Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-11-06

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

All times shown according to UTC.

Time Nick Message
00:07 jeffreykegler joined #marpa
02:03 ronsavage joined #marpa
02:13 aredridel joined #marpa
03:26 ronsavage jdurand: Sorry - I did not answer your question: http://irclog.perlgeek.de/marpa/2014-10-30#i_9590417. Yes, I use Tree::DAG_Node to stash progress, so I don't care about the value() result. I do care whether or not it's defined, of course.
06:49 shadowpaste "ronsavage" at 217.168.150.38 pasted "BNF problem" (253 lines) at http://fpaste.scsys.co.uk/438181
06:52 ronsavage Why does $shadowpaste match node_statement but not also attribute_statement? rank makes no difference. And priority doesn't work on prioritized rules.
07:35 lwa joined #marpa
11:38 koo6 joined #marpa
15:00 koo6 joined #marpa
16:27 jeffreykegler joined #marpa
16:39 jeffreykegler ronsavage: re http://irclog.perlgeek.de/marpa/2014-11-06#i_9618347
16:39 jeffreykegler Looked at this only quickly, I don't know if the following comments address your question
16:40 jeffreykegler There is a "node_name" event defined, but it is never checked for.
16:42 jeffreykegler In event loops, by the way, I *always* check for spurious events -- if an event is intended to be ignored, I'll code that explicitly so that an event not explicitly handled is always a fatal error.
16:43 jeffreykegler If this doesn't help (and I'm far from sure I'm addressing your issue) please annotate the script so it's clear what behavior is expected, and where it is not happening.
16:45 jeffreykegler Test::More can be a framework for this, but a check or sometimes just a comment can be OK as long as it's totally clear.
16:51 rns_ joined #marpa
16:52 rns_ Adding square brackets negation to string lexeme s{ string ~ [^\s]+ } { string ~ [^\s\[\]]+ } makes it parse the input.
16:53 rns_ left #marpa
16:53 jeffreykegler rns: Thanks!
18:08 jluis_ joined #marpa
19:15 ronsavage joined #marpa
20:27 ronsavage Many thanx. You're both right. This is a small part of the upcoming GraphViz2::Marpa V 2.00. The original parses the grammar and puts the event names into a hashref, which is checked in _validate_event(). Cutting down lost that logic. This problem here is not the problem I'm having with the code. A new problem arose here, so I patched the gist and forgot to add back the check for the node_name event. "square brackets negation" is a different
22:01 flaviu joined #marpa
22:34 ronsavage I've just found a case where the input is 83 lines and Marpa takes just over 7 seconds. I'm amazed - normally it's 1 second max.
22:48 idiosyncrat joined #marpa
22:48 idiosyncrat ronsavage: Heavily event-driven?
22:50 idiosyncrat That's all done with Perl callbacks, and yes it's expensive compared to the stuff Marpa can do in pure C.
22:50 idiosyncrat With Kollos I hope events will become cheaper
23:03 ronsavage Yes, heavily event driven. And it's 3 times the size of the next largest file, which takes 0.678 secs.
23:06 ronsavage Massive progress with GraphViz2::Marpa. It now passes all of its tests which are expected to pass. I'll move on to the tests the re-write used to fail on.

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