Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-01-22

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

All times shown according to UTC.

Time Nick Message
23:05 jeffreykegler joined #marpa
01:59 ronsavage joined #marpa
03:41 ronsavage joined #marpa
04:49 jeffreykegler Phase 2 is going well.  The LR(0) states now correspond one-to-one with Earley items, which is another way of saying that they've been eliminated.
04:49 jeffreykegler Marpa will now be a simpler algorithm, one that looks much more like Jay Earley's original.
04:50 jeffreykegler Phase 2, alas, won't produce any visible external value added -- that should happen in Phase 3.
04:52 jeffreykegler Remaining in Phase 2 is some followup -- putting the test suite back into regular order, and cleaning up *some* of the overheads that the LR(0) states required.
04:52 jeffreykegler It's amazing how much more complicated the LR(0) states made things.
04:53 jeffreykegler Good night!
04:53 jeffreykegler left #marpa
06:04 ronsavage joined #marpa
08:19 _rns joined #marpa
09:24 LLamaRider joined #marpa
09:27 _rns Hi Deyan
09:44 LLamaRider Morning _rns :)
09:50 LLamaRider I just repeated the exercise from yesterday on my office PC, so now I have build Marpa successfully twice
09:50 LLamaRider that's a good sign :)
09:52 _rns Morningm, LLamaRider, sure thing!
10:13 LLamaRider So, I'm running your panda1 test and the rule_closure print doesn'
10:13 LLamaRider doesn't return anything
10:18 _rns That's because it has no closures for now; the real test so far is in cpan/t/sl_null_example.t
10:20 LLamaRider good to know
10:20 LLamaRider ah, excellent!
10:21 _rns Yep; panda1 isn't in manifest for a reason :) I have a version with closures to test rule_closure() in ASF traversers.
10:22 LLamaRider Since your test just works, I will try to write a mock grammar that does something meaningful using the rule closures
10:22 LLamaRider but i'm mixing it up with $work, so it's an asynchronous development process :D
10:23 LLamaRider i'll try to adapt one of my toy grammars : https://gist.github.com/dginev/6419166
10:24 _rns Once rh_values() and rule_closure() make it to the upstream.
10:26 LLamaRider Once that happens, then what? I didn't follow.
10:26 _rns Seems like a good test.
10:26 _rns Then they can be used together in ASF traversers.
10:27 LLamaRider right! But I built your github repo, so I get to try it now :) Hopefully Jeffrey merges it soon too
10:30 _rns That was the original idea as I saw it then: to call a rule closure from ASF traverser to have semantic values (parse values after action is applied). )
10:30 _rns Sure, hope I didn't mess with the internals too much. :)
10:39 LLamaRider this is pretty interesting actually. I need ambiguity-aware semantic routines now, if I just invoke them inside full_traversal
10:43 _rns The grammar is ambiguous because it does nto use precedence. Is it by design?
10:43 _rns does not use precedence, sorry. :)
10:44 LLamaRider yes, it is designed to be ambiguous :)
10:44 LLamaRider it's the tiniest toy ambiguous grammar i could factor out of my huge math expression gramamr
10:47 _rns Well, I see. :) The goal is to study how to handle ambuguity in Marpa? Does your huge math grammar uses prececence, if I may ask :)?
10:51 _rns I mean, just in case, that precedence is easy with Scanless::G grammars.
10:52 LLamaRider It doesn't use precedence, since they are not naturally modelling language
10:53 LLamaRider it uses hierarchies of categories, just as any classic context-free grammar
10:53 LLamaRider i'm modeling math in scientific publications btw, not in programming languages
10:57 _rns Formulas from math papers, ok. Must be an interesting thing to model. Oh, sorry, I'll be AFK for an hour. See you later.
10:59 LLamaRider have fun afk
11:01 _rns :)
11:45 _rns BTK.
12:27 LLamaRider joined #marpa
13:45 LLamaRider have fun afk
13:45 LLamaRider ah, sorry, didn't mean to write that again
13:54 _rns No problem: fun afk, fun btk. :)
14:11 _rns left #marpa
18:58 jeffreykegler joined #marpa
19:07 jdurand joined #marpa
19:33 jeffreykegler left #marpa
20:19 jdurand Hello, aside cpretty that I promissed, I have another idea using MarpaX::Languages::C::AST : make a c2xs, whose goal would be an alternative of the bloated
20:20 jdurand and untrusty (when used with another compiler than gcc, guaranteed) h2xs
20:20 jdurand Opinions welcome
21:08 ronsavage joined #marpa
21:11 rns_ joined #marpa
21:31 jeffreykegler joined #marpa
21:32 jeffreykegler jdurand: c2xs -- I like the idea.  You want to consider the support issue.  If you're not going to be a heavy user yourself, and ...
21:33 jeffreykegler you don't quickly attract users, you'll either have to support it or give up on it.
21:34 jeffreykegler XS is a kind of moving target, so you'd either have to continually update along with Perl development, or have the tool become useless.
21:36 jeffreykegler Build tools have a history of becoming quagmires for their developers.
21:36 jeffreykegler left #marpa
21:38 jdurand jeffreykegler: noted - thx for the feedback
22:45 ronsavage jdurand: A c2xs tool means people might not need to learn XS. That would be a big plus, if it could automate writing interfaces to C libraries

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