Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2015-10-09

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

All times shown according to UTC.

Time Nick Message
01:47 ilbot3 joined #marpa
01:47 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
05:22 ronsavage joined #marpa
07:43 sivoais joined #marpa
07:55 sivoais joined #marpa
08:05 sivoais joined #marpa
08:15 sivoais joined #marpa
08:25 lwa joined #marpa
08:26 sivoais joined #marpa
08:36 sivoais joined #marpa
08:46 sivoais joined #marpa
08:56 sivoais joined #marpa
09:06 sivoais joined #marpa
09:10 lwa FYI: Yet Another Earley Parser: “This is the fastest implementation of Earley parser which I know”, “YEAP without scanner is up to 20 times faster Marpa and requires up to 200 times less memory” https://github.com/vnmakarov/yaep
09:16 sivoais joined #marpa
09:27 sivoais joined #marpa
09:37 sivoais joined #marpa
09:47 sivoais joined #marpa
09:58 sivoais joined #marpa
10:08 sivoais joined #marpa
10:18 sivoais joined #marpa
10:29 sivoais joined #marpa
10:39 sivoais joined #marpa
10:50 sivoais joined #marpa
11:00 sivoais joined #marpa
11:11 sivoais joined #marpa
11:21 sivoais joined #marpa
11:31 sivoais joined #marpa
11:42 sivoais joined #marpa
11:52 sivoais joined #marpa
12:03 sivoais joined #marpa
12:13 sivoais joined #marpa
12:25 sivoais joined #marpa
12:36 sivoais joined #marpa
12:46 sivoais joined #marpa
12:56 sivoais joined #marpa
12:58 pczarn joined #marpa
13:03 pczarn Marpa could use lookahead for eventless positions, too.
13:06 sivoais joined #marpa
13:12 pczarn YAEP was written for a programming language. Does it construct parse forests? Either way, memory consumption this small can't be achieved with source links.
13:16 sivoais joined #marpa
13:17 lwa pczarn: according to the readme, it can construct all parse trees: “the parse result can be an abstract tree or all possible abstract trees”. From reading the source, it seems to have a "yaep_alt" node for the parse trees when the parse is ambiguous.
13:26 pczarn I don't understand the implementation, but results seem great.
13:26 sivoais joined #marpa
13:36 sivoais joined #marpa
13:47 sivoais joined #marpa
13:57 sivoais joined #marpa
14:07 sivoais joined #marpa
14:17 sivoais joined #marpa
14:21 MerittClarksdale joined #marpa
14:27 sivoais joined #marpa
14:34 MerittClarksdale hi
14:37 sivoais joined #marpa
14:47 sivoais joined #marpa
14:48 aredridel joined #marpa
14:58 sivoais joined #marpa
14:58 aredridel joined #marpa
15:00 Idiosyncrat joined #marpa
15:04 Idiosyncrat https://github.com/vnmakarov/yaep
15:05 Idiosyncrat Implements a traditional Earley algorithms (no Marpa, no Leo, no SLIF) in pure C.
15:05 Idiosyncrat Lots faster than Marpa
15:05 aredridel Ooh, neat.
15:06 Idiosyncrat He didn't run a comparison with Libmarpa, AFAICT, just did the C vs. Perl thing.
15:10 sivoais joined #marpa
15:19 lwa Idiosyncrat: no, the benchmark uses libmarpa directly. Unfortunately, the code is hidden in a shell script: https://github.com/vnmakarov/yaep/blob/master/src/compare_parsers.tst.in#L12096
15:20 sivoais joined #marpa
15:25 Idiosyncrat Hmmmm
15:28 Idiosyncrat A lot of the Libmarpa overhead at this point is due to tracking events, and other added features.
15:30 Idiosyncrat It would seem that Vladimir's optimization on the stripped down Earley version may have paid off.
15:30 sivoais joined #marpa
15:31 Idiosyncrat Folks are going to want features like events and the abiliity to handle right recursion without going quadratic, is my guess.
15:38 Idiosyncrat If someone has the time to investigate Vladimir's work, please report back.
15:38 Idiosyncrat I'll be interested to hear.
15:41 sivoais joined #marpa
15:41 Idiosyncrat I note that YAEP does have an interface language, but it is not self-parsed -- if I read the code correctly, he uses YACC to parse it.
15:50 Idiosyncrat Still skimming ...
15:51 Idiosyncrat Vladimir's optimization sound very interesting.
15:51 sivoais joined #marpa
15:51 Idiosyncrat I'd speculated that things of the sort he talks about might have a real payoff.
15:52 Idiosyncrat If Vladimir's numbers are right (and I can't find anything wrong with them so far), he's found a good set.
15:52 Idiosyncrat "set" == "set of optimizations"
16:01 sivoais joined #marpa
16:11 sivoais joined #marpa
16:22 sivoais joined #marpa
16:23 Idiosyncrat Oops -- http://irclog.perlgeek.de/marpa/2015-10-09#i_11346190
16:23 Idiosyncrat Missed the backlog -- pczarn and lwa were already discussing YAEP
16:32 sivoais joined #marpa
16:42 sivoais joined #marpa
16:52 sivoais joined #marpa
17:03 sivoais joined #marpa
17:04 Pursuit joined #marpa
17:13 sivoais joined #marpa
17:25 sivoais joined #marpa
17:34 pczarn joined #marpa
17:35 sivoais joined #marpa
17:45 sivoais joined #marpa
17:55 sivoais joined #marpa
18:05 sivoais joined #marpa
18:15 sivoais joined #marpa
18:25 sivoais joined #marpa
18:36 sivoais joined #marpa
18:46 sivoais joined #marpa
18:56 sivoais joined #marpa
19:06 sivoais joined #marpa
19:16 sivoais joined #marpa
19:27 sivoais joined #marpa
19:32 Idiosyncrat Re YAEP -- perhaps once my writeup of Leo's algorithm is out, we'll see more implementations of it.
19:32 Idiosyncrat That's what I am occupied with right now.
19:37 sivoais joined #marpa
19:47 sivoais joined #marpa
19:57 MerittClarksdale hi
19:57 sivoais joined #marpa
19:59 MerittClarksdale I am maintaining an IBM z/os mainframe at the moment, it does have COBOL code with JCL and JES control. I wonder what parser IBM used. The compiler may be in assembler
19:59 MerittClarksdale Marpa would probably run on this thing
20:02 sivoais joined #marpa
21:40 Idiosyncrat MerritClarksdale: COBOL came out before the first parser paper was published IIRC
21:41 Idiosyncrat Traditionally, it is parsed using line-by-line hacks, the way folks typically parse config files.
21:43 Idiosyncrat Debian made an attempt to port to z/os -- I think that's what is meant by the s390x architecture
21:49 Idiosyncrat I can't find any mention of it now on the Debian Marpa pages: https://packages.debian.org/jessie/libmarpa-r2-perl
21:49 Idiosyncrat Don't know if omission means that s390x is no longer supported, or just that they didn't get the port working.
21:50 Idiosyncrat A major obstacle for Debian was that nobody had access to an s390x box for testing.
21:51 Idiosyncrat I'm sentimental about s390x -- when I was in grad school that was the OS most computers ran.
23:08 ronsavage joined #marpa

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