[daboehme]

Should fix the issue but not tested with Python 3.15 yet.

Here's what Claude says:

Fixed Python 3.15 compatibility issue with re.Scanner by replacing capturing groups with non-capturing groups:

Changes made:

1. OuterCommentLexer (lines 306-309): Changed (.|[\r\n]) to (?:.|[\r\n]) in the multiline and single-line comment patterns, and ({(?!{)...}) to (?:{(?!{)...}) in the text pattern.

2. InnerLexer (lines 313-318): Replaced the complex pattern with backreferences (["\'])?((?:(?!\1)[^\\]|\\.)*)\1 with two simpler patterns:

   • r'"(?:[^"\\]|\\.)*"' for double-quoted strings
   • r"'(?:[^'\\]|\\.)*'" for single-quoted strings

   This avoids the backreference \1 which required a capturing group.

Why these changes work:

• Python 3.15 raised restrictions on re.Scanner to disallow capturing groups since the Scanner API doesn't provide access to captured subgroups anyway
• Non-capturing groups (?:...) provide the same grouping behavior for alternation and repetition without creating captures
• The InnerLexer change splits one complex pattern into two simpler ones, which is actually clearer and avoids the backreference entirely

The quoted_id callback function at line 323 will still work correctly because it processes the entire matched token string (with quotes) using re.sub() to strip the quotes.