Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2015-05-26

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

All times shown according to UTC.

Time Nick Message
01:48 ilbot3 joined #marpa
01:48 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
03:21 ronsavage joined #marpa
03:45 Aria \o/ I got leo optimization working in lotsawa!
03:47 jeffreykegler Now that I think about it, is this just the 3rd time ever?
03:47 jeffreykegler Leo's 1991 paper reports an experimental implementation of some kind
03:48 jeffreykegler (I've been in touch with Leo about it, and apparently it's not worth looking at, if it could be found now)
03:48 jeffreykegler And of course there is Marpa
03:49 Aria Could be.
03:49 Aria I may still have some bugs, but my test parses work.
03:50 jeffreykegler It doesn't seem like that hard an algorithm, but the Leo modification is very hard on the intuition
03:50 Aria Hmm. I don't find it so now, but it was really hard to grasp from the literature.
03:51 jeffreykegler There are lots of test examples in the Marpa test suite.
03:51 Aria I'm gonna have to start porting those.
03:52 jeffreykegler Do you try to handle cycles?
03:52 jeffreykegler A ::= B ; B ::= A -- that sort of thing
03:52 Aria Not short cycles like that. I bet that'd do funky things.
03:53 jeffreykegler Marpa handles them OK -- if there are productive, it rewrites them into just the productive part.
03:54 jeffreykegler Anyway, you can just write your own direct and indirect right recursions and see what happens
03:54 Aria Yeah. Direct works well.
03:56 Aria Indirect, not so much.
03:56 Aria (Falls back to non-leo.)
03:56 Aria Looks like I should work on that.
03:57 jeffreykegler Yes, indirect is actually important.
03:57 Aria Probably by explicit linking of the leo items. Rather than by symbols.
03:58 jeffreykegler My t/sl_leo2.t might be a useable test (its grammar that is)
03:59 jeffreykegler A lot of my tests are wierd or special-purpose in some way, and don't focus on right recursion by itself
04:00 jeffreykegler Also t/leo_unit.t
04:00 Aria Sounds good to me.
04:00 jeffreykegler and t/leo3.t
04:02 jeffreykegler May debug_leo.t, but that look easy and focused on testing the debug messages
04:02 jeffreykegler (By the way I *do* test my debug messages, and I am glad I do, because they have more regression issues than any other feature.)
04:03 jeffreykegler With regard to indirect --
04:04 jeffreykegler I wound up precomputing which symbols are in which right recursive series.
04:04 jeffreykegler "right recursive series" meaning a bunch of symbols which mutually right recurisve.
04:04 Aria Yep.
04:05 jeffreykegler s/recurisve/recurse/
04:06 Aria Aha. I know exactly where I have to tweak it.
04:07 jeffreykegler Fortunately "x right recurses to y" is transitive, symmetric and reflexive, so they're an equivalence class.
04:07 Aria Yess.
04:07 Aria It would be a pain in the ass if not.
04:08 jeffreykegler Aria: by the way if you've ever studied math & philosophy, mathematical counter-factuals like the you expressed are very strange beasts.
04:08 Aria Hehe. Indeed they are.
04:08 Aria (I have indeed ... let's just say Hofstadter was a regular appearance on my bookshelf as a teenager)
04:09 jeffreykegler That is, take any theory of language and semantics and test it against http://irclog.perlgeek.de/marpa/2015-05-26#i_10655104
04:09 jeffreykegler Most of them train-wreck on it.
04:10 jeffreykegler It is very hard to describe why I understand http://irclog.perlgeek.de/marpa/2015-05-26#i_10655104, and if I do, exactly what it is that I understand by it.
04:10 jeffreykegler Because by the usual account of mathematics and language it means I am modeling a world where the truth is not true.
04:12 * Aria grins.
04:12 jeffreykegler So anyway, I can now report to Leo that implementations of his 1991 are becoming thick on the ground. :-)
04:12 Aria Yay!
04:16 jeffreykegler Good night, all
15:09 jeffreykegler joined #marpa
15:23 koo6 joined #marpa
18:02 lwa joined #marpa
22:20 ronsavage joined #marpa
22:24 jeffreykegler joined #marpa

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