Exclude pragma comments from measured line width (#7008)

Co-authored-by: Micha Reiser <[email protected]>

Author: cnpryer

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

@@ -1,6 +1,29 @@
+# Pragma reserved width fixtures
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # noqa: This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # type: This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # pyright: This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # pylint: This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # noqa This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # nocoverage: This should break
+
+
+# Pragma fixtures for non-breaking space (lead by NBSP)
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # noqa: This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # type: This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # pyright: This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # pylint: This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # noqa This shouldn't break
+i = ("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa",)  # nocoverage: This should break
+
+
 # As of adding this fixture Black adds a space before the non-breaking space if part of a type pragma.
 # https://github.com/psf/black/blob/b4dca26c7d93f930bbd5a7b552807370b60d4298/src/black/comments.py#L122-L129
-i2 = ""  #         type: Add space before leading NBSP followed by spaces
-i3 = ""  #type: A space is added
-i4 = ""  #  type: Add space before leading NBSP followed by a space
-i5 = ""  # type: Add space before leading NBSP
+i = ""  #         type: Add space before leading NBSP followed by spaces
+i = ""  #type: A space is added
+i = ""  #  type: Add space before leading NBSP followed by a space
+i = ""  # type: Add space before leading NBSP
+i = ""  #  type: Add space before two leading NBSP
+
+
+# A noqa as `#\u{A0}\u{A0}noqa` becomes `# \u{A0}noqa`
+i = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"  #  noqa

crates/ruff_python_formatter/src/comments/format.rs

@@ -357,17 +357,33 @@ impl Format<PyFormatContext<'_>> for FormatTrailingEndOfLineComment<'_> {
 
         let normalized_comment = normalize_comment(self.comment, source)?;
 
-        // Start with 2 because of the two leading spaces.
-        let mut reserved_width = 2;
-
-        // SAFE: The formatted file is <= 4GB, and each comment should as well.
-        #[allow(clippy::cast_possible_truncation)]
-        for c in normalized_comment.chars() {
-            reserved_width += match c {
-                '\t' => f.options().tab_width().value(),
-                c => c.width().unwrap_or(0) as u32,
+        // Trim the normalized comment to detect excluded pragmas (strips NBSP).
+        let trimmed = strip_comment_prefix(&normalized_comment)?.trim_start();
+
+        let is_pragma = if let Some((maybe_pragma, _)) = trimmed.split_once(':') {
+            matches!(maybe_pragma, "noqa" | "type" | "pyright" | "pylint")
+        } else {
+            trimmed.starts_with("noqa")
+        };
+
+        // Don't reserve width for excluded pragma comments.
+        let reserved_width = if is_pragma {
+            0
+        } else {
+            // Start with 2 because of the two leading spaces.
+            let mut width = 2;
+
+            // SAFETY: The formatted file is <= 4GB, and each comment should as well.
+            #[allow(clippy::cast_possible_truncation)]
+            for c in normalized_comment.chars() {
+                width += match c {
+                    '\t' => f.options().tab_width().value(),
+                    c => c.width().unwrap_or(0) as u32,
+                }
             }
-        }
+
+            width
+        };
 
         write!(
             f,
@@ -442,11 +458,7 @@ fn normalize_comment<'a>(
 
     let trimmed = comment_text.trim_end();
 
-    let Some(content) = trimmed.strip_prefix('#') else {
-        return Err(FormatError::syntax_error(
-            "Didn't find expected comment token `#`",
-        ));
-    };
+    let content = strip_comment_prefix(trimmed)?;
 
     if content.is_empty() {
         return Ok(Cow::Borrowed("#"));
@@ -476,6 +488,17 @@ fn normalize_comment<'a>(
     Ok(Cow::Owned(std::format!("# {}", content.trim_start())))
 }
 
+/// A helper for stripping '#' from comments.
+fn strip_comment_prefix(comment_text: &str) -> FormatResult<&str> {
+    let Some(content) = comment_text.strip_prefix('#') else {
+        return Err(FormatError::syntax_error(
+            "Didn't find expected comment token `#`",
+        ));
+    };
+
+    Ok(content)
+}
+
 /// Format the empty lines between a node and its trailing comments.
 ///
 /// For example, given:

crates/ruff_python_formatter/tests/snapshots/black_compatibility@simple_cases__comments6.py.snap

@@ -156,7 +156,7 @@ aaaaaaaaaaaaa, bbbbbbbbb = map(list, map(itertools.chain.from_iterable, zip(*ite
      )
  
  
-@@ -108,11 +112,20 @@
+@@ -108,11 +112,18 @@
      "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
  )
  
@@ -176,10 +176,7 @@ aaaaaaaaaaaaa, bbbbbbbbb = map(list, map(itertools.chain.from_iterable, zip(*ite
 +    ],  # type: ignore
  )
  
--aaaaaaaaaaaaa, bbbbbbbbb = map(list, map(itertools.chain.from_iterable, zip(*items)))  # type: ignore[arg-type]
-+aaaaaaaaaaaaa, bbbbbbbbb = map(
-+    list, map(itertools.chain.from_iterable, zip(*items))
-+)  # type: ignore[arg-type]
+ aaaaaaaaaaaaa, bbbbbbbbb = map(list, map(itertools.chain.from_iterable, zip(*items)))  # type: ignore[arg-type]
 ```
 
 ## Ruff Output
@@ -313,9 +310,7 @@ call_to_some_function_asdf(
     ],  # type: ignore
 )
 
-aaaaaaaaaaaaa, bbbbbbbbb = map(
-    list, map(itertools.chain.from_iterable, zip(*items))
-)  # type: ignore[arg-type]
+aaaaaaaaaaaaa, bbbbbbbbb = map(list, map(itertools.chain.from_iterable, zip(*items)))  # type: ignore[arg-type]
 ```
 
 ## Black Output

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

@@ -300,17 +300,6 @@ last_call()
  )  # note: no trailing comma pre-3.6
  call(*gidgets[:2])
  call(a, *gidgets[:2])
-@@ -142,7 +143,9 @@
- xxxx_xxxxx_xxxx_xxx: Callable[..., List[SomeClass]] = classmethod(  # type: ignore
-     sync(async_xxxx_xxx_xxxx_xxxxx_xxxx_xxx.__func__)
- )
--xxxx_xxx_xxxx_xxxxx_xxxx_xxx: Callable[..., List[SomeClass]] = classmethod(  # type: ignore
-+xxxx_xxx_xxxx_xxxxx_xxxx_xxx: Callable[
-+    ..., List[SomeClass]
-+] = classmethod(  # type: ignore
-     sync(async_xxxx_xxx_xxxx_xxxxx_xxxx_xxx.__func__)
- )
- xxxx_xxx_xxxx_xxxxx_xxxx_xxx: Callable[..., List[SomeClass]] = classmethod(
 ```
 
 ## Ruff Output
@@ -461,9 +450,7 @@ very_long_variable_name_filters: t.List[
 xxxx_xxxxx_xxxx_xxx: Callable[..., List[SomeClass]] = classmethod(  # type: ignore
     sync(async_xxxx_xxx_xxxx_xxxxx_xxxx_xxx.__func__)
 )
-xxxx_xxx_xxxx_xxxxx_xxxx_xxx: Callable[
-    ..., List[SomeClass]
-] = classmethod(  # type: ignore
+xxxx_xxx_xxxx_xxxxx_xxxx_xxx: Callable[..., List[SomeClass]] = classmethod(  # type: ignore
     sync(async_xxxx_xxx_xxxx_xxxxx_xxxx_xxx.__func__)
 )
 xxxx_xxx_xxxx_xxxxx_xxxx_xxx: Callable[..., List[SomeClass]] = classmethod(