[pyupgrade] Fix parsing named Unicode escape sequences (UP032)

[phongddo]

Summary

Fixes #19771

Fixes incorrect parsing of Unicode named escape sequences like Hey \N{snowman} in FormatString, which were being incorrectly split into separate literal and field parts instead of being treated as a single literal unit.

Problem

The FormatString parser incorrectly handles Unicode named escape sequences:

• Current: Hey \N{snowman} is parsed into 2 parts Literal("Hey \N") & Field("snowman")
• Expected: Hey \N{snowman} should be parsed into 1 part Literal("Hey \N{snowman}")

This affects f-string conversion rules when fixing UP032 that rely on proper format string parsing.

Solution

I modified parse_literal to detect and handle Unicode named escape sequences before parsing single characters:

• Introduced a flag to track when a backslash is "available" to escape something.
• When the flag is true, and the text starts with N{, try to parse the complete Unicode escape sequence as one unit, and set the flag to false after parsing successfully.
• Set the flag to false when the backslash is already consumed.

Manual Verification

"\N{angle}AOB = {angle}°".format(angle=180)

Result

 def foo():
-    "\N{angle}AOB = {angle}°".format(angle=180)
+    f"\N{angle}AOB = {180}°"

Would fix 1 error.

"\N{snowman} {snowman}".format(snowman=1)

Result

 def foo():
-    "\N{snowman} {snowman}".format(snowman=1)
+    f"\N{snowman} {1}"

Would fix 1 error.

"\\N{snowman} {snowman}".format(snowman=1)

Result

 def foo():
-    "\\N{snowman} {snowman}".format(snowman=1)
+    f"\\N{1} {1}"

Would fix 1 error.

Test Plan

• Added test cases (happy case, invalid case, edge case) for FormatString when parsing Unicode escape sequence.
• Updated snapshots.

[astral-sh-bot]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[ntBre]

Thanks! This makes sense to me, I just had a couple of small suggestions about the tests.

I took a closer look at this code today, and I'm feeling much less wary than in #19774 (review). Our parser will have already flagged weird invalid cases like these that I brought up last time:

"\N{{angle}}".format(angle="angle")
"\N{LATIN {SMALL} LETTER A}"

So we only need to handle valid cases, which this PR seems to do in a nice way. I also ran the fuzzer for a little while, just in case.

[astral-sh-bot]

Diagnostic diff on typing conformance tests

No changes detected when running ty on typing conformance tests ✅

[astral-sh-bot]

mypy_primer results

Changes were detected when running on open source projects

scikit-build-core (https://github.com/scikit-build/scikit-build-core)
+ src/scikit_build_core/build/wheel.py:98:20: error[no-matching-overload] No overload of bound method `__init__` matches arguments
- Found 41 diagnostics
+ Found 42 diagnostics

Memory usage changes were detected when running on open source projects

sphinx (https://github.com/sphinx-doc/sphinx)
- WARN expected `heap_size` to be provided by Salsa query `class_based_items`
- WARN expected `heap_size` to be provided by Salsa query `class_based_items`
- WARN expected `heap_size` to be provided by Salsa query `class_based_items`
- WARN expected `heap_size` to be provided by Salsa query `class_based_items`

prefect (https://github.com/PrefectHQ/prefect)
+ WARN expected `heap_size` to be provided by Salsa query `class_based_items`
+ WARN expected `heap_size` to be provided by Salsa query `class_based_items`
+ WARN expected `heap_size` to be provided by Salsa query `class_based_items`
+ WARN expected `heap_size` to be provided by Salsa query `class_based_items`

[phongddo]

Hey @ntBre 👋 just in case you missed the updates on this PR. I've addressed your comments about the tests. Please take a look when you have a moment and let me know if there's anything else I should adjust 🙏

[ntBre]

Thank you!

This looks good to go, just one small test comment update, and I think I preferred your old tests (to avoid adding a dev-dependency to this crate). Sorry for the flip-flop and the delay.

[ntBre]

Thank you!