Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-03-29

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

All times shown according to UTC.

Time Nick Message
01:12 jeffreykegler joined #marpa
01:15 jeffreykegler jest: re http://irclog.perlgeek.de/​marpa/2014-03-28#i_8508866 -- I am not forgetting this.  I am working in the same area of code on some other issues, and once I'm done Ill move straight on to checking out the "miraculous behavior".  That might be in a couple of days.
03:15 jeffreykegler joined #marpa
03:17 jeffreykegler jest: I have the answer re http://irclog.perlgeek.de/​marpa/2014-03-28#i_8508866 -- <qstring content> is not looked for as a lexeme, because it is *not* a lexeme.
03:18 jeffreykegler <qstring content> is an internal symbol in the lexer's grammar, and will only be looked for in the context of a <qstring> -- this is regardless of whether LTM or LATM is in use.
03:19 jeffreykegler Only top-level symbols in L0 are lexemes, and only they are subject to the LATM or LTM discipline.
03:46 jeffreykegler1 joined #marpa
04:51 jeffreykegler joined #marpa
06:21 jest joined #marpa
06:24 jest jeffreykegler: OK, I see it, it gets "included" inside quotes, so the full lexem recognized always starts with " and <qstring content> is only needed from DSL's point of view.
06:24 jest thanks
10:56 jeffreykegler joined #marpa
10:58 jeffreykegler jdurand: re XML parsing, I glanced at your code and I think you're matching tags using events.  (Let me know if that's wrong.)
10:59 jeffreykegler That approach is likely to kill your performance with event processing.  Here's an alternative that I implemented in Marpa::R2::HTML:
10:59 jeffreykegler First, lex the entire input using HTML::Parser.
11:00 jeffreykegler Second, as you lex (or in a 2nd pass) do a census of all the tags.
11:01 jeffreykegler Third, write your grammar, using only tags present in the actual input, and parse with that.
11:03 jeffreykegler Marpa::R2::HTML uses this approach, and gets quite fancy -- since the grammar is written on the fly, I made the exact interpretation of liberal HTML configurable.
11:05 jeffreykegler I blogged on this a while ago: http://jeffreykegler.github.io/Ocean-of-Aware​ness-blog/metapages/annotated.html#PARSE_HTML
11:07 jeffreykegler The blog was written pre-SLIF, but the ideas can be transferred to a SLIF-based approach.
11:25 jest joined #marpa
11:39 LLamaRider joined #marpa
11:46 LLamalessRider joined #marpa
16:29 jeffreykegler joined #marpa
20:26 LLamaRider joined #marpa
20:27 LLamalessRider joined #marpa
22:58 ronsavage joined #marpa
23:37 jeffreykegler joined #marpa
23:38 jeffreykegler I've just uploaded a new developer's release to CPAN: Marpa-R2-2.083_001
23:39 jeffreykegler It removes the requirement for "long long int", which should allow compilation with some MS compilers which do not have it.
23:41 jeffreykegler Also fixed is an issue with LATM -- in a few cases it fell back on the older, less efficient LATM implementation, which I had not removed.  Now only the newest, efficient LATM implementation remains.  Some applications might see measureable gains in speed.
23:42 jeffreykegler Finally, I've updated the documentation so that all the examples use LATM.
23:42 jeffreykegler Marpa-R2-2.083_001 is a release candidate, and testing is appreciated.  Thanks!

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