fix: remove invalid unicode escapes from regExpEscape

[abhu85]

Summary

Fixes #273 - Invalid regex generated when glob includes comma and character class.

The regExpEscape function was escaping characters like ,, #, -, and whitespace with backslash (e.g., \,). However, these escape sequences are invalid in unicode regex mode (/u flag).

When a pattern includes a POSIX character class like [[:space:]], the unicode flag is required because POSIX classes use unicode property escapes (e.g., \p{Z}). This caused patterns like ,[[:space:]] to throw:

SyntaxError: Invalid regular expression: /^\,[\p{Z}\t\r\n\v\f]$/u: Invalid escape

Changes

• Modified regExpEscape in src/ast.ts to only escape characters that are both:

  1. Regex metacharacters (need escaping)
  2. Valid to escape in unicode mode

• Added comprehensive test suite in test/posix-class-with-comma.js covering:

  • Comma with POSIX character classes
  • Hash character with POSIX classes
  • Space character with POSIX classes
  • Hyphen outside character class with POSIX classes
  • Multiple special characters combined

Test Plan

• Reproduced the original issue - minimatch('foo', ',[[:space:]]') threw SyntaxError
• Verified the fix resolves the issue - no longer throws
• Verified correct matching behavior (comma-space matches, comma-x doesn't)
• All existing tests pass (6249 total)
• Updated snapshots for changed regex patterns (semantically equivalent)

Generated with Claude Code