Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-01-29

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

All times shown according to UTC.

Time Nick Message
23:23 jeffreykegler joined #marpa
23:24 jeffreykegler jdurand: re http://irclog.perlgeek.de/​marpa/2014-01-28#i_8188893 -- thanks!
00:10 ronsavage Marpa::R2 V 2.079013
00:10 ronsavage Counts: Tests: 542. Modules: 8. Passes: 8. Fails: 0
00:10 ronsavage Duration: 1 minute and 34 seconds
00:10 ronsavage Also, this is just after switching to Perl V 5.18.2
00:15 jeffreykegler ronsavage: Thanks!
01:28 jeffreykegler1 joined #marpa
01:33 jeffreykegler joined #marpa
05:12 ronsavage joined #marpa
06:42 jeffreykegler left #marpa
07:49 sivoais joined #marpa
16:51 jeffreykegler joined #marpa
19:13 jdurand joined #marpa
19:14 jdurand Jeffrey, is there a possbility to bypass the fact that \$input must be a reference to a SCALAR ? I am thinking to tied variable, and in particular IO::String
19:14 jdurand or IO::File
19:15 jeffreykegler jdurand: not really
19:16 jeffreykegler Also, tied variables create a lot of difficulties inside the XS code.
19:16 jeffreykegler Basically, for all the magic you want, I have to write the C code to make sure it happens.
19:17 jdurand Let's go betong tied notion. A nice notion would be to have another option in the recce giving a callback method that must return the character at the position required by the SLR
19:17 jdurand "beyond" - sorry switching between swiss and french keyboards -;
19:19 jeffreykegler Marpa is going to need a scalar string.  It does not at this point handle streams.
19:19 jeffreykegler I'm guessing some kind of streaming behavior is what you're after.
19:20 jdurand yes - the fact that an input data of hundreds of MB has to stored in memory is not practical
19:21 jeffreykegler If you divide the world into SAX-ish parsers and DOM-ish parsers, Marpa is DOM-ish
19:21 jdurand Ok - I have SAX-ised a bit my other grammar but then I raised another issue, perhaps easier:
19:22 jdurand would it possible that Marpa does not raise an exception if the parse end but this is not the end of the input. Something optional.
19:22 jdurand use case is:
19:22 jdurand the input is to be parsed bw two consecutive grammars. And of course the user does not know where the first grammar will end.
19:23 * jeffreykegler is listening
19:23 * jdurand heu... explanation done
19:23 * jdurand diner time, my wife says - will be back soon
19:26 jeffreykegler judrand: assuming you're out of the room, but will be backlogging ...
19:27 jeffreykegler 1.) you can always catch the exception.
19:29 jeffreykegler 2.) Perhaps better, you can use an event -- for example, a completion event on the rule you're looking for.
19:41 jdurand_ joined #marpa
19:43 jdurand_ Jeffrey, yes I have done an implementation with eval
19:44 jeffreykegler I've toyed with the idea of adding parse exhaustion as another event.
19:45 jdurand_ and was thinking to event tonight when driving (!) but event is not ok as soon as the start rule is in the form this+ -
19:45 jdurand_ this+
19:45 jdurand_ where this = a | b |  etc...
19:46 jeffreykegler There's also a call which checks for parse exhaustion, so that you can catch completion events and continue if not exhausted.
19:46 jdurand_ Jeffrey - yes, parse exhaustion would be a nice event in general - not just because this is my current use case
19:46 jdurand_ ah... this call is ?
19:47 jeffreykegler OK, maybe it's just a NAIF call.  It should be a SLIF call as well.  Could you file an issue?
19:47 jdurand_ exhausted() in NAIF
19:48 jdurand_ Sure - that will be an added value to SLIF
19:49 jeffreykegler And if you could write a test/use case, that'd be very helpful.  I'll get to it after Phase 3.
19:50 jeffreykegler Have to go.  AFK.
19:50 jdurand_ ok - thx
19:50 jeffreykegler left #marpa
20:08 jdurand_ re - trigger an event on start completion seems ok a priori for my workflow -;
20:09 jdurand_ ok, I mya tell you what I've written - this is the callgrind grammar - why? Because I'd love to see callgrind supported by nytprof - kcachegrind is very pretty but not so easy to use

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