11:19 <kares[m]> lopex: hey, just curious - joni's regexps are interruptible are how is it actually solved to not run 'forever' ?

11:25 <lopex> kares[m]: https://github.com/jruby/joni/blob/master/test/org/joni/test/TestInterrupt.java#L85

11:25 <lopex> but we can hack it whatever we want

11:26 <lopex> like test time every n bytecode loop

11:27 <kares[m]> oh nice to have an actual test to see this, thanks

11:28 <kares[m]> will poke around the impl a bit :)

11:28 <kares[m]> maybe JRuby should showcase how it can deal with user-input regexps

11:29 <kares[m]> ... might be interesting to some Ruby folks

11:29 <lopex> kares[m]: https://github.com/jruby/joni/blob/master/src/org/joni/ByteCodeMachine.java#L173

11:29 <lopex> enebo[m]: added this

11:30 <kares[m]> why did he do that I mean it makes sense but what was the motivation

11:30 <kares[m]> guess some Ruby-ness let's see what git history says ;)

11:31 <lopex> there was an issue for that, I forgot what it was

14:01 <enebo[m]> kares: I just added parameterization so you could pick to interupt or not. I think we improved some internal paths/uses of match by disabling the interupt feature (since we know it was not needed). The wrapper around the match call ended up being significant overhead

14:02 <enebo[m]> kares: in fact I really think this was us on the date parsing stuff which uncovered that

14:17 <kares[m]> interesting, thanks for the feedback

14:35 <lopex> yeah, like internal regexps etc

14:37 <lopex> .match(..) by default doesnt check for interrupts

14:37 <lopex> only matchInterruptible does

14:38 <lopex> afair it was from when were optimizing =~

14:38 <lopex> *we were

14:39 <lopex> ah, no, it that issue so captures were ignored

14:40 <lopex> er it was "match?" actually

14:43 <enebo[m]> lopex: ah yeah match? was definitely the main reason

14:44 <enebo[m]> lopex: inasmuch that he wanted to eliminate backend costs in joni for setting up region match logic/setup

14:44 <enebo[m]> since there was none