[Cherith1222]

Why

rf"\{foobar:>3o}" is valid Python. It evaluates to a backslash plus the formatted value. parso treated that backslash as escaping the brace, so parse(..., error_recovery=False) raised ParserSyntaxError on the closing brace.

What changed

• An f-string escape now matches a backslash before { without consuming the brace.
• \N{SNOWMAN} and a failed name such as \N{ BULLET } stay on the old path.
• Regression cases cover raw strings, format specs, and a triple-quoted string.

Scope

• Covers a backslash immediately before a replacement field.
• Leaves the pep8 normalizer crash on a backslash-newline comment alone.
• Does not change format-spec rules.

Tradeoffs

A bare backslash alternative also matched a failed \N escape and made f"\N{ BULLET }" parse. The pattern uses a lookahead so the extra match applies only when { is next.

Blast Radius

Callers that parse f-strings, including Jedi, now accept this form instead of reporting a syntax error. The existing escape and invalid-name tests still pass.

Verification

• Before the fix, test_valid[rf"\\{foobar:>3o}"] failed with ParserSyntaxError on } at column 15.
• After the fix, py -3.13 -m pytest -q reported 1993 passed.
• The reported snippet parses, and get_code() returns the same source.

Fixes #207

Drafted with Cursor.