Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2016-04-14

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

All times shown according to UTC.

Time Nick Message
00:00 idiosyncrat_ If we do find whoever suggested an AATM, I'll give him credit for the suggestion in the future.  (Note AATM is my name, the OP just wanted the SLIF lexer to recognizer "all" tokens).
01:33 idiosyncrat_ === Methods for input, G1 spans ===
01:34 idiosyncrat_ I'm now working with the methods for finding substrings of the input using G1 and L0 (input) locations, and I realize I am free to change them to something sensible.
01:35 idiosyncrat_ In Marpa::R2 I created them gradually, and some of the early choices of name proved unfortunate.
01:35 idiosyncrat_ One thing I will do is have every method which uses or returns G1 locations, have g1 in its name.
01:38 idiosyncrat_ I also have better start a new Changes POD document, documenting the changes between Marpa::R2 and Marpa::R3
01:39 idiosyncrat_ s/have better/had better/
01:50 ronsavage joined #marpa
03:25 rns joined #marpa
03:56 ronsavage joined #marpa
04:07 rns idiosyncrat: re http://irclog.perlgeek.de/m​arpa/2016-04-14#i_12333717 -- the closest thing I can remember is this -- https://youtu.be/XaScLywH2CI?t=1187 -- moment in hobbs talk about Marpa where he talked about passing all (overlapping also) tokens to a Marpa parser and letting it to figure them out.
04:08 rns In one slide he did it with one character at a time in another (Brute force lexer) with regexes (lexer only, without use of alternative()/earleme_complete().
04:13 idiosyncrat_ Yes, I think that's it ... I had in mind someone else who asked for this feature on (I think) IRC, but this talk would preceed that
04:14 idiosyncrat_ hobbs: dang, you get credit for another one
04:14 rns re http://irclog.perlgeek.de/m​arpa/2016-04-13#i_12333555 -- yes, it is this repo -- https://github.com/rns/kollos-luif-doc (click on manual.md for then-current version) and issues to it.
04:15 rns Except those specific to D2L and Lua.
04:16 idiosyncrat_ Yes, we'll back off some of that, but that's good ...
04:17 idiosyncrat_ I'd tried to remember my final opinion on how to handle grouping vs. hiding symbols. :-)
04:18 idiosyncrat_ And looks like I did settle on making parens '()' group without hiding, and square brackets '[]' group *with* hiding.
04:19 idiosyncrat_ And the other thing is the use of a full set of repetition specifiers, based on those in Perl 6
04:20 idiosyncrat_ Oh gee, we did a full syntax in BNF, and everything.  Cool.  It wasn't that long ago, but I'd forgotten.
04:20 rns you also mentioned seamless interface -- http://irclog.perlgeek.de/m​arpa/2015-08-22#i_11099277 -- on this channel
04:22 idiosyncrat_ Actaully the version of sl_randal.t I just pushed is seamless -- you can do that in the SLIF already, though your strings had better not be very long, because it's not really efficient.
04:23 idiosyncrat_ In the IRC discussion you referenced, I'm talking about using "strand parsing" to deal with the efficiency ...
04:23 idiosyncrat_ but that won't happen in Marpa::R3.
04:24 idiosyncrat_ But in Marpa::R3 a lot of things where I said in effect YAGNI, we can now say YAGNI or not, if you want it you can have it.
04:25 idiosyncrat_ By the way, I rarely actually say YAGNI or anything equivalent, because I think it's presumptuous to think you know what the user is going to need.
04:27 idiosyncrat_ We're down to 18 "rework" tests.
04:28 rns Great, let me know if where can help.
04:28 rns BTW the recently push sl_randal.t shows how seamless can be done in SLIF -- G1-only rules and doing strings in single character literals (<die k0> ::= 'd' 'i' 'e') with concatenate action as needed.
04:29 idiosyncrat_ I've started the Changes doc -- https://github.com/jeffreykegler/Marpa​--R3/blob/master/cpan/pod/Changes.pod
04:29 idiosyncrat_ We've already got a lot of them.
04:30 idiosyncrat_ re http://irclog.perlgeek.de/m​arpa/2016-04-14#i_12334484 -- and note especially the handling of whitespace, which is the trickiest part, and requires a lot of details
04:31 idiosyncrat_ The basic strategy for whitespace is, with each rule, handle only *interior* whitespace ...
04:31 idiosyncrat_ except at the top, where you have to deal with any leading or trailing whitespace you intend to allow.
04:32 idiosyncrat_ There are several details to watch, such as with separated sequences ...
04:33 idiosyncrat_ The separator can have optional whitespace before and after, and that will deal with whitespace around separators, but ...
04:33 idiosyncrat_ separation must be proper, not Perl-style -- Perl-style allows a final "separator", which is really a terminator, and that means your rule has exterior whitespace.
04:34 rns Oops, missed that <opt ws> stuff -- very cool -- reading it now.
04:34 idiosyncrat_ The reason you have to watch all this is that if there is more than one way to soak up white space, you introduce ambiguity ...
04:34 idiosyncrat_ and under the scheme I just described, it's always unambiguous where any particular piece of whitespace goes.
04:35 idiosyncrat_ Yes, the <opt ws> was the tricky part -- I've been thinking about this for years now.
04:37 idiosyncrat_ I'll look at the rework tests to see if there are any others I want to ask you to take a look at.
04:38 VsyachePuz joined #marpa
04:38 idiosyncrat_ ah2.t those tests are probably best folded into sl_ah2.t, if you could take that on
04:39 rns Ok, I'll see what I can do.
04:39 idiosyncrat_ That is, I trying to say, ah2.t and sl_ah2.t are the same grammar, but ah2.t has diagnostics that sl_ah2.t does not.
04:40 rns Well, how about sl_ah2_diag.t?
04:40 rns I can take at least initial conversion.
04:41 idiosyncrat_ Maybe best just fold the tests into sl_ah2.t I think -- it's just more diagnostics for the same grammar.
04:41 idiosyncrat_ bocage.t is also more diagnostics for that same grammar.
04:41 idiosyncrat_ I'm thinking just add all the diagnostics to sl_ah2.t
04:42 idiosyncrat_ By the way, ah2 is the grammar from the Aycock-Horspool paper, and was the first one I got working.
04:42 hobbs idiosyncrat_: I inspire people by having no clue what I'm doing :-P
04:43 rns hobbs: :))
04:44 rns idiosyncrat: re ah2.t -- it has show_isys which is missing from Scanless::G
04:44 idiosyncrat_ With show_isys, do the todo_skip thing again.
04:45 idiosyncrat_ One final one, if you're willing, chaf.t.
04:45 idiosyncrat_ Which I think fits the paradigm for the tests in sl_gia.t and can be added to that as one more example.
04:47 idiosyncrat_ 10 PM California time, time to say
04:47 idiosyncrat_ Good night!
04:47 hobbs idiosyncrat_: and yeah, I actually asked about AATM while preparing my talk
04:47 idiosyncrat_ rns: Many, many thanks!
04:47 hobbs #marpa_20140619.log:[14:28:51] <hobbs> I'm reading through some of the improvements in Marpa in 2014, and updating my talk, and I have to wonder, could the machinery for LATM also support "AATM" (the brute-force lexing technique from my 2013 talk)?
04:47 idiosyncrat_ hobbs: AATM
04:48 hobbs that question was asked poolside in Orlando :)
04:48 idiosyncrat_ Did you use that acronym?
04:48 hobbs yes
04:49 idiosyncrat_ Was the above a reference to this IRC channel?
04:49 hobbs yes
04:50 hobbs http://irclog.perlgeek.de/​marpa/2014-06-19#i_8901979
04:50 hobbs through to 19:44
04:51 hobbs damn, that doesn't feel like two years ago :)
04:51 idiosyncrat_ I was about to say, I didn't realize this IRC channel was that old.
04:52 idiosyncrat_ In answer to your question, no the AATM does not just pop out of the machinery for LATM -- it requires support in the Libmarpa ...
04:53 idiosyncrat_ which it already has, but I have to do stuff like making sure Libmarpa's idea of earlemes matched the upper layers idea of input location ...
04:53 hobbs well, I'm kind of glad the question paid off in the end, even if it was nontrivial :)
04:54 idiosyncrat_ so that all the C pointers dereference somewhere nice :-)
04:54 idiosyncrat_ Yes, basically in that part of the 2013 talk, I just glazed over.
04:55 hobbs I didn't pace things super well for that talk, I put a bunch of code on slides that I could only briefly wave my hands at
04:55 hobbs I figured anyone really interested could go back and look at it online :)
04:55 idiosyncrat_ Marpa is a hard thing to present
04:58 idiosyncrat_ Anyway, it is time for me to settle in for the night.
04:58 idiosyncrat_ Good night folks!
04:58 hobbs night
05:00 rns Good night
06:09 hobbified joined #marpa
07:04 beaugunderson joined #marpa
09:34 VsyachePuz joined #marpa
10:11 rns joined #marpa
10:48 kaare_ joined #marpa
10:59 kaare_ joined #marpa
11:32 Pursuit joined #marpa
11:39 btyler joined #marpa
11:44 btyler joined #marpa
11:46 btyler joined #marpa
12:13 aredridel joined #marpa
13:18 btyler joined #marpa
13:48 kaare_ joined #marpa
14:08 JPGainsborough joined #marpa
14:26 kaare__ joined #marpa
16:15 Idiosyncrat joined #marpa
16:53 Idiosyncrat joined #marpa
17:52 JPGainsborough joined #marpa
17:56 teatime joined #marpa
18:55 koo7 joined #marpa
19:00 idiosyncrat_ joined #marpa
19:31 choroba joined #marpa
20:21 ernimril joined #marpa
22:02 JPGainsborough joined #marpa
22:33 ronsavage joined #marpa
23:30 idiosyncrat_ joined #marpa

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