RE#: how we built the world's fastest regex engine in F#
10 points by wareya
10 points by wareya
I'm suspicious of one of the benchmarks mentioned in the referenced paper (it seems like it would stress-test the performance of utf-8 validity or boundary checks more than the regex engine itself), but I've seen some interesting discussions happen around this project.
Haven't read the paper yet, but the dotnet implementation did quite well on burntsushi's rebar benchmarks two months ago https://github.com/burntsushi/rebar.
Ok one thing to note here is that they’re dropping some of the features that contribute to the performance costs of other re engines.
There’s another difference which is that by being on .net, which doesn’t effect their performance claims (they have the perf, how does not matter), but things like PCRE don’t use JITs - as doing so comes with significant security costs so it’s not a trivial trade off.
PCRE2 actually has a JIT. It's not enough to make it particularly competitive, though, in the cases that I've tested.