Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2015-06-25

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

All times shown according to UTC.

Time Nick Message
00:04 LLamaRider ok, will dig in more offline, thanks for the help!
00:25 Aria jeffreykegler: I am indeed a rust fan.
00:26 jeffreykegler Probably LLamaRider is backlogging.  Do you know the answer to his question?
00:27 Aria I don't. I hadn't looked, though considered rust a good target for marpa
00:27 Aria nice to see someone else poking it.
00:27 Aria I'll have to investigate when I've free time.
00:28 jeffreykegler It'd be nice to see Marpa projects get beyond being matters of a "lone fanatic" :-)
00:28 jeffreykegler I've always believed fanatics are better in groups. :-)
01:46 Aria Fanatics in groups are much better!
04:13 ronsavage1 joined #marpa
04:24 jeffreykegler joined #marpa
04:25 jeffreykegler I'm still in search of a Lua code formatter (like perltidy for Perl) -- the one I found proved to be buggy.
06:00 ronsavage1 Perhaps if I can get your Lua's grammar into a nice tree, we could write one :-).
06:29 ronsavage joined #marpa
06:46 ronsavage In MarpaX::Languages::Lua::Parser I've used my module Data::RenderAsTree to turn the output of Marpa into a Tree::DAG_Node-based tree. This eliminated sub decode_value() and all subs and variables used by it. Simple!
09:31 ronsavage joined #marpa
09:42 koo5 joined #marpa
10:15 ronsavage I've gotten the Lua parser output and logging options working nicely. Next: docs.
14:03 RichardMichaels hi
15:50 jeffreykegler joined #marpa
17:09 RichardMichaels hi
17:22 RichardMichaels How is Kollos coming?
17:23 jeffreykegler slowly :-)
17:24 jeffreykegler It's a big project -- will take a few months.
17:24 jeffreykegler Right now I'm working on the code that does grammar "hygene" -- making sure all the rules are obeyed, such as no duplicate rules, no unproductive symbols, etc., etc.
17:25 jeffreykegler Someone who is considering helping out, can consider learning Lua.
17:26 jeffreykegler This will be useful for other reasons -- of course, it will be the semantics language for Marpa once Kollos is complete.
17:26 jeffreykegler But Lua is a very useful and interesting language on its own.
17:27 jeffreykegler It's beautfully designed -- it's one of the few technologies where hitting an obstacle is to some extent enriching ...
17:28 jeffreykegler because with Lua there almost always is a deep reason for that obstacle.
17:28 jeffreykegler As opposed to the usual "for historical reasons", or being just a matter of arbitrary choice.
17:29 RichardMichaels i have heard of Lua, I have not looked closely at it. I have used Perl rather extensively
17:30 jeffreykegler Lua can be thought of a C extension library, which happens to include a small Turing-complete language ...
17:30 RichardMichaels Sounds like TCL
17:30 jeffreykegler It is (literally, I think) unique, in being the only language with minimalism as a main goal.
17:30 jeffreykegler You could also think of it as TCL done right.
17:31 * jeffreykegler wished he'd said "done better" instead of "done right" -- Ousterhout is a friend.
17:36 RichardMichaels Yes I have just heard of him
17:38 jeffreykegler We come from the same area of the US and an ancestor of mine and his (so the legend goes) were close friends, but on opposite sides of the Revolutionary War.
17:38 jeffreykegler It was clear that by the time it was over, one or the other would have to emigrate, so they made a deal.
17:39 jeffreykegler Whoever got to stay would buy the other's farm.
17:39 jeffreykegler My ancestor wound up in Canada.
17:40 jeffreykegler But I'm American because his son didn't share his Dad's loyalism.
17:41 jeffreykegler The son stayed in New York, and retired on a pension provided for veterans of the Revolution -- the pension documentation survives, and is the only authenticated part of this story.
17:42 jeffreykegler The parent's deal is not described in it, and may be just a great yarn.
17:46 RichardMichaels fascinating
17:49 jeffreykegler Also, TCL's goals were a bit different.
17:49 jeffreykegler Ousterhout was heavily focused on TCL being a good CLI, and scripting was secondary.
17:50 jeffreykegler Lua is a scripting language, and does not seem to be intended to be used as a CLI
17:54 Aria The tension between CLI and programming language is really interesting to me, and I think still under-researched.
17:55 jeffreykegler Aria: good point
17:56 jeffreykegler It does seem hard to design a language that is good at both, and I am not clear as to why.
17:56 jeffreykegler It's a bit of a mystery
17:56 Aria Same. I think parsing strategy has something to do with it.
17:56 jeffreykegler ?
17:56 Aria Or tokenizing, at least. CLIs tend to have unusually simple tokenizing rules.
17:57 Aria And barewords-are-strings vs barewords-are-symbols, too, gramatically.
17:57 Aria CLIs tend to resolve those late, and contextually. Programming languages tend to resolve those early, if it even allows barewords to be strings. (*hem* perl.)
17:59 jeffreykegler Natural language seems to work for both -- quick interactions like this, which can be thought of of NLP CLI.
18:00 jeffreykegler But also extended presentations, like articles and blog posts, which can be thought of as "scripts"
18:03 Aria In theory. But look at the trouble we have specifying precise logic in natural language.
18:04 Aria A tangled mess of references. Reading, for example, legal code... ow.
18:07 RichardMichaels Thats why TCL looks a lot like a scripting language to me :-)
18:34 koo5 what do blog posts and legal code have to do with shells and programming?
18:35 koo5 Aria, forget the term "natural language", the basis is *controlled* natural language
18:37 * koo5 is bored
18:44 jeffreykegler NLP is the central inspiration behind Marpa -- not its central goal, but its inspiration
18:45 jeffreykegler In order to put the years into inquiring for better parsing algorithms, I had somehow to convince myself that they existed -- almost everyone as of 10 years ago believed that the parsing problem was solved ...
18:45 jeffreykegler we had gone as far as we could go.
18:45 koo5 Programming or Processing?
18:46 jeffreykegler But I looked at what humans do in NLP, and I supposed that our language processing was based on something more powerful than the parsers in the textbooks.
18:47 lwa joined #marpa
18:47 jeffreykegler Which meant that what I was searching for actually existed.
18:49 jeffreykegler This is not to make any statement about how useful Marpa will eventually be for NLP -- on that I am agnostic.
22:34 ronsavage joined #marpa
23:57 jeffreykegler joined #marpa

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