Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-10-07

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:23 flaviu1 joined #marpa
01:45 flaviu1 joined #marpa
01:47 ilbot3 joined #marpa
01:47 Topic for #marpa is now Start here: http://savage.net.au/Marpa.html - Pastebin: http://scsys.co.uk:8002/marpa - Jeffrey's Marpa site: http://jeffreykegler.github.io/Marpa-web-site/ - IRC log: http://irclog.perlgeek.de/marpa/today
01:55 jeffreykegler joined #marpa
01:55 jeffreykegler jdurand: re http://irclog.perlgeek.de/​marpa/2014-10-06#i_9465377
01:55 flaviu1 joined #marpa
01:56 jeffreykegler trigger_events() does nothing except trigger events, I believe, and therefore should be fast if there are no events --
01:57 jeffreykegler Are there numbers showing that the cost is high in the no-event case?
02:49 ronsavage joined #marpa
02:56 flaviu1 joined #marpa
03:12 lwa joined #marpa
03:38 CQ_ joined #marpa
04:26 rns joined #marpa
05:06 rns ronsavage: re http://irclog.perlgeek.de/​marpa/2014-10-06#i_9465793 — good question.
05:08 rns At http://lua-users.org/wiki/LuaGrammar they say "Those reimplementing Lua may be interested in the test suite for 5.1" so probably the test suite might be sufficient to exercise the parser well.
05:08 jdurand joined #marpa
05:08 rns Also, googling for lua parser tests gives plenty of links so some corner cases can be pulled from there.
05:09 jdurand Re http://irclog.perlgeek.de/​marpa/2014-10-07#i_9466880 - indeed you are right.
05:10 jdurand With no event, it is almost invisible in the profile. I.e. consumption of trigger_events() is < 1% of marpa_r_earleme() profiling
05:12 jdurand With 17 nullable events on nullable symbols, it takes 45% of marpa_r_earleme() profile.
05:12 jdurand Quick conclusion: events are powerful but have a true cost
05:14 rns left #marpa
05:14 rns joined #marpa
05:14 rns left #marpa
05:16 jdurand joined #marpa
05:17 jdurand Comes to my mind another question: do all type of events (prediction, nullable, completion) have the same cost ?
05:19 ronsavage joined #marpa
05:21 ronsavage rns: Thanx.
05:21 jeffreykegler joined #marpa
05:22 jeffreykegler re http://irclog.perlgeek.de/​marpa/2014-10-07#i_9467277 -- events can be deactivated and reactivated.  Perhaps that will help.
05:28 jeffreykegler Jumping ahead, once Kollos reaches a stage where the LUIF incorporates most of the SLIF, I plan to move a lot of stuff that's in the parse engine to an upper layer ...
05:29 jeffreykegler which will open the road to beaucoup optimizations.
05:30 jeffreykegler We'll be able to look at the basic data structures and see if something a little faster can't be done.
05:30 jeffreykegler Meantime, it'd be hard to measure any of this stuff, because of the overhead in the upper layers.
06:06 rns joined #marpa
06:46 rns left #marpa
09:43 hernanuo1 joined #marpa
10:39 rns joined #marpa
10:40 rns ronsavage: re sample grammars: perhaps this date parser code from my answer on StackOverflow http://stackoverflow.com/questions/26226838/ho​w-would-you-parse-a-date-of-any-reasonable-for​mat-from-a-big-chunk-of-text/26233929#26233929 can qualify.
10:41 rns It's a quick hack of sorts I've been trying recently with Marpa.
10:41 rns left #marpa
15:52 jeffreykegler joined #marpa
19:38 flaviu1 joined #marpa
21:41 ronsavage rns: PHP: That's what I use as $work. And yes, thanx, it makes another great example.
23:28 jeffreykegler joined #marpa

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary