Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-10-05

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

All times shown according to UTC.

Time Nick Message
03:41 CQ_ joined #marpa
06:47 ronsavage joined #marpa
07:26 lwa joined #marpa
11:23 flaviu1 joined #marpa
14:12 flaviu1 joined #marpa
17:40 jeffreykegler joined #marpa
17:51 jdurand joined #marpa
18:04 jdurand Jeffrey, it seems that having events is quite a killer for marpa_r_earleme() - is there a way to simulate events without using them but coding explicitely something in the lexing phase
18:04 jdurand I meant marpa_r_earleme_complete(), sorry
18:08 jeffreykegler jdurand: you mean execute the actions as part of lexing.
18:09 jeffreykegler ?
18:14 jdurand yes
18:15 jdurand I meant: since I know which lexeme is accepted, I can trigger the action myself without relying on Marpa
18:15 jeffreykegler That's what the Kollos and Lua thing is all about.
18:16 jdurand Ha ha... I should be invovled in that then! Because I will very probably move to this solution
18:16 jeffreykegler It's been my hope to entice folks into the Lua/Kollos effort
18:17 jeffreykegler I've been unable to make progress due to maintenance for Marpa::R2 & Libmarpa
18:17 jeffreykegler One thing someone can do is write a Libmarpa library for Lua
18:18 jdurand What do you mean more precisely - a binding of libmarpa to Lua, or a rewrite in Lua native language
18:18 jeffreykegler The 1st.
18:19 jdurand IMHO there is no need to work on that. SWIG should be able to do it.
18:19 jdurand http://www.swig.org/Doc1.3/Lua.html
18:20 jeffreykegler Having time to figure out stuff like what strategy to use is where others can help.
18:20 jeffreykegler I'd also want some of the Thin interface features, which I dunno is compatible with a SWIG strategy
18:21 jeffreykegler For example, having a flag that allows errors to be either returned or thrown, the way Marpa's THIF does it.
18:21 jdurand Per def SWIG takes a header file, and turns it into a binding. If you put that in libmarpa, this will be in the binding.
18:21 jeffreykegler Btw, I'd hope others besides Jean-Damien would want to join in.
18:21 jdurand Nothing prevent to do more semantics around of course.
18:22 jdurand Ok, i'll do a proof of concept of using SWIG if you don't mind.
18:22 jdurand AFK for lunch -;
18:22 jeffreykegler If someone would write it, I'd pull it into Libmarpa
18:24 jeffreykegler Of course, in the process of taking it into Libmarpa it would probably be revised a lot -- I'd be fussy about it's exact features.
18:24 jeffreykegler But a Lua wrapper of Libmarpa is pretty straightforward -- there are not really a lot of decisions involved.
18:26 jeffreykegler I'd also very much hope others would get interested in this -- it's your chance to really hack Marpa close to the internals level which ...
18:27 jeffreykegler if you're one of those thinking to revise/fork/port the actual parse engine is an excellent way to get familar with the engine, and to get comfortable with the code.
18:40 jeffreykegler Another note -- as we move into Kollos/Lua development, I see high possibilities of forks and counter-forks --
18:41 jeffreykegler Basically I see that as a sign of more active, less centralized development, and as A Good Thing.
18:42 jeffreykegler That is, when I grab stuff somebody did into Marpa, I am likely to do so as a fork from it.  And also ...
18:43 jeffreykegler When you see something like the Lua wrapper of Libmarpa, and it's not done the way you want in my version, it'd become easier to just grab it and hack it in your own fork.
18:45 jeffreykegler I say this because forks are often seen as unfriendly, and I want that not to be the case in the Marpa community.
18:46 jeffreykegler Bottom line, Marpa is an algorithm, or even an approach, and the tendency with those is that it's often bascially cool, but just not quite right, and needs your own hacks.
18:49 jdurand In theory a SWIG wrapper is just a .i file (libmarpa.i to name it), with usually few per-languages bindings specificities if needed. I already played with that. So will do the proff of concept. I want to do deeper into libmarpa as well, since I am nor far away from it -;
18:49 jdurand Ok AFK I'll read the backlog as usual
18:50 jdurand joined #marpa
18:51 jdurand Oups, another question though. If the grammar has no event, will trigger_events consume time anway? Looking to the generated code it is not obvious if the answer is yes or no.
18:54 jeffreykegler Yes, there's some overhead for trigger_events even if there are no events.
19:00 jdurand Ok - thx
19:00 jdurand AFK for good
19:22 jeffreykegler jdurand: I assume you're backlogging --
19:23 jeffreykegler I suggest starting a Github repo for the Lua wrapper -- might as well call it Kollos, if you like the name.
19:23 jeffreykegler Also, for this I think Lua's license might be best -- it's an MIT license, looser than LGPL, but perhaps not unlike the Perl license.

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