Uh oh!
There was an error while loading. Please reload this page.
Skip RegexOptions.NonBacktracking phase 3 for some patterns - #65531
Conversation
If a pattern doesn't have any captures and if any match of that pattern will always be the same length, we can skip the Phase 3 computation as, given the computed starting position of the match, we know exactly where it's going to end.
ghost
commented
Feb 17, 2022
Tagging subscribers to this area: @dotnet/area-system-text-regularexpressions Issue DetailsIf a pattern doesn't have any captures and if any match of that pattern will always be the same length, we can skip the Phase 3 computation as, given the computed starting position of the match, we know exactly where it's going to end. Also took the opportunity to add some comments. Fixes #65383 (mostly)
|
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
| if (startat == input.Length) | ||
| { | ||
| // Covers the special-case of an empty match at the end of the input. | ||
| uint prevKind = GetCharKind(input, startat - 1); | ||
| uint nextKind = GetCharKind(input, startat); | ||
| bool emptyMatchExists = _pattern.IsNullableFor(CharKind.Context(prevKind, nextKind)); | ||
| return emptyMatchExists ? | ||
| return _pattern.IsNullableFor(CharKind.Context(prevKind, nextKind)) ? | ||
| new SymbolicMatch(startat, 0) : | ||
| SymbolicMatch.NoMatch; | ||
| } |
There was a problem hiding this comment.
Looking at the new comment I realized this optimization should additionally handle the case where there are capture groups and do the same ApplyEffects thing FindEndPositionCapturing is doing. The difference will be visible for some patterns with nullable capture groups that have anchors in them.
Edit: Oh actually any nullable patterns with nullable capture groups.
There was a problem hiding this comment.
Thanks, @olsaarik. That's pre-existing this PR, yes?
There was a problem hiding this comment.
Yes, shouldn't block this PR, just something I noticed.
Improvements in dotnet/perf-autofiling-issues#3665, dotnet/perf-autofiling-issues#3672 |
If a pattern doesn't have any captures and if any match of that pattern will always be the same length, we can skip the Phase 3 computation as, given the computed starting position of the match, we know exactly where it's going to end.
Also took the opportunity to add some comments.
Fixes#65383 (mostly... I think the rest is #65532)