicodeEscape::CharacterName,
+            _ => return None,
+        })
+    }
+
+    /// Normalises `\u..`, `\U..`, `\x..` and `\N{..}` escape sequences to:
+    ///
+    /// * `\u`, `\U'` and `\x`: To use lower case for the characters `a-f`.
+    /// * `\N`: To use uppercase letters
+    fn normalize(self, input: &str) -> Option<Cow<str>> {
+        let mut normalised = String::new();
+
+        let len = match self {
+            UnicodeEscape::Hex(len) => {
+                // It's not a valid escape sequence if the input string has fewer characters
+                // left than required by the escape sequence.
+                if input.len() < len {
+                    return None;
+                }
+
+                for (index, c) in input.char_indices().take(len) {
+                    match c {
+                        '0'..='9' | 'a'..='f' => {
+                            if !normalised.is_empty() {
+                                normalised.push(c);
+                            }
+                        }
+                        'A'..='F' => {
+                            if normalised.is_empty() {
+                                normalised.reserve(len);
+                                normalised.push_str(&input[..index]);
+                                normalised.push(c.to_ascii_lowercase());
+                            } else {
+                                normalised.push(c.to_ascii_lowercase());
+                            }
+                        }
+                        _ => {
+                            // not a valid escape sequence
+                            return None;
+                        }
+                    }
+                }
+
+                len
+            }
+            UnicodeEscape::CharacterName => {
+                let mut char_indices = input.char_indices();
+
+                if !matches!(char_indices.next(), Some((_, '{'))) {
+                    return None;
+                }
+
+                loop {
+                    if let Some((index, c)) = char_indices.next() {
+                        match c {
+                            '}' => {
+                                if !normalised.is_empty() {
+                                    normalised.push('}');
+                                }
+
+                                // Name must be at least two characters long.
+                                if index < 3 {
+                                    return None;
+                                }
+
+                                break index + '}'.len_utf8();
+                            }
+                            '0'..='9' | 'A'..='Z' | ' ' | '-' => {
+                                if !normalised.is_empty() {
+                                    normalised.push(c);
+                                }
+                            }
+                            'a'..='z' => {
+                                if normalised.is_empty() {
+                                    normalised.reserve(c.len_utf8() + '}'.len_utf8());
+                                    normalised.push_str(&input[..index]);
+                                    normalised.push(c.to_ascii_uppercase());
+                                } else {
+                                    normalised.push(c.to_ascii_uppercase());
+                                }
+                            }
+                            _ => {
+                                // Seems like an invalid escape sequence, don't normalise it.
+                                return None;
+                            }
+                        }
+                    } else {
+                        // Unterminated escape sequence, don't normalise it.
+                        return None;
+                    }
+                }
+            }
+        };
+
+        Some(if normalised.is_empty() {
+            Cow::Borrowed(&input[..len])
+        } else {
+            Cow::Owned(normalised)
+        })
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::borrow::Cow;
+
+    use crate::string::{QuoteChar, StringPrefix, StringQuotes};
+
+    use super::{normalize_string, UnicodeEscape};
+
+    #[test]
+    fn normalize_32_escape() {
+        let escape_sequence = UnicodeEscape::new('U', true).unwrap();
+
+        assert_eq!(
+            Some(Cow::Owned("0001f60e".to_string())),
+            escape_sequence.normalize("0001F60E")
+        );
+    }
+
+    #[test]
+    fn normalize_hex_in_byte_string() {
+        let input = r"\x89\x50\x4E\x47\x0D\x0A\x1A\x0A";
+
+        let normalized = normalize_string(
+            input,
+            StringQuotes {
+                triple: false,
+                quote_char: QuoteChar::Double,
+            },
+            StringPrefix::BYTE,
+            true,
+        );
+
+        assert_eq!(r"\x89\x50\x4e\x47\x0d\x0a\x1a\x0a", &normalized);
+    }
+}

crates/ruff_python_formatter/src/verbatim.rs

@@ -873,11 +873,11 @@ impl Ranged for LogicalLine {
     }
 }
 
-struct VerbatimText {
+pub(crate) struct VerbatimText {
     verbatim_range: TextRange,
 }
 
-fn verbatim_text<T>(item: T) -> VerbatimText
+pub(crate) fn verbatim_text<T>(item: T) -> VerbatimText
 where
     T: Ranged,
 {

crates/ruff_python_formatter/tests/snapshots/black_compatibility@cases__preview_long_strings.py.snap

@@ -902,7 +902,7 @@ log.info(f"""Skipping: {'a' == 'b'} {desc['ms_name']} {money=} {dte=} {pos_share
  )
  
  dict_with_lambda_values = {
-@@ -524,61 +383,54 @@
+@@ -524,65 +383,58 @@
  
  # Complex string concatenations with a method call in the middle.
  code = (
@@ -941,7 +941,7 @@ log.info(f"""Skipping: {'a' == 'b'} {desc['ms_name']} {money=} {dte=} {pos_share
  log.info(
 -    "Skipping:"
 -    f" {desc['db_id']} {foo('bar',x=123)} {'foo' != 'bar'} {(x := 'abc=')} {pos_share=} {desc['status']} {desc['exposure_max']}"
-+    f'Skipping: {desc["db_id"]} {foo("bar",x=123)} {"foo" != "bar"} {(x := "abc=")} {pos_share=} {desc["status"]} {desc["exposure_max"]}'
++    f'Skipping: {desc["db_id"]} {foo("bar", x=123)} {"foo" != "bar"} {(x := "abc=")} {pos_share=} {desc["status"]} {desc["exposure_max"]}'
  )
  
  log.info(
@@ -981,6 +981,18 @@ log.info(f"""Skipping: {'a' == 'b'} {desc['ms_name']} {money=} {dte=} {pos_share
  )
  
  log.info(
+-    f"""Skipping: {"a" == 'b'} {desc["ms_name"]} {money=} {dte=} {pos_share=} {desc["status"]} {desc["exposure_max"]}"""
++    f"""Skipping: {"a" == "b"} {desc["ms_name"]} {money=} {dte=} {pos_share=} {desc["status"]} {desc["exposure_max"]}"""
+ )
+ 
+ log.info(
+@@ -590,5 +442,5 @@
+ )
+ 
+ log.info(
+-    f"""Skipping: {'a' == 'b'} {desc['ms_name']} {money=} {dte=} {pos_share=} {desc['status']} {desc['exposure_max']}"""
++    f"""Skipping: {"a" == "b"} {desc["ms_name"]} {money=} {dte=} {pos_share=} {desc["status"]} {desc["exposure_max"]}"""
+ )
 ```
 
 ## Ruff Output
@@ -1394,7 +1406,7 @@ log.info(
 )
 
 log.info(
-    f'Skipping: {desc["db_id"]} {foo("bar",x=123)} {"foo" != "bar"} {(x := "abc=")} {pos_share=} {desc["status"]} {desc["exposure_max"]}'
+    f'Skipping: {desc["db_id"]} {foo("bar", x=123)} {"foo" != "bar"} {(x := "abc=")} {pos_share=} {desc["status"]} {desc["exposure_max"]}'
 )
 
 log.info(
@@ -1422,7 +1434,7 @@ log.info(
 )
 
 log.info(
-    f"""Skipping: {"a" == 'b'} {desc["ms_name"]} {money=} {dte=} {pos_share=} {desc["status"]} {desc["exposure_max"]}"""
+    f"""Skipping: {"a" == "b"} {desc["ms_name"]} {money=} {dte=} {pos_share=} {desc["status"]} {desc["exposure_max"]}"""
 )
 
 log.info(
@@ -1430,7 +1442,7 @@ log.info(
 )
 
 log.info(
-    f"""Skipping: {'a' == 'b'} {desc['ms_name']} {money=} {dte=} {pos_share=} {desc['status']} {desc['exposure_max']}"""
+    f"""Skipping: {"a" == "b"} {desc["ms_name"]} {money=} {dte=} {pos_share=} {desc["status"]} {desc["exposure_max"]}"""
 )
 ```
 

crates/ruff_python_formatter/tests/snapshots/black_compatibility@cases__preview_long_strings__regression.py.snap

@@ -832,7 +832,7 @@ s = f'Lorem Ipsum is simply dummy text of the printing and typesetting industry:
  
  some_commented_string = (  # This comment stays at the top.
      "This string is long but not so long that it needs hahahah toooooo be so greatttt"
-@@ -279,36 +280,25 @@
+@@ -279,37 +280,26 @@
  )
  
  lpar_and_rpar_have_comments = func_call(  # LPAR Comment
@@ -852,31 +852,32 @@ s = f'Lorem Ipsum is simply dummy text of the printing and typesetting industry:
 -    f" {'' if ID is None else ID} | perl -nE 'print if /^{field}:/'"
 -)
 +cmd_fstring = f"sudo -E deluge-console info --detailed --sort-reverse=time_added {'' if ID is None else ID} | perl -nE 'print if /^{field}:/'"
++
++cmd_fstring = f"sudo -E deluge-console info --detailed --sort-reverse=time_added {'{{}}' if ID is None else ID} | perl -nE 'print if /^{field}:/'"
  
 -cmd_fstring = (
 -    "sudo -E deluge-console info --detailed --sort-reverse=time_added"
 -    f" {'{{}}' if ID is None else ID} | perl -nE 'print if /^{field}:/'"
 -)
-+cmd_fstring = f"sudo -E deluge-console info --detailed --sort-reverse=time_added {'{{}}' if ID is None else ID} | perl -nE 'print if /^{field}:/'"
++cmd_fstring = f"sudo -E deluge-console info --detailed --sort-reverse=time_added {{'' if ID is None else ID}} | perl -nE 'print if /^{field}:/'"
  
 -cmd_fstring = (
 -    "sudo -E deluge-console info --detailed --sort-reverse=time_added {'' if ID is"
 -    f" None else ID}} | perl -nE 'print if /^{field}:/'"
 -)
-+cmd_fstring = f"sudo -E deluge-console info --detailed --sort-reverse=time_added {{'' if ID is None else ID}} | perl -nE 'print if /^{field}:/'"
- 
 +fstring = f"This string really doesn't need to be an {{{{fstring}}}}, but this one most certainly, absolutely {does}."
-+
+ 
  fstring = (
 -    "This string really doesn't need to be an {{fstring}}, but this one most"
 -    f" certainly, absolutely {does}."
 +    f"We have to remember to escape {braces}." " Like {these}." f" But not {this}."
  )
--
--fstring = f"We have to remember to escape {braces}. Like {{these}}. But not {this}."
  
+-fstring = f"We have to remember to escape {braces}. Like {{these}}. But not {this}."
+-
  
  class A:
+     class B:
 @@ -364,10 +354,7 @@
          def foo():
              if not hasattr(module, name):
@@ -979,7 +980,13 @@ s = f'Lorem Ipsum is simply dummy text of the printing and typesetting industry:
  )
  
  # The parens should NOT be removed in this case.
-@@ -518,88 +494,78 @@
+@@ -513,93 +489,83 @@
+ 
+ 
+ temp_msg = (
+-    f"{f'{humanize_number(pos)}.': <{pound_len+2}} "
++    f"{f'{humanize_number(pos)}.': <{pound_len + 2}} "
+     f"{balance: <{bal_len + 5}} "
      f"<<{author.display_name}>>\n"
  )
  
@@ -1103,7 +1110,13 @@ s = f'Lorem Ipsum is simply dummy text of the printing and typesetting industry:
      "6. Click on Create Credential at the top."
      '7. At the top click the link for "API key".'
      "8. No application restrictions are needed. Click Create at the bottom."
-@@ -613,55 +579,40 @@
+@@ -608,60 +574,45 @@
+ 
+ # It shouldn't matter if the string prefixes are capitalized.
+ temp_msg = (
+-    f"{F'{humanize_number(pos)}.': <{pound_len+2}} "
++    f"{f'{humanize_number(pos)}.': <{pound_len + 2}} "
+     f"{balance: <{bal_len + 5}} "
      f"<<{author.display_name}>>\n"
  )
  
@@ -1688,7 +1701,7 @@ class X:
 
 
 temp_msg = (
-    f"{f'{humanize_number(pos)}.': <{pound_len+2}} "
+    f"{f'{humanize_number(pos)}.': <{pound_len + 2}} "
     f"{balance: <{bal_len + 5}} "
     f"<<{author.display_name}>>\n"
 )
@@ -1773,7 +1786,7 @@ message = (
 
 # It shouldn't matter if the string prefixes are capitalized.
 temp_msg = (
-    f"{F'{humanize_number(pos)}.': <{pound_len+2}} "
+    f"{f'{humanize_number(pos)}.': <{pound_len + 2}} "
     f"{balance: <{bal_len + 5}} "
     f"<<{author.display_name}>>\n"
 )