[07:26:29] *** Joins: drbean_ (~drbean@TC210-63-209-187.static.apol.com.tw) [07:27:38] *** Quits: drbean (~drbean@TC210-63-209-198.static.apol.com.tw) (Ping timeout: 258 seconds) [07:51:19] *** Joins: drbean (~drbean@TC210-63-209-62.static.apol.com.tw) [07:51:47] *** Quits: drbean_ (~drbean@TC210-63-209-187.static.apol.com.tw) (Ping timeout: 258 seconds) [09:50:03] *** Joins: laurettem (92401c5a@puddleglum.dhcp.meraka.csir.co.za) [09:59:22] Hello! I have a question: is there a difference in how the compiler deals with "my_morph : Feature => Str = table ..." and "my_morph : Feature -> Str = \f -> ..." Does it come down to the same thing? I have many potential tables that would be very sparsely populated, so I'm wondering if using opers instead would help at all. [10:07:46] hello! [10:09:22] Github Pages (with your blog) is down... I was hoping your piece on compiling to PMCFG would help me. :P [10:09:56] tables are functions from a finite input (i.e. the param type) to some output, and opers are functions from a potentially infinite input to some output. that's the biggest reason why one can't use functions that do something on the arguments to a lin [10:10:18] I actually don't know how its handled when an oper has param as its only argument [10:10:27] does your solution compile? [10:10:57] are you using my_morph only in lexicon, or also in syntactic funs/lins? [10:11:03] Yes, it works. I'm just not sure if I'm actually gaining anything and if I should continue in this way. [10:11:32] this works for me https://inariksit.github.io/gf/2018/06/13/pmcfg.html [10:11:41] Also in funs and lins. In fact, especially there, since I'm treating morphology as syntax and using BIND to build my verbs. [10:11:49] okay [10:11:58] you can try too compile your grammar with -v 3 [10:12:08] Huh. I get "GitHub Pages is temporarily down for maintenance." Thanks though! [10:12:15] or use gftest -g MyGrammar.pgf --show-cats -d [10:12:22] I can paste the text [10:12:49] That would be great, thanks! I'll try gftest as well. [10:12:50] https://github.com/inariksit/inariksit.github.io/blob/master/_posts/2018-06-13-pmcfg.markdown [10:13:19] https://github.com/inariksit/inariksit.github.io/blob/master/_posts/2018-09-22-grammar-blowup.markdown#verbosity [10:13:51] you don't even need to know what those numbers are exactly if you compile with -v=3, just check if they are bigger in one of the versions :-D [10:14:01] Thank you so much!! Exactly what I need. [10:15:13] cool! [10:15:42] let me know what you find, I could add it to some post :-P [10:16:12] I will! [10:16:34] I'm just writing python since october, I feel like my GF is rusting :-D [10:17:01] btw, are you guys interested in coming to the summer school in singapore? [10:19:48] Very much, but funding is problematic for my institution at the moment. [10:20:17] we're trying to sort out travel grants [10:21:17] I'll keep an eye out! [10:21:37] great! I'll post to gf-dev as soon as we have something [10:44:35] For what it's worth, tables and opers of this kind give the same output on -v=3. Thanks! [10:46:51] good to know! [10:48:52] you can also try gftest -g MyGrammar.pgf --show-funs -d [10:56:27] *** Quits: laurettem (92401c5a@puddleglum.dhcp.meraka.csir.co.za) (Ping timeout: 260 seconds) [11:19:16] *** Joins: laurettem (92401c5a@puddleglum.dhcp.meraka.csir.co.za) [11:50:25] laurettem: I'm asking aarne right now [11:50:27] your question [11:50:54] Thanks! [11:52:17] does "sparsely populated" mean that many of the entries are the same? [11:52:37] Yes and no: many entries are empty. [11:52:43] unique strings are the same in the PGF, so that's not a problem [11:54:20] I still struggle to get my head around that, though. Ito PGF size (filesize) I understand that that's a win. Does it present problems at runtime? I can imagine resolution is tricky with large tables mostly containing empty strings. [11:54:44] "a win" - same strings are cheap to store [11:55:18] same strings are cheap to store in any case [11:55:42] if you have table {_ => "foo"}, there's only one "foo" in the PGF [11:56:03] ask this on gf-dev, I bet krasimir has the best current understanding of these things! [11:56:38] Okay, I'll think a bit more about what exactly I'm asking and then send it to gf-dev. Thanks! [11:58:39] also: oper foo : Number => Str = \n -> case n of {Sg => "sg" ; Pl => "pl"} is expanded into a table [11:58:59] table {Sg => "sg" ; Pl => "pl"} ! n [11:59:21] the whole thing: [11:59:22] [11:59:24] oper foo : Number => Str = \n -> table {Sg => "sg" ; Pl => "pl"} ! n [12:01:30] Do you perhaps mean: [12:01:35] oper foo : Number -> Str = \n -> case n of {Sg => "sg" ; Pl => "pl"} [12:01:42] sorry, yes I mean that [12:02:09] Okay. I think that actually answers my question! At least as far as I've thought about it. [12:02:22] alright! [12:15:47] Inari, can I ask you a different question? What are your thoughts (or best practice, if applicable) for putting strings into polarity and tense categories? I'm toying with the idea making those morphemes live there instead of in massive tables on the Cl level. [12:17:26] Oh, I see my "thanks" for the previous question never actually sent. So: thanks! [13:32:02] *** Quits: laurettem (92401c5a@puddleglum.dhcp.meraka.csir.co.za) (Remote host closed the connection) [13:49:37] *** Joins: laurettem (92401c5a@puddleglum.dhcp.meraka.csir.co.za) [13:49:54] seems completely fine! [13:50:51] it's even closer to the implicit rule of GF that every argument needs to leave a string to the linearisation, otherwise you get question marks when parsing [13:51:13] this thing https://github.com/inariksit/inariksit.github.io/blob/master/_posts/2018-08-28-gf-gotchas.markdown#metavariables-or-those-question-marks-that-appear-when-parsing [13:55:14] Yes! Thanks also for the blog post - I know those question marks well, and I was actually going to bug you about prodrop next. :P [13:56:56] haha, I'm glad I wrote those posts :-D [13:59:14] Haha, me too! While I'm here, do you know of somewhere I can look at ways in which optional prodrop has been handled? [14:15:15] *** Quits: laurettem (92401c5a@puddleglum.dhcp.meraka.csir.co.za) (Remote host closed the connection) [14:16:29] optional prodrop: https://groups.google.com/forum/#!topic/gf-dev/RSB0PwSiCHU [14:17:25] *** Joins: laurettem (92401c5a@puddleglum.dhcp.meraka.csir.co.za) [14:17:28] optional prodrop: https://groups.google.com/forum/#!topic/gf-dev/RSB0PwSiCHU [14:22:04] Thanks, I thought I had searched the mailing list at some point... Evidently not! [14:48:48] *** Quits: laurettem (92401c5a@puddleglum.dhcp.meraka.csir.co.za) (Remote host closed the connection)