Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2015-01-17

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

All times shown according to UTC.

Time Nick Message
02:45 jeffreykegler I'm annotating Andrew's FLOSS interview.
02:45 jeffreykegler At 27:00 he's asked what's Marpa's biggest data volume application to date.  Andrew didn't know and neither do I.
02:45 jeffreykegler Any guesses out there?
02:45 jeffreykegler It's probably one of the proprietary, and secret applications, but among public ones, it might be that Tivoli log parser over at IBM.
03:42 jeffreykegler jdurand: at 41:11 in the talk, Randal asks how you'd handle C typedefs in Marpa.  How did you actually end up dealing with them?
04:06 jeffreykegler https://gist.github.com/jeffreykegler/3563fd82932663837f01
04:06 jeffreykegler My notes on Andrew's FLOSS talk, as a Github gist.
04:32 RandalSchwartz joined #marpa
04:32 RandalSchwartz just reading jk's annotation of the floss episode
04:33 RandalSchwartz I had to guess what a couple of the items were referencing, but having been there in the first person, it wasn't that hard. ;)
04:34 jeffreykegler Ah!  I was going to send it on eventually, once I'd gathered feedback from the other Marpaistas.
04:34 RandalSchwartz I think I can get a link to it from the twit.tv page
04:34 jeffreykegler I thought you might find my remark on Perl 6 subgrammars of particular interest.
04:34 RandalSchwartz yeah
04:35 jeffreykegler The only combine nicely if all the backtracking and ratcheting, etc., happen to work out, which is possible but not likely ...
04:35 jeffreykegler and it becomes less possible as you scale.
04:36 RandalSchwartz yeah, it's like designing a good framework for subclassing
04:36 RandalSchwartz generally, you mostly hack until it works for you
04:36 RandalSchwartz anything overly abstract is useless, but sadly, anything under-abstract is also limited.
04:36 RandalSchwartz finding the sweet spot is an art.
04:36 jeffreykegler Whereas in writing Marpa grammars, we cut and paste BNF all the time, and it will just work.
04:36 RandalSchwartz or if it doesn't work, you'll find out fast. :)
04:37 jeffreykegler Because Marpa is BNF driven.
04:37 jeffreykegler It will work in the sense you'll get what the BNF describes.
04:37 RandalSchwartz does the underlying C engine call back to perl to do the perl regex match for tokens?
04:37 RandalSchwartz or are even the token regexes parsed out
04:38 jeffreykegler Yes, but it memoizes for efficiency.
04:38 jeffreykegler That is, it matches character classes, by callback to Perl.
04:39 jeffreykegler But memoizes what character classes match each character.
04:39 RandalSchwartz I'm glad while thinking on my feet that I was able to come up with "rfc822 format" as a familiar thing that has a precise BNF.
04:39 jeffreykegler This trick allows it to do full Unicode efficiently.
04:40 jeffreykegler Though I think the BNF actually may be in a different RFC. :-)  Let me check.
04:40 RandalSchwartz well, we're two revs past 822
04:40 RandalSchwartz it was 2822 first, and then something else later
04:40 RandalSchwartz I lost track
04:41 jeffreykegler Nope you're right, the BNF is in RFC822
04:41 RandalSchwartz yeah - I remember staring at that
04:41 RandalSchwartz it was the most complex bnf I had seen at the time
04:41 RandalSchwartz and yet affected every one of us every day
04:42 jeffreykegler When I took Comp. Sci, regular expression were new and advanced, and BNF was old and well-known.
04:42 RandalSchwartz even just defining what an email address looks like...
04:42 RandalSchwartz and the fact that you can't parse an email address with regex that isn't recursive (because of nested comments)... were both fascinating to me
04:43 jeffreykegler I don't think anybody has done an RFC822 Marpa parser yet.
04:43 RandalSchwartz also that merlyn (yeah that is me)@(at at at)stonehenge(the rock pile).(dot com?) com(!)
04:43 RandalSchwartz is legal
04:43 RandalSchwartz well... my first tackle will be undump
04:43 RandalSchwartz but I seem to recall someone else was working on that
04:44 jeffreykegler Not that I know of, actually.
04:44 jeffreykegler No volunteers emerged.
04:44 RandalSchwartz ok - I misremembered
04:44 RandalSchwartz the twit guys want me to hack some code to parse their apache log files
04:44 RandalSchwartz but I bet it'll be two modules and 10 lines of Perl
04:45 RandalSchwartz not a challenge for Marpa yet
04:45 jeffreykegler That's one problem with finding parsing apps in the Perl world ...
04:45 RandalSchwartz the thing I talked about in the show... parsing an XSD and generating the next grammar that would parse the XML into an AST the way I want...
04:45 jeffreykegler everybody has already solved most of the problems ... they did the hard way, but it is sunk costs.
04:45 RandalSchwartz *that* would be useful.
04:45 jeffreykegler I thought about that.
04:46 jeffreykegler Probably you want a liberal XML to tree parser for both the source and the XSB,
04:46 jeffreykegler XSB -> XSD
04:46 jeffreykegler and then work from the AST's, which means not really an app for a new parser.
04:46 RandalSchwartz yeah, since the XSD is also XML itself
04:47 RandalSchwartz and maybe this solves most of it already: http://search.cpan.org/dist/XML-Validator-Schema/
04:47 jeffreykegler That'd be my guess.
04:47 jeffreykegler I looked at leveraging the XSD in the parse itself,
04:47 jeffreykegler and it is really not a win.
04:48 jeffreykegler Because it makes sense to separate parsing and validation.
04:48 jeffreykegler That was my 1st guess, anyway.
04:48 RandalSchwartz well - I'd like at some point to provide annotations to say "and, I want this to be a hash, but that to just be a list of the contents.
04:49 RandalSchwartz I suppose I could just get the AST and write my custom tree-walkers
04:49 jeffreykegler That later place is where a parser might come in handy -- have a configuration language to describe how you interpret the XSD
04:49 RandalSchwartz Oooh!
04:49 * RandalSchwartz perks up his ears
04:49 jeffreykegler I was hoping you'd like that idea.
04:50 jeffreykegler Notice by the way that it is LOP.
04:50 jeffreykegler Solving problems with small languages.
04:50 RandalSchwartz Yeah - I solved a lot of issues with Config::General on my longest-running client
04:50 RandalSchwartz I really liked how this simple language combines in beautiful ways
04:51 jeffreykegler Config::General you mean?
04:51 RandalSchwartz No - actually, that's not it
04:51 RandalSchwartz darn it... what is it.
04:51 jeffreykegler Note that a lot of problem are caused by the emphasis on big languages.
04:51 jeffreykegler Like the Perl 6 vs. Perl 5 thing.
04:52 RandalSchwartz ahh.  Config::Scoped
04:52 jeffreykegler Languages have been so hard, they have to be big, and be all things to all people.
04:52 jeffreykegler Whereas if you can write one in a few hours, you can focus on dolng a single task well.
04:53 RandalSchwartz whoa... I forgot entirely about eval { ... }
04:54 RandalSchwartz that would have saved me some time recently. :)
04:54 RandalSchwartz and made my debugging that much harder
04:54 * RandalSchwartz relocates... back online in a few
04:59 RandalSchwartz ugh... was sitting at an airport bar, and left to hit the delta sky club... it closed at 7:30pm!  WTF
05:00 jeffreykegler Where are you
05:00 jeffreykegler just curious
05:01 RandalSchwartz PDX
05:01 RandalSchwartz enroute to ATL and MIA
05:01 RandalSchwartz and then Havana!
05:02 RandalSchwartz this trip has been planned for a year
05:02 jeffreykegler Enjoy.  I was there during the Carter administration.
05:02 RandalSchwartz and then they open up the rules the day before we travel there!
05:02 RandalSchwartz I can bring stuff back!
05:02 jeffreykegler And so just anybody and everybody will be there. :-)
05:03 RandalSchwartz not initially
05:03 RandalSchwartz there's some hysteresis
05:03 RandalSchwartz interesting.. Config::Scoped uses P::RD inside
05:03 RandalSchwartz I wonder what a Marpa version might be like
05:03 jeffreykegler Somebody will find out someday. :-)
05:04 jeffreykegler I'd wonder how robust a P::RD implementation is.
05:04 RandalSchwartz whoa... use base 'Parse::RecDescent';
05:04 RandalSchwartz it doesn't just *use* it... it subclasses it.
05:04 RandalSchwartz that's... scary.
05:05 jeffreykegler By the way, I've gotten some heat, but my docs say *not* to subclass Marpa::R2
05:05 RandalSchwartz I can understand
05:05 RandalSchwartz # Do you to understand this grammar?
05:05 RandalSchwartz # Be warned, dragons ahead, recursive brain damage possible!
05:05 jeffreykegler The namespace issues scared me.
05:05 RandalSchwartz that's in the Config::Scoped source
05:06 jeffreykegler There has almost developed an attitude that parsing *should* be hard and risky.
05:06 RandalSchwartz egad... now that I've stared at Marpa for a while, P::RD's arcanity boggles me
05:07 RandalSchwartz especially the <commit> for ratcheting
05:07 jeffreykegler Ah, you're getting into the mindset.
05:07 jeffreykegler I look at things like "ratcheting" and find it hard to realize that most people call that a *feature*.
05:07 RandalSchwartz I wanted to bring up something like
05:08 RandalSchwartz oops, nevermind talked about that last night
05:09 RandalSchwartz have you seen my "spew", which generates random strings based on weighted bnf-like rules?
05:09 RandalSchwartz also using PRD to parse that
05:09 jeffreykegler I don't remember it, if so.
05:09 RandalSchwartz ... https://metacpan.org/pod/Inline::Spew
05:10 jeffreykegler By the way, your examples are scattered throughout my test suite.
05:10 RandalSchwartz gawd.  code of mine from 12 years ago.
05:11 jeffreykegler https://metacpan.org/source/JKEGL/Marpa-R2-2.102000/t/randal.t
05:11 RandalSchwartz heh... The cost to compile a grammar is roughly a second on a reasonably speedy machine, so the grammar compilation is cached by the Inline mechanism.
05:11 RandalSchwartz I wonder how accurate that is now.
05:12 jeffreykegler Where is that from?
05:12 RandalSchwartz from Inline::Spew
05:12 RandalSchwartz heh... I recognize the famous sin/time argument
05:12 jeffreykegler I think you could be able to dispense with the caching.
05:13 RandalSchwartz so marpa can go from bnf to useful in pretty short time?
05:13 jeffreykegler In the words of Andrew Rodland, "yes"
05:14 jeffreykegler Assuming you call an AST useful.
05:14 RandalSchwartz I presume the bootstrapped marpa bnf is precompiled already
05:14 RandalSchwartz and how did you do that... iteratively?
05:14 RandalSchwartz or can you go from first principles still?
05:14 jeffreykegler No it is taken from an intermediate form.
05:14 jeffreykegler And the intermediate form is only used because of chicken-egg problems in development.
05:14 RandalSchwartz of course
05:15 RandalSchwartz I think Perl 6 can start from scratch, presuming you have perl 5 nearby
05:15 jeffreykegler I didn't follow that last.  "from scratch"?
05:16 RandalSchwartz as in... although some parts of perl6 are written in perl6, they're written in a subset, that is compiled by a more limited perl6, and that goes down a level or two until you get to perl5 doing the lowest bits.
05:16 jeffreykegler I don't know.
05:16 RandalSchwartz it'd be like "how would you write a C compiler written in itself today?"
05:17 jeffreykegler I think they had efficiency problems with the original vision of parsing Perl 6 in itself.
05:17 RandalSchwartz yeah - it's a few levels now
05:17 RandalSchwartz NQP comes to mind
05:17 RandalSchwartz I may be mis-membering
05:17 jeffreykegler Whereas of course, Marpa's SLIF parses itself quite easily.
05:19 jeffreykegler "Against top-down parsing, even geniuses contend in vain"
05:21 jeffreykegler RandalSchwartz: listening to the FLOSS thing for the 3rd time, I much appreciated your enthusiasm for the idea of Earley-driven parsing.
05:24 jeffreykegler RandalSchwartz: maybe to get you started,
05:24 jeffreykegler rewriting the spew parser might be a good place.
05:24 jeffreykegler I think it will be simple (don't remember, just guessing)
05:25 jeffreykegler and it's also an example of a 2nd order language.
05:28 jeffreykegler Listening to you on the FLOSS thing, I see now that my early presentation, which emphasized ambiguity, was probably not a good idea.
05:28 jeffreykegler I focused on it, because I needed severe tests of the algorithm,
05:29 jeffreykegler But putting it up front in describing what Marpa was about may have told folks, "This is something you do not need to care about"
05:30 jdurand joined #marpa
05:31 jdurand jeffreykegler, Re http://irclog.perlgeek.de/marpa/2015-01-17#i_9957999 - typedefs are resolved using the technique "Resolving Typedefs in a Multipass C Compiler" as per http://www.cs.dartmouth.edu/~mckeeman/references/JCLT/ResolvingTypedefsInAMultipassCCompiler.pdf
05:31 jdurand Implementation is mentionning explicitely this source as a credit, c.f. https://github.com/jddurand/MarpaX-Languages-C-AST/blob/master/lib/MarpaX/Languages/C/AST/Scope.pm
05:31 jeffreykegler jdurand: Thanks
05:32 jeffreykegler Great.  I'll add this to the notes on the FLOSS interview.
05:36 jeffreykegler jdurand: trying for a quick summary
05:36 jeffreykegler it looks like the paper proposes solving the problem in post-processing.
05:37 jdurand I have adapted this to lex phase using pauses
05:37 RandalSchwartz it's odd that this is one of those odd facts that I remembered about parsing while doing the interview
05:37 RandalSchwartz for some reason, I've always had a distant fascination with parsing
05:38 jeffreykegler Randal: yes, I was a bit shocked when you turned the conversation to typedefs.
05:38 RandalSchwartz I think it's because I was always a bit perplexed that lexx and yacc were separate.
05:38 jeffreykegler jdurand: so you solve it in lexing, using events?
05:38 jdurand jeffrekegler: yes
05:39 jeffreykegler Randal: events are another really cool feature of Marpa, which did not come up.
05:39 jeffreykegler jdurand: which events?  I'm curious.
05:40 RandalSchwartz yeah - I wasn't quite grokking them
05:40 jeffreykegler The idea is you can tell Marpa, stop processing when you see an <X> symbol, and hand control over to me.
05:41 jeffreykegler It's like RecDescent, except your parser can tell you exactly where you are.
05:41 jeffreykegler The best of both syntax-driven & procedural logic.
05:41 jeffreykegler In addition to spotting symbols, you can spot rules, or label arbitrary locations in rules using nulling symbols.
05:42 RandalSchwartz yeah - at that part, I sorta understood.  Kinda how larry solved the op vs value issue
05:42 jeffreykegler You can also ask for an event when something is predicted, so that you say, "When you are ready for an <X>, let me know".
05:43 jeffreykegler In Perl 5, you mean?
05:43 RandalSchwartz yeah
05:43 jdurand jeffreykegler: at every parameter declaration or declaration event, the identifier either obscures or enters a typedef definition
05:43 RandalSchwartz although it was more just a global
05:43 jeffreykegler OK, Marpa events are like that, except pain-free.
05:44 jeffreykegler jdurand: prediction events?  completion events?
05:44 RandalSchwartz you know, I'm gonna be thoroughly distracted by Marpa for at least the next month or two now. :)
05:45 jeffreykegler You mean it will no longer be our little secret?
05:45 jeffreykegler :-)
05:45 jdurand jeffrekegler: obscuring or entering a typedef is a completion event
05:46 jeffreykegler jdurand: thanks!  I could have looked at the code, but I'm being a bit lazy.
05:46 jdurand and identifier ambiguity is a prediction event, based on previous completion events that indeed obscured or entered typedefs -;
05:46 jdurand the same logic applies to enum, with one enormous exception: enums, ones defined, are not scope based
05:47 jeffreykegler Randal: getting all that?
05:47 RandalSchwartz sorta!
05:47 RandalSchwartz after all, I'm "not a member of the educated class"
05:47 jeffreykegler Randal: Jean-Damien, by the way, was very important in the development of the event mechanism ...
05:48 jeffreykegler many design suggestions, in addition to acting as a guinea pig, and being the 1st victim of all my bugs.
05:48 RandalSchwartz that's a reference to Tom Christiansen's description of me, as quoted in http://archive.wired.com/wired/archive/8.10/cruise_pr.html
05:48 RandalSchwartz scan that for "educated class".
05:48 RandalSchwartz and then read back a few lines.
05:48 jeffreykegler Christiansen went to school, I take it?
05:49 RandalSchwartz I would suspect so.
05:49 RandalSchwartz that comment actually made it into a footnote in Learning Perl. :)
05:50 RandalSchwartz something about some complex overloading, "but to understand this, you would have to be a member of the educated class" :)
05:50 jeffreykegler Cause he takes me to task for using big words.
05:50 RandalSchwartz I would suspect he takes everyone to task for something.
05:51 RandalSchwartz our public falling-out was inspired by some stupid things I did, but for which he took far more personally than I had expected.
05:51 jeffreykegler So I guess the theory is there's a reasonable middle ground, that avoids the two extremes. :-)
05:51 jdurand RandalSchwartz: the possibilities with Marpa are huge and great, I'll come back to C package soon providing a hopefully novel way to generate C bindings to any target language. Perl6 grammar was also on my list but with lower priority. Nevertheless that can change if people like you express a serious desire to have such thing.
05:51 RandalSchwartz I've offered the oak branch many times, but never accepted.
05:51 jeffreykegler That's too bad.
05:52 RandalSchwartz well - hopefully, I can get you resources to do it
05:52 RandalSchwartz and hopefully, the 50k downloads of floss weekly over the next week or two will help.
05:52 jeffreykegler You get 50K downloads?
05:52 RandalSchwartz that's what they say
05:53 RandalSchwartz primarily audio
05:53 RandalSchwartz but definitely quite a few video
05:53 jeffreykegler Any idea how many listen to a major portion of it?  how downloads translate to listeners?
05:53 RandalSchwartz FLOSS Weekly is the number one podcast in the open source space
05:53 RandalSchwartz no idea.
05:53 RandalSchwartz except that more people approach me at OSCON about floss weekly than about Perl now, if that's any indication.
05:54 jeffreykegler You'd suspect 1 to 1, or close, because I listen to most things I download.
05:54 RandalSchwartz I listen to at least the first few minutes of everything I download
05:54 jeffreykegler I hope it's an indication about FLOSS, instead of about Perl. :-)
05:55 RandalSchwartz I suspect it's because after 321 shows, I'm a pretty good interviewer.
05:55 RandalSchwartz and I get the most interesting projects on.
05:55 jeffreykegler And you have great taste in parsers!
05:55 * RandalSchwartz lists that on his linkedin
05:55 jeffreykegler So the response will be ramping up about now?
05:55 jeffreykegler And peak over the next week?
05:56 RandalSchwartz yeah... it'll likely peak soon
05:56 RandalSchwartz given past experience
05:56 RandalSchwartz it was a pretty geeky show
05:56 RandalSchwartz not something like pfsense
05:56 RandalSchwartz which was a lot more enduser friendly
05:56 jeffreykegler Hopefully it will pay off long-run -- people will go back to it.
05:57 RandalSchwartz yes... I am happy that my show is a magazine show, not a news show.
05:57 RandalSchwartz the back catalog is extremely useful.
05:57 * jeffreykegler is reminded he needs to go back and listen to the Lua FLOSS
05:59 RandalSchwartz ok - gotta go get on my flight
05:59 RandalSchwartz but read http://archive.wired.com/wired/archive/8.10/cruise_pr.html
05:59 jeffreykegler Enjoy!
05:59 jeffreykegler I will.
06:06 jdurand When there will be another talk on Marpa, I guess a great point would be how to deal with ambiguities with it
06:06 jdurand IMHO there are three main implementations:
06:07 * jeffreykegler is listening
06:07 jdurand a grammar that is obviously not ambiguous - the"perfect" world, then comes the read world:
06:07 jdurand the events, to take decisions are lexing phases
06:08 jdurand and the forest tree, extremely powerful and still underused: you can navigate in the parsee tree valueS and take decision on what is the correct value
06:08 jdurand -.
06:08 jdurand am I missing something?
06:08 jeffreykegler Nope, that sounds like the list.
06:08 jdurand "decisions at lexing phase"
06:08 jdurand Ok (ouff -.)
06:09 jeffreykegler ASF's were an effort to make traversing the parse forest easier.
06:09 jeffreykegler and I think evaluation in Kollos will be ASF based.
06:10 jdurand Yes. I believe that ASF is the correct answer to "native language", where the correct "value" is very much context-sensitive, sometimes depending on what is after, what is before - Jeffrey I do not remember, tell me: can we inspect what possible values are "after" in the value phase without yet taking a decision on them
06:11 jeffreykegler "native language" -- you mean NLP?
06:12 jdurand I mean english, french, etc.. Needless to say, in native languages, value can be writen valueS as well.
06:12 jeffreykegler OK, yes, NLP
06:12 jeffreykegler NLP= natural language processing
06:13 jeffreykegler Right now, you can enumerate trees of ambiguous parses, but ...
06:13 jeffreykegler that's not necessarily practical for NLP, where a sentence could be highly ambiguous.
06:14 jdurand ah ok, thx
06:14 jeffreykegler So the answer to your question is, in theory, "yes", but in practice there are serious efficiency issues,
06:15 jeffreykegler working around which was the motivation behind ASF's.
06:15 jeffreykegler With ASF's you could dive into the forest, starting at the top, and "prune" the forest as you went.
06:15 * jdurand is thinking to an amazing (yet) another project... IA with Marpa, where a grammar is a live object that evolves during its lifetime
06:16 jeffreykegler IA?
06:16 jdurand Hemm.... in french: "intelligence artificielle", i.e. the program is learning
06:17 jeffreykegler OK.  In English that's AI. :-)
06:17 jeffreykegler Which I actually have a degree in.
06:17 jdurand ARgh, this is AI in english, probably - yes! sorry
06:17 jdurand Ah ah.... I am not surprised
06:18 jdurand I believe that a BNF can be a live object. When decisions are taken, the program is informed about it and adapts its grammar
06:18 jeffreykegler Actually, with strand parsing, you'll be able to do that.
06:19 jdurand Ok... will have to read on studies on this subject.
06:20 jeffreykegler Which would make my use of DNA terminology appropriate.
06:20 jeffreykegler "studies" -- you mean on strand parsing?
06:20 jdurand Anyway, Marpa is so powerfull that it can easily make me diverge. I have to concentrate on finishing what is started.
06:20 jdurand Yes, I meant that
06:20 jeffreykegler Actually, strand parsing is something I'm in the process of inventing.
06:21 jeffreykegler For Kollos.
06:22 jeffreykegler It was originally motived by flaviu's question about the cost of keeping the whole parse in memory,
06:22 jeffreykegler and I'm coming up with this scheme whereby you can create partial parse forests, transcribe them and wind them together, kind of like DNA strands.
06:23 jeffreykegler But anyway, there's no reason the grammar cannot change, as long as you have a mapping between rules that allows you to wind strands together.
06:23 jdurand Googling on it, there is the "JBrowse" project, that matches "strand parsing" in the search engine
06:24 jdurand But ok... So I'll look forward _your_ studies -;
06:24 jeffreykegler It's my own coinage, so if they mean the same thing, it's an interesting coincidence.
06:25 ronsavage joined #marpa
06:25 jdurand OK -;
06:25 jeffreykegler Looked it up.
06:26 jeffreykegler What they mean is parsing strands of DNA, the actual stuff taken from living cells.
06:26 jeffreykegler Whereas I'm using the term to describe a mathematical technique, and just using the analogy/metaphor.
06:28 jeffreykegler It's late CA time.  Good night, all.
06:29 jdurand Thkx - good night
07:10 rns joined #marpa
07:17 rns jdurand: re http://irclog.perlgeek.de/marpa/2015-01-16#i_9956823 -- I saw the same "* THE MARPA SEMANTICS PRODUCED A FATAL ERROR ..." msg in pattern.t under windows (perl 5.20.1, Marpa::R2 2.102) and cygwin (perl 5.14.2, same Marpa); doesn't break the install for me though, so no cpantester report. Should I look into it?
07:20 jdurand rns: this failure is a predicted one and the test is ok. It failed upper in t/jquery.t
07:23 rns Yes, in cpantester report. t/jquery-1.10.2.t.
07:24 rns BTW, the report says Parse errors: Bad plan.  You planned 2 tests but ran 1.
07:24 rns So, perhaps no_plan and/or done_testing() would help?
07:25 jdurand Ah... from http://www.cpantesters.org/cpan/report/545a870c-98fd-11e4-b221-9e126cbd7f71 I see something else: Out of memory! followed by # Looks like you planned 2 tests but ran 1.
07:25 jdurand Nice point.. many thanks
07:26 jdurand I now believe that Test::More does not handle well the case of "Out Of memory" -; !
07:26 rns :)
07:27 rns But blames you for running less tests than planned. :)
07:28 jdurand That's it. The number of planned tests is ok. But if "out of memory", it said the test did not ran, while ahem... it ran but failed too early for him isn't !?
07:29 jdurand Anyway... this is understood - big thanks - something I can drop out from my todo list (yessss -;) !
07:30 rns BTW, report says they have Test::More 1.001014 (latest version -- https://metacpan.org/pod/Test::More) and I run 1.001009 on both cygwin and windows.
07:30 rns I saw talks about latest Test::More breaking things.
07:31 jdurand Ah... interesting
07:32 rns Yes, but anyway, out of memory -- out of *their* memory -- so they possibly just don't have enough, lol. :)
07:34 jdurand yep he he! Glad you helped me - thx!
07:34 rns Absolutely welcome.
07:34 rns P.S. report without 'out of memory' -- http://www.cpantesters.org/cpan/report/545a870c-98fd-11e4-b221-9e126cbd7f71
07:35 rns Sometimes they have just enough, he he ...
07:36 jdurand I see the "Out of memory!" in this link (?)
07:38 rns Ooops, my bad, it was above the marpa message so I missed it, sorry and forget that link.
07:39 jdurand ok
07:44 rns Oops again -- found that "no out of memory report" -- http://www.cpantesters.org/cpan/report/76ad6a9a-99f2-11e4-b221-9e126cbd7f71
07:45 rns just say "Failed 1/2 subtests" after juery test.
07:46 jdurand Oh... it was probably missing the memory so much that it could not even print out "Ouf of memory"... puzzled
07:50 rns puzzled too. it's from the same tester 'root at klanker', but 5.18.4-only.
07:52 rns jquery-1.10.2.t is the same as jquery-1.10.2.min.t, just without minification? I mean jquery-1.10.2.min.t runs ok.
07:53 jdurand Exactly. It just consumes more memory because of all the spaces stuff.
07:53 jdurand May I conclude with high probablity this is the same problem of "Out Of Memory". All other tests are OK on all platforms, including openbsd with other perl versions.
07:56 rns Well yes, I think. BTW, do you discard spaces in your grammar?
07:58 jdurand Yes, I do
07:59 rns Just as I expected. :)
08:05 clacke joined #marpa
08:07 koo5 joined #marpa
10:46 ilbot3 joined #marpa
10:46 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
11:19 basiliscos joined #marpa
12:14 koo5 joined #marpa
14:22 basiliscos joined #marpa
14:37 jdurand rns: thx for the pull request: Re https://github.com/jddurand/MarpaX-Languages-ECMAScript-AST/pull/6/files#diff-8230e4058df6b205c5367648dcd8b0b0L1015 for example, you mean that this is consuming less memory from Marpa point of view ?
15:20 lwa joined #marpa
18:41 flaviu joined #marpa
19:40 jeffreykegler joined #marpa
20:41 jeffreykegler I updated my notes on Andrew's FLOSS talk, with responses to 2 questions that did not get answered: https://gist.github.com/jeffreykegler/3563fd82932663837f01
20:42 jeffreykegler At 27:00, I answer (sort of) "What is the biggest application, by volume, on Marpa to date?"
20:43 jeffreykegler AT 41::11, I answer "How does Marpa handle something like C typedefs?", basically summarizing what Jean-Damien told me on this channel last night (CA time).
20:53 jdurand jeffreykegler: ok. I'll ask at my work if we use TIvoli btw, in the hope to see it in action with Marpa, who knows
21:29 ronsavage joined #marpa
21:48 ronsavage jeffreykegler: Re http://irclog.perlgeek.de/marpa/2015-01-17#i_9957936. At 22:10 you've mis-typed 'powered', and at 44:00, I got 3 JSON grammars, 2 by you.
21:50 jeffreykegler ronsavage: Fixed.  Thanks.
22:40 basiliscos joined #marpa

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