Files
Ahsan-Sarbaz ac797a05ec [PERF] Remove \b anchor from escaped-unicode codePointPat
The leading \b blocked the regexp engine's literal-prefix fast path, so
codePointPat was tried at every offset in every chunk. Apply the word-boundary
check in Go instead, gate each pattern on its fixed literal anchor, and let
FromChunk reuse the match list it already computed rather than scanning a
second time to decode.

Adds an equivalence test comparing the new implementation against the old one
on generated, hand-picked, and fuzzed inputs.
2026-09-16 14:58:23 +05:00
..