Fix exponential backoff in multiline string regexp #157
Conversation
Since the middle of the replaced regexp was a greedy match, it required exponential time to match (or fail to match) a very long literal. For instance, if we had a `"""` followed by ~700 lines of code that didn't contain a multine string, trying to match that with the regexp would not terminate.
|
|
|
Thank you for the PR. Yes please modify the |
|
Just checked, making that regex lazy breaks highlighting. Sample: https://github.com/lampepfl/dotty/blob/master/compiler/test/dotty/tools/vulpix/ParallelTesting.scala#L95. Everything from that line onwards is HL-ed as string. Does this ring a bell? I'll look into it later. |
|
Ah, I was wrong in my analysis. The start re actually matches the whole string, not just the start.
|
|
Now, if you make that reluctant, then it will stop before the end of the multi-line string, and thus the next |
Hello there!
We've found an issue where emacs would freeze when using scala-mode scrolling to line https://github.com/lampepfl/dotty/blob/master/compiler/src/dotty/tools/dotc/typer/Namer.scala#L1199. If you put the point at the beginning of the trailing """ and call scala-syntax:looking-at-simplePattern-beginning, Emacs will freeze. I've traced this down to the
scala-syntax:multiLineStringLiteral-reregexp.Since the middle of the replaced regexp was a greedy match, it required exponential time to match (or fail to match) a very long literal.
After making
*lazy, things no longer freeze.I tried to make the change as unintrusive as possible.
scala-syntax:multiLineStringLiteral-start-recould probably be changed as well. Let me know if I should continue to dig into this, I'd be willing to do it if I had some help.