Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2015-03-02

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

All times shown according to UTC.

Time Nick Message
00:18 ronsavage Using and over-using Git, etc: http://bitquabit.com/post/unorthodocs-abandon-your-dvcs-and-return-to-sanity/
01:00 jeffreykegler joined #marpa
01:01 jeffreykegler rns: I went ahead and did the MIT re-licensing -- it was all either judgment calls that I needed to make, or working with idiosyncratic code which will never have to be touched again -- either way stuff I should do.
01:02 jeffreykegler If you could double-check my work, that would be very much appreciated.
01:02 jeffreykegler In the process of testing this, I discovered our changes to Libmarpa had broken Marpa::R2 -- the fault being mine.
01:03 jeffreykegler marpa_b_new allows -1 for an ordinal argument -- it means use the Earley set at the current earleme if there is one.
01:03 jeffreykegler You did not take this into account -- quite reasonably, since I nowhere mentioned the use of the -1 value in the documentation.
01:04 jeffreykegler So I fixed the documentation, and then the code, and then the tests, and now all is OK.
01:33 jeffreykegler rns: What do you have in mind for Libmarpa tests in the near future?
01:34 jeffreykegler I ask because the MIT re-licensing seems as good an occasion for a new Marpa::R2 release as anything
01:34 * jeffreykegler should clarify that only Libmarpa will be under the MIT license, and that the licensing of the upper layers of Marpa::R2 is NOT changing
01:35 jeffreykegler So I wonder if there's something you have in mind worth waiting for, or if I should cut the Marpa::R2 release candidate shortly.
01:41 ronsavage After a long detour, including applying a local patch to Date::Calc because a module I use depends on it, I'm now again able to build by web site properly (i.e. afer switching to Perl 5.20.2). I've also just released 2 new modules (not Marpa-related) File :: BOM :: Utils and XML :: Tiny :: Tree.
01:41 ronsavage Next, I need to check the W3C definition of SVG to ensure they only contain features supported by XML :: TIny. Then I can patch MarpaX :: Languages :: SVG :: Parser to use XML :: Tiny. Then I'll get back to my other 2 Marpa modules which didn't run under 5.20.2. Then, or ignoring those issues, I'll get back to testing installation of Marpa::R2.
01:44 ronsavage Some other mysteries arose along the way, which I need to clean up. (1) sqlite3 comes in 2 incompatible versions, from Oracle and from SQLite.org. The latter one I can't get to even run any more. I rarely use it, but it worries me. Also, I have some local CGI DBs which don't now export their data perfectly. These (exports) are small issues, so they don't worry me.
01:44 ronsavage 'Also' => '(2) Also'
01:46 ronsavage (3) There are other small but messy things which I'd better clean up now rather than later. As I think I mentioned, I'm rectifying all the Build.PL/Makefile.PL code in all my modules, but patching various files per distro can't always be automated.
02:06 jeffreykegler joined #marpa
02:47 ilbot3 joined #marpa
02:47 Topic for #marpa is now Start here: http://savage.net.au/Marpa.html - Pastebin: http://scsys.co.uk:8002/marpa - Jeffrey's Marpa site: http://jeffreykegler.github.io/Marpa-web-site/ - IRC log: http://irclog.perlgeek.de/marpa/today
03:04 rns joined #marpa
03:10 rns jeffreykegler: re http://irclog.perlgeek.de/marpa/2015-03-02#i_10203259 -- the census is mostly done, I was thinking about the final round of checks -- I have nothing specific, just reading the doc and looking at the code to make sure I didn't miss something.
03:11 rns I don't expect it (1) will produce some considerable changes or (2) will produce them soon, so -- yes, Marpa::R2 release candidate seems to be a good idea.
03:12 rns re licensing checks -- I'll take a look, will keep you informed.
03:14 jeffreykegler rns: Perhaps it'd best if you looked over my re-licensing changes before I cut the release candidate ...
03:15 jeffreykegler If you do discover some mistakes, it's probably best I don't include that version with a Marpa::R2 indexed release.
03:15 jeffreykegler s/it'd best/it would be best/
03:16 jeffreykegler I have automated licensing checks, and also did some grepping for things I might have missed.
03:17 rns Ok, I'll put licensing checks first thing on my queue. Hope we'll be able to catch something, if any, at release candidate stage, before the indexed release.
03:24 rns I mean I'll try to look into licensing pretty soon so that you can start cutting the release candidate.
03:25 jeffreykegler To others: I ask rns because he's the one who has gotten into Libmarpa the deepest at this point.
03:26 jeffreykegler But anyone else who wants to glance into Libmarpa, just to check that I really caught all the license language and changed it properly, please do so, and thanks in advance!
03:27 jeffreykegler I know there is at least one reader whose employer forbade him to read LGPL'd code, and who will be happy to see the conversion to the MIT license ...
03:27 jeffreykegler and I suspect there are others.
03:28 Aria Indeed.
04:48 rns left #marpa
04:49 ronsavage joined #marpa
08:13 pczarn joined #marpa
09:30 lwa joined #marpa
09:39 koo7 joined #marpa
10:24 rns joined #marpa
10:33 rns left #marpa
10:33 rns joined #marpa
10:35 rns jeffreykegler: marpa_g_symbol_is_nulled_event_set(g, sym_id) doesn't check if a sym_id is nullable -- is this by design?
10:35 rns I take it if a symbol is not nullable it can't produce an empty string.
12:24 rns ronsavage: re http://irclog.perlgeek.de/marpa/2015-03-02#i_10203270 -- is it all related to 5.20.2 or you're just "cleaning out your closet"? I mean is an upgrade to 5.20.2 any good?
12:31 jeffreykegler joined #marpa
12:32 jeffreykegler rns: re http://irclog.perlgeek.de/marpa/2015-03-02#i_10204647 -- I cannot recall if it was design or happy accident. :-)
12:34 jeffreykegler There are cases where it'd convenient to set a nulled event for a non-nullable symbol -- the grammar may be automatically generated, for example.
12:34 jeffreykegler Note that if you really want to exclude that case, you can explicitly check the symbol for nullability.
12:35 jeffreykegler Nonetheless I should document this behavior.
12:38 rns jeffreykegler: thanks for explaining, so I'd better leave this behaviour as is. :)
12:48 jeffreykegler Documented in commit 2c77a3eb04b3e6476409a10e9cf68143e9ddec10
12:51 rns Great, this sorts that out very nicely.
12:52 jeffreykegler I hope there will be increasing use of languages that generate languages -- REAL higher-order languages -- and this behavior is, from their point of view, orthogonal.
12:54 rns Is it right that a nulled event set on a non-nullable symbol will never be triggered?
12:55 jeffreykegler Yes, such an event would never be triggered.
12:56 * jeffreykegler goes on to press a pedagogic point, however.
12:56 jeffreykegler Note that the "never" in the above statement is "never" from the point of view of a specific grammar.
12:58 jeffreykegler If you take the point of view of a THOL (true higher-order-language), so that your "universe" of cases includes sets of grammars, ...
12:58 rns Interesting ...
12:59 jeffreykegler then from the THOL's point of view there are cases where events for non-nullables might be triggered.
12:59 jeffreykegler I'm trying to get folks thinking at this level of abstraction.
13:00 jeffreykegler But it was a non-obvious behavior, and needed to be documented.
13:00 jeffreykegler rns: Good catch!
13:03 rns My pleasure, really. Re THOL -- A blog post on that kind of thinking would make a real nice read, I think, mind-boggling in a good sense of the word, like Larry's Perl as a postmodern computer language.
13:03 jeffreykegler For some reason, folks will listen to this kind of abstraction when Larry and Damian Conway do it, ...
13:04 jeffreykegler but are perplexed if I deliver anything except working code ...
13:04 jeffreykegler and sometimes even then.
13:04 jeffreykegler I've done a few (not using the term THOL) and these have been my *least* successful blog posts.
13:05 jeffreykegler The reaction was so bad that even the trolls ignored them. :-)
13:05 jeffreykegler But the SLIF, for example, is a THOL.
13:06 jeffreykegler That's how its precedenced statements work -- by rewriting them into another language.
13:07 rns Re THOL-like blog posts is it one of them -- http://blogs.perl.org/users/jeffrey_kegler/2012/11/a-grammar-that-exemplifies-describes-and-parses-itself.html?
13:09 jeffreykegler Yes -- in this case it describes self-parsing, which is something a THOL can do -- and something the SLIF does.
13:10 rns Ok, got it.
13:10 rns re http://irclog.perlgeek.de/marpa/2015-03-02#i_10205256 -- There is that idea in Grune 2008 that a parse result is really a grammar -- so there is only a source grammar, the input and the target grammar.
13:10 rns Grune & Jacobs 2nd ed, that is.
13:10 jeffreykegler Back in the 70's people, and I mean even COBOL programmers, talked about self-parsing languages all the time.
13:12 jeffreykegler Yes, the G&J idea is a specific application (one I have not really figured out a way to leverage yet)
13:13 rns trying to think along the THOL lines -- so it can be more useful, rather than a monolythic entity, to think about a complex grammar like a composition of smaller grammars, in which symbols may have different meaning depending on context?
13:13 jeffreykegler Another application was in Marpa::R2::HTML, which is configurable.
13:15 jeffreykegler rns: re http://irclog.perlgeek.de/marpa/2015-03-02#i_10205292 -- that'd be one way of thinking about a subset of the possibilities.
13:15 jeffreykegler More pedestrian was what I started on with HTML, for dealing with liberal HTML.
13:16 jeffreykegler If a table is broken, do you construct a new one?  And if so, how elaborate?  Different applications might want to try stray <td> tags as cruft ...
13:17 jeffreykegler rather than build a whole table around it.
13:17 jeffreykegler So this is a good case for configuring the parser.
13:18 jeffreykegler In this case, there's no big theoretical scaffolding -- just a few small tweaks to the grammar.
13:19 rns Yep, census first, then grammar generation, than parsing and it basically becomes an application-specific semantics for broken HTML.
13:19 jeffreykegler So THOL can go all the way from a cosmic scaffolding of language within language, to ...
13:19 jeffreykegler sticking very close to a basic grammar, and just allowing a tweak or two.
13:20 rns Yes, Going back in the conversation a bit --
13:20 rns re http://irclog.perlgeek.de/marpa/2015-03-02#i_10205287 -- I was thinking that libmarpa can build a Marpa_Grammar in the valuator and return it as an ast -- with suitable accessors to make it more convenient to use as a tree. Pure Marpa, of sorts.
13:20 jeffreykegler Yes, (re HTML), and that is how Marpa::R2::HTML is able to efficiently handle all kinds of tags -- it censuses the tags, then writes the grammar to suit.
13:21 rns Yes.
13:22 jeffreykegler re http://irclog.perlgeek.de/marpa/2015-03-02#i_10205331 -- an interesting idea.
13:24 jeffreykegler Going back to sleep -- AFK
13:24 rns Good mid-night!
16:34 jeffreykegler joined #marpa
16:39 pczarn joined #marpa
18:25 Aria YAY MIT LICENSE!
18:25 Aria I should work on a binding to node.
19:58 jeffreykegler joined #marpa
19:59 jeffreykegler Marpa-R2 2.103_009 is now up on CPAN
19:59 jeffreykegler It is a release candidate.
20:00 jeffreykegler It includes the new "discard events" and has at its core an MIT-licensed Libmarpa.
20:00 jeffreykegler Testing is appreciated!
20:57 ronsavage rns: Re http://irclog.perlgeek.de/marpa/2015-03-02#i_10205129. Perl V 5.20.2 is /almost/ OK. Some modules depend on Devel::Size directly or indirectly, and it's DOA. Some modules depend on Date::Calc, and it's tests started failing with the transition from 2014 to 2015 because of hardcoded 1964 in there. Some modules started failing because of dropped (long ago deprecated) methods in Data::Locale. Perhaps this is business-as-usual, but it's al
21:01 jeffreykegler ronsavage: FYI, your longer messages get cut off.
21:01 jeffreykegler This happens on the IRC log as well as in my IRC client, so I suspect it's at your end.
21:18 ronsavage (Continued:) Perhaps this is business-as-usual, but it's also painful.
21:19 ronsavage jk: Yes. I'm trying to retrain my typist tendencies.....
23:34 jeffreykegler joined #marpa
23:43 jeffreykegler I'm thinking seriously about a blog post on PEG.
23:44 jeffreykegler I'd resisted because I think I'm better employed improving Marpa, rather than describing its alternatives, and
23:44 jeffreykegler because given the misery of working with it, I figured PEG would go away.
23:45 jeffreykegler It was the hot new thing in 2007 when I started Marpa and I kept my mouth shut about it,
23:46 jeffreykegler but while PEG has not grown in popularity, it seems to attract a new wave of users as quickly as it disgruntles its current one, so that it is still around.
23:48 flaviu joined #marpa
23:58 ronsavage A blog post on PEG makes sense, if only to summarize its problems as a warning to newbies. Whether or not people ignore your post is their problem, since you can't force them to read it or act on it.
23:59 jeffreykegler Actually, my posts get attention these days.
23:59 jeffreykegler And people actually *ask* me about PEG.

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