Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2016-04-03

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

All times shown according to UTC.

Time Nick Message
01:46 idiosyncrat_ choroba: I am looking for examples of your problem with Unicode in the SLIF DSL ...
01:46 idiosyncrat_ so that I can fix it in Marpa::R3, but all your pastes have expired.
01:46 idiosyncrat_ choroba:  Never mind!!  Found it: https://github.com/jeffreyk​egler/Marpa--R2/issues/268
01:53 ronsavage joined #marpa
03:18 ronsavage joined #marpa
03:19 idiosyncrat_ ronsavage: re my conversion to EU::MM ...
03:20 idiosyncrat_ I think a major reason was documentation & examples ... I find the EU::MM docs better than the Module::Build ones, and ...
03:21 idiosyncrat_ for complex cases like Marpa I often wanted to look at examples -- how others had solved their problems in similarly complex cases.
03:21 kaare_ joined #marpa
03:21 idiosyncrat_ But almost all the people who were really good at Perl packaging stuck with EU::MM -- as a Module::Build user I was pretty much on my own.
03:22 idiosyncrat_ Since the conversion to EU::MM, I've already solved several problems by looking at the Makefile.PL for other modules on CPAN.
03:23 idiosyncrat_ Finally, I think the design of EU::MM is more "right", at least in the Marpa case.
03:24 idiosyncrat_ By producing a Makefile as output, EU::MM makes it easy to figure out what's going on, at least for me ...
03:25 idiosyncrat_ and limiting Marpa's builds to systems with a working 'make' is no problem -- libmarpa requires make anyway.
03:26 teatime Did R3 releasE?
03:27 idiosyncrat_ teatime: I'm creating "trial" branches, which can be used on a "alpha" and "experimental" basis.
03:28 idiosyncrat_ I haven't decided what to do about CPAN releases yet.
03:28 teatime Is the bugfix in for unicode chars in scripts/
03:28 idiosyncrat_ But, you caught me just as I was about to make an announcement.
03:28 teatime ?
03:28 idiosyncrat_ Yes, that was the announcement.
03:28 teatime lol; sorry to steal the thunder :)
03:29 idiosyncrat_ Actually, I think of you of creating a fanfare, and I am very appreciative.
03:29 idiosyncrat_ I went for years releasing Marpa and being totally ignored, so "getting my thunder stolen" in this way is a very pleasant change.
03:29 teatime ahaha
03:30 idiosyncrat_ Anyway ....
03:30 idiosyncrat_ The 'trial' branch of Marpa::R3 on Github contains the fix to the Unicode problem in SLIF DSL scripts.
03:30 idiosyncrat_ It passes the test suite.
03:31 idiosyncrat_ jdurand: thanks for spotting the right section of code for the fix.
03:31 idiosyncrat_ My fix was slightly different from yours.
03:32 idiosyncrat_ For the benefit of others: the problem was that I used a Marpa-internal table of characters to speed up lexing, which I constructed on a just-in-time basis.
03:33 idiosyncrat_ This worked beautifully, and allowed me to handle Unicode with almost no additional cost to programs which didn't use it, but the problem was ...
03:33 idiosyncrat_ that at one point I passed the new character around as a number, that is ...
03:34 idiosyncrat_ I converted it from a string to a number, and then back for pattern matching.
03:35 idiosyncrat_ This worked beautifully, except that Perl also tries to get the best of efficient 8-bit strings and ability to support Unicode, and it does this by tagging the strings according to whether they
03:35 idiosyncrat_ were UTF8 or not.
03:35 idiosyncrat_ My conversion from string to number and back lost this information.
03:37 idiosyncrat_ I solved the problem by falling back to grabbing the character via Perl's substr() call, which preserves UTF8'ness.
03:39 idiosyncrat_ Simply upgrading the character to UTF8 would not work, because if someone is trying to use one of the old  ISO/IEC 8859 encodings, and you interpret it as UTF8, it comes out wrong.
03:40 idiosyncrat_ Folks using Perl 5 are encouraged to decode these, but not everyone does, and for backward compatibility they are not forced to decode ISO/IEC 8859 into UTF8 if they don't want to.
03:42 idiosyncrat_ This also solves choroba's problem with his Karel scripts ...
03:42 idiosyncrat_ and I thank choroba for finding the problem and creating the test example, ...
03:43 idiosyncrat_ and Jean-Damien for finding the exact section of code that needed to be fixed.
03:44 idiosyncrat_ Good night!
08:00 koo7 joined #marpa
08:17 jdurand joined #marpa
08:17 jdurand Re http://irclog.perlgeek.de/m​arpa/2016-04-03#i_12277987 - yep much better (and correct, definitely) fix - thx
09:56 koo7 joined #marpa
11:47 rns I've tried to build Marpa-R3-1.001_000 manually under windows; findings:
11:48 rns 1) Makefile.PL tries to run configure (cmd /x /c configure --with-pic --disable-shared) even if Config::AutoConf is used.
11:49 rns 2) cd engine/perl_ac_build && $(MAKE) fails because no Makefile is written due to 1)
14:09 koo7 joined #marpa
14:26 Idiosyncrat joined #marpa
18:11 idiosyncrat_ joined #marpa
18:11 idiosyncrat_ rns: re http://irclog.perlgeek.de/m​arpa/2016-04-03#i_12278767
18:14 idiosyncrat_ Makefile.PL in the Config::Autoconf build directory is written by this script: https://github.com/jeffreykegler/Marpa--R3/b​lob/master/cpan/engine/cf/write_makefile.pl
18:15 idiosyncrat_ You can change it -- make sure you don't break the releng target.
18:15 idiosyncrat_ I don't think the releng target uses Makefile.PL, in which case we are pretty safe.
18:34 rns idiosyncrat: cpan/engine/cf/write_makefile.pl isn't in MANIFEST and doesn't get in the dist
18:35 rns when I copied it there manually, it builds libmarpa.lib
18:36 rns ok but main Makefile (as written by main Makefile.PL) tries to copy libmarpa.lib from
18:37 rns engine/perl_ac_build/.libs/ which doesn't exist and thus fails.
18:38 idiosyncrat_ I didn't update cpan/engine/cf/write_makefile.pl for this release, because I just use the releng target ...
18:38 idiosyncrat_ and I don't really even know what cpan/engine/cf/write_makefile.pl is for.
18:39 idiosyncrat_ So feel free to make the changes you need to it, and to add it to the MANIFEST, so long as 'releng' does not break.
18:40 idiosyncrat_ My understanding is that cpan/engine/cf/write_makefile.pl is useful for doing development on Windows, which I don't do.
18:43 idiosyncrat_ IIRC the 'releng' target do not use or run the cpan/engine/perl_ac_build/Makefile.PL ...
18:44 idiosyncrat_ I think the solution is to create a cpan/engine/cf/write_makefile.pl that writes a cpan/engine/perl_ac_build/Makefile.PL that expects things to be where the 'releng' target puts them.
18:45 idiosyncrat_ But I may misunderstand, in which case, let me know.
18:47 rns ok, I'll see what can be done and let you know.
19:45 idiosyncrat_ AFK
22:02 koo7 im starting to realize marpa isnt too useful for autocompletion
22:06 koo7 given a grammar like number "+" number, if i enter "3 +", marpa will accept my tokens, and then later tell me that the input is incomplete...
22:07 koo7 do i have a way to tell what symbol it is expecting next?
22:09 koo7 im not too interested in knowing only the expected token..
22:13 koo7 ah
22:13 koo7 MARPA_EVENT_SYMBOL_PREDICTED
22:20 koo7 getting back to my experiments after a while:)
22:28 idiosyncrat_ joined #marpa
22:35 ronsavage joined #marpa
22:50 idiosyncrat_ I'm going to remove the "latm" adverb from the SLIF.
22:51 idiosyncrat_ This will break most Marpa::R2 scripts, which I think is a feature ...
22:52 idiosyncrat_ If you use the wrong script for Marpa::R3, rather than a lot of subtle errors over time, you will get one obvious error immediately.
22:52 idiosyncrat_ I'm starting a census of the Marpa::R3 tests -- of those which test the NAIF, I want to figure out which should be ported to the SLIF, and which should simply be deleted.
22:53 idiosyncrat_ Once this census is done, I will declare Marpa::R3 open to refactorings which break the NAIF.
23:23 koo7 does marpa_r_predicted_symbol_event_set exist?
23:23 koo7 http://jeffreykegler.github.io/Marpa-web-s​ite/libmarpa_api/stable/api_one_page.html only mentions it
23:26 koo7 github.com:jeffreykegler/libmarpa.git master sources dont
23:44 idiosyncrat_ koo7: re http://irclog.perlgeek.de/m​arpa/2016-04-03#i_12280683
23:44 idiosyncrat_ You're right -- marpa_r_predicted_symbol_event_set() is in the docs, but does not exist.
23:45 ronsavage joined #marpa
23:45 koo7 i think expected nonterminals is exactly what i need

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