Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-05-02

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

All times shown according to UTC.

Time Nick Message
01:20 jeffreykegler joined #marpa
01:21 jeffreykegler jdurand: That should be marpa_g_symbol_is_completion_event_set().  I've just pushed a doc fix.
01:23 jeffreykegler jdurand: This is not the first time you've been the 1st to thoroughly explore an aspect of Marpa, and of course your "reward" is shaking out all sorts of nits.
01:23 jeffreykegler Anyway, I am most grateful.  It does make things smoother for the next person to come along. :-)
03:29 ronsavage joined #marpa
03:31 ronsavage I just added https://jeffreykegler.github.io/Marpa-w​eb-site/libmarpa_api/latest/index.html to the FAQ in http://savage.net.au/Perl-​modules/html/marpa.papers/
04:24 jeffreykegler ronsavage: Actually could you make that a link to http://jeffreykegler.github.com/Marpa-web-​site/libmarpa_api/cpan_indexed/index.html -- that's the one from the stable, indexed release.
04:27 jeffreykegler The links jdurand and I throw back and forth at each other are often from the bleeding edge development version of the Libmarpa API doc, which is from a version of Marpa which is not even guaranteed to compile correctly.
04:27 jeffreykegler Thanks
04:28 jeffreykegler ronsavage: Depending on what you are doing in the particular FAQ, you might find this more suitable -- it's the Libmarpa "home" page --
04:28 jeffreykegler http://jeffreykegler.github.io​/Marpa-web-site/libmarpa.html
04:57 jeffreykegler Here is an academic blogger talking about a course where he taught regular expressions, CFL'
04:58 jeffreykegler CFL's and computability theory including P and NP
04:58 jeffreykegler http://blog.computationalcomplexity.org/2014​/04/changing-order-you-teach-things-in.html
05:00 jeffreykegler He mentions in passing he's just dropped CFL's from the course.
05:00 jeffreykegler "I dumped CFLs-- don't worry, they cover them some in the Prog Langs course"
05:03 jeffreykegler It's a sign of how deeply unfashionable parsing theory has become -- he teaches at a good school.
07:42 ronsavage joined #marpa
08:53 jeffreykegler joined #marpa
08:53 yxhuvud joined #marpa
08:53 sivoais joined #marpa
08:53 lucs joined #marpa
08:53 shadowpaste joined #marpa
10:25 jeffreykegler joined #marpa
10:31 jeffreykegler joined #marpa
13:06 jeffreykegler joined #marpa
14:47 LLamaRider joined #marpa
15:35 jeffreykegler joined #marpa
15:52 yxhuvud jeffrey: yikes. dropping CFLs seem really strange.
15:53 jeffreykegler It certainly means graduates from that school will probably think little about the issue of parsing -- they may not even realize there is one.
15:54 jeffreykegler And I re-emphasize -- it's not that it's a bad school, not at all -- this seems to be going on all over.
16:00 jeffreykegler The effect of course is to freeze innovation in parsing, which is what has happened.
16:00 jeffreykegler Recursive descent is the state of the art, and it was fully developed in 1961.
16:02 jeffreykegler Not that the literature has not gone on -- I depended heavily on Leo's 1991 result, and also used Aycock and Horspool.
16:02 jeffreykegler But it does explain why Leo's wonderful result just sat for 20 years, waiting for me to use it.
16:42 LLamaRider jeffreykegler: it is amazing how often I get asked "Why this regex can't parse balanced parantheses?" by people without a CS degree
16:42 LLamaRider Luckily my university teaches foundations of formal languages early on, but somehow people still don't have an appreciation for parsers
16:43 jeffreykegler Outside the US, I think parsing has stayed in the curriculum longer.  We Americans tend to think in practical terms ...
16:44 jeffreykegler and if a technology doesn't get used for 50 years, US schools will tend to stop teaching it.
16:45 jeffreykegler You don't need to know CFL's to do recursive descent or regular expressions, and LALR (yacc/bison) has essentially been abandoned, so CFL's are seen as pointless.
16:46 jeffreykegler But of course without them you can't even describe the problem Marpa tackles, much less try to solve it.
16:52 yxhuvud I'm certainly glad I took the automata course (regexen through CFLs to Turing machines) as a freshman. it is a very nice foundation.
19:04 LLamaRider joined #marpa
19:06 LLamaRider jeffreykegler: How does the building of the C libmarpa work during Marpa::R2 installation?
19:07 LLamaRider I am writing an XS wrapper for a C library and using Marpa's XS as a guide for the advanced parts.
19:09 jeffreykegler It's a vanilla autoconf build, called from Marpa::R2's build procedure.
19:09 jeffreykegler This is assuming you don't have one of the systems that needs Config::Autoconf, in which case things get more complex.
19:12 jeffreykegler In the standard Marpa::R2 build, configures command line is used to instruct autoconf to build a static libmarpa, but with PIC, and then Perl code links this into the R2 shared library.
19:13 jeffreykegler Jean-Damien and I on this channel have been working on shared library support, which I think is well along, and this allows another approach, which is to install libmarpa somewhere as a shared library, and have your XS code link to that shared library.
19:13 LLamaRider jeffreykegler: Thanks. It's a curious dance.
19:14 LLamaRider Having it as a shared library that your package manager brings along will certainly make installation simpler (though you need separate versioning for libmarpa and Marpa::R2 then)
19:14 jeffreykegler I must say having a combined Module::Build and GNU autoconf build has not been pleasant.
19:14 jeffreykegler Libmarpa has separate versioning.
19:15 LLamaRider cool.
19:15 LLamaRider I am trying to avoid using autoconf, but that may be too optimistic
19:15 jeffreykegler Libmarpa is currently 6.0.4
19:16 jeffreykegler Jean-Damien is working on a Debian package, so that might not be too optimistic.
19:16 LLamaRider Oh, I am sure it would work great for Marpa! Best of success to Jean-Damien
19:17 jeffreykegler So if you can install libmarpa from a package, you can avoid autoconf.
19:17 jeffreykegler The other avoidance strategy is Config::Autoconf, a Perl package which emulates part of autoconf.
19:18 jeffreykegler LLamaRider: you may want to ask Jean-Damien'
19:18 jeffreykegler s advice when he's online.
19:18 LLamaRider jeffreykegler: thanks, I think my current questions are well answered :)
19:19 LLamaRider I am also happy to see there are some efforts to write grammars in pure C.
19:19 LLamaRider Ultra-performant grammars are something I can definitely showcase in my community, they keep asking for speed
19:20 LLamaRider sadly running out of battery; gotta go for a while; thanks for the help!
19:20 jeffreykegler I know.  Much of the early advice I received when I started writing Marpa was the standard "do not optimize"
19:20 jeffreykegler Like most of the advice I got, it was very wrong -- when it comes to parsers, people value speed.
19:21 jeffreykegler LLamaRider: bye
20:29 LLamaRider joined #marpa
21:39 LLamaRider joined #marpa
23:51 jeffreykegler joined #marpa

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