My personal build of CMark ✏️
Only memoize during inline parsing.
This cuts the performance hit.
With memoization, we get roughly constant behavior in the fuzztest. Without it, not.
1 file changed, 4 insertions, 4 deletions
| Status | File Name | N° Changes | Insertions | Deletions |
| Modified | js/stmd.js | 8 | 4 | 4 |
diff --git a/js/stmd.js b/js/stmd.js @@ -455,7 +455,7 @@ } } - if ((next_inline = this.parseInline())) { + if ((next_inline = this.parseInline(true))) { Array.prototype.push.apply(current, next_inline); } else { break; @@ -743,10 +743,10 @@ // Parse the next inline element in subject, advancing subject position // and returning the inline parsed. - var parseInline = function() { + var parseInline = function(memoize) { var startpos = this.pos; - var memoized = this.memo[startpos]; + var memoized = memoize && this.memo[startpos]; if (memoized) { this.pos = memoized.endpos; return memoized.inline; @@ -793,7 +793,7 @@ res = [{t: 'Str', c: c}]; } - if (res) { + if (res && memoize) { this.memo[startpos] = { inline: res, endpos: this.pos }; }