From 5de4cc56086493689701e86aa0ccf6a4a4a87d75 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Chris=20Hasi=C5=84ski?= Date: Sun, 11 Jan 2026 14:12:07 +0100 Subject: [PATCH] Fix regexp performance regression for patterns starting with s/k Commit 981ee02c7c ("Fix performance problem with /k/i and /s/i") was merged for Ruby 4.0 to enable partial Boyer-Moore optimization for patterns containing 's' or 'k' by using the prefix before those characters. However, when 's' or 'k' appears at the start of a pattern (no usable prefix), set_bm_skip() returns 0 and the code returned early without setting any optimization mode, leaving reg->optimize at ONIG_OPTIMIZE_NONE. This caused up to 30x slowdown for patterns like /slackware/i when matched against strings with non-ASCII characters. This patch keeps the improvement from 981ee02c7c for patterns with 3+ char prefix, while fixing the regression by falling back to ONIG_OPTIMIZE_EXACT_IC with the full pattern when the usable prefix is less than 3 characters. Before: /\bslackware\b/i with non-ASCII string: 2.24 us/op After: /\bslackware\b/i with non-ASCII string: 0.70 us/op (3.2x faster) [Bug #21824] --- regcomp.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) diff --git a/regcomp.c b/regcomp.c index 18b2c97eb6..320cf520e9 100644 --- a/regcomp.c +++ b/regcomp.c @@ -5264,18 +5264,24 @@ set_optimize_exact_info(regex_t* reg, OptExactInfo* e) if (e->ignore_case > 0) { if (e->len >= 3 || (e->len >= 2 && allow_reverse)) { + int orig_len = e->len; e->len = set_bm_skip(reg->exact, reg->exact_end, reg, reg->map, 1); - reg->exact_end = reg->exact + e->len; if (e->len >= 3) { + reg->exact_end = reg->exact + e->len; reg->optimize = (allow_reverse != 0 ? ONIG_OPTIMIZE_EXACT_BM_IC : ONIG_OPTIMIZE_EXACT_BM_NOT_REV_IC); } - else if (e->len > 0) { + else { + /* Even if BM skip table can't be built (e.g., pattern starts with + 's' or 'k' which have multi-byte case fold variants), we should + still use EXACT_IC optimization with the original pattern. + Without this fallback, patterns like /slackware/i have no + optimization at all, causing severe performance regression + especially with non-ASCII strings. See [Bug #21824] */ + e->len = orig_len; /* Restore original length for EXACT_IC */ reg->optimize = ONIG_OPTIMIZE_EXACT_IC; } - else - return 0; } else { reg->optimize = ONIG_OPTIMIZE_EXACT_IC;