[12:57:46] morning all [13:11:20] Morning Ironholds [13:12:24] * YuviPanda waves [13:14:41] hey halfak, YuviPanda :) [13:14:52] * YuviPanda has been consumed with ops work the last few weeks [13:15:06] Have you officially made your switch? [13:16:00] halfak: no, not until november [13:17:01] Gotcha. [13:17:29] YuviPanda, where in the world are you this week? [13:18:10] halfak: London [13:18:39] Wut. I didn't know you were still in the UK. Any chance you'll make your move before Oct. 6th? [13:19:33] * halfak almost has the feature extractors for a text revision scorer completed. [13:19:57] halfak: I'm in the UK till sep 30 [13:20:04] halfak: and won't be in the US until Dec / Jan [13:20:18] I just added LANG_LONG_TOK and LANG_CHAR_REP. [13:20:20] Gotcha [13:20:32] Ref. http://repository.upenn.edu/cgi/viewcontent.cgi?article=1515&context=cis_papers for definitions [13:26:43] Ironholds, any thoughts on that intertime stuff? [13:28:08] not at the moment, solely because my brain is currently fried. I've got a call with Shilad/Heather/Brent in ~30, then I'll be thinkin' on it. [13:28:28] but basically the thoughts are: I need a walkthrough of how the methodology here works, I think [13:29:01] kk sounds good. I'll drop something on our calendars. We can do a quick screensharing session. [13:30:31] cool! [13:55:03] * halfak just pushed up 7 more features. I'm almost feature complete with A.G.West's classifier. [13:55:13] I'll be skipping the WikiTrust features though. [13:55:17] We'll see what we get. [13:57:33] It would be nice to spool up a WikiTrust instance in labs though. Seems like A.G.West could help us with that. [13:57:36] YuviPanda, ^ [13:57:51] oh sure [13:57:57] I can help out later today [13:58:02] is WikiTrust open source, btw? [13:58:06] It is. [13:58:09] But it is OCaml [13:58:10] cool [13:58:12] hah! [13:58:17] good challenge, I suppose? :) [13:58:19] link, halfak? [13:58:20] Yup [13:58:34] I wrote a compiler in ML so I should be half competent at OCaml [13:58:44] hah! :) [13:58:56] I suppose setting it up shouldn't be too much of a problem [13:59:02] halfak: do you already have a project for this? [13:59:05] or should we create a new one? [13:59:15] "project" as in repo? [13:59:20] no [13:59:23] project as in labs project [13:59:23] Or project as in labs-project [13:59:28] Gotcha. I don't. [13:59:32] Setting one up would be great. [13:59:35] * YuviPanda is labsadmin, and can create projects now [13:59:39] :D [13:59:52] halfak: what should we call it? [13:59:57] (we are going to create so much cool shit in labs) [13:59:58] we should use this for your scoring stuff as well [14:00:19] Oh... um... Hmm. I've been calling the repo "Revision-Scoring" and the python module "revscores". [14:00:34] But we could go broadly and call it "AI as a service" [14:00:35] revscoring? [14:00:39] too broad :P [14:00:42] aaas [14:00:45] hahaha [14:00:53] revscoring sounds good to me [14:02:31] halfak: done and added you as projectamin [14:02:45] * halfak flexes muscles. [14:02:51] Cool. Thanks, dude. [14:02:57] yw [14:03:03] https://github.com/halfak/Revision-Scoring [14:03:05] FYI [14:03:09] halfak: if you're doing things to it, log it with !log revscoring on the labs channel [14:03:25] * halfak has no idea how such logs work. [14:03:34] halfak: they write it out on a page in wikitech [14:03:43] makes it useful for debugging / reconstructing / documenting [14:04:00] halfak: also if you're setting up the ocaml setup, document your actions in an etherpad so I can turn it into a puppet module later on [14:04:25] So if I go "!log revscoring spooled up a new instance to test out WikiTrust (called wikitrust)", would that be appropriate? [14:05:28] halfak: indeed [14:07:16] OK. Will do. [14:07:24] halfak: Ironholds also looks like newer R and python coming to stat1002 soon :) [14:07:33] cool :) [14:07:37] (from ops@) [14:34:00] YuviPanda, can you but a second set of eyes on this: https://gist.github.com/halfak/638a70f78046c0770a0b [14:34:25] I'm trying to build a unique index and I get a duplicate key error. However, the key doesn't actually duplicate! [14:35:29] Anything obvious? [18:50:17] halfak, http://blog.ironholds.org/joining-between-databases-in-hive/ btw - the aliases problem. [18:53:34] :P Looks like standard SQL to me. [18:53:54] I must be missing something. [18:55:03] halfak, the fact that it's necessary [18:55:44] unless mysql requires aliases for inter-database joins too and I've just not encountered it :/ [18:55:56] which is plausible [18:57:29] Hmm... It doesn't require aliases, but it does require the use of .. [18:58:22] gotcha [18:58:36] yeah, the distinction is if you try .. in hive, hive will freak the eff out [18:58:45] that is, if you do it in the JOIN/ON statement [18:59:09] you need to set an alias for each . and then ON . = . [18:59:29] Oh! I see. You *need* the alias. [18:59:33] yep [18:59:38] thass the distinction! [18:59:52] (bringing it up now because I mentioned it in the meeting but wasn't sure how clear I was being. Evidently; not very! :D) [19:00:18] so ON database1.table1.column1 = database2.table2.colum2 won't work? [19:05:14] nope [19:05:27] it starts getting confused and going "I can't find a column called table1, watchu talkin' bout willis?" [19:06:13] Ahh. Lame hive. Lame. [21:01:38] Ironholds: running late [21:03:26] kk [21:07:06] probably 10 more mins [21:25:21] DarTar, time: you're bad at it ;p [21:36:45] hey [21:36:57] yeah, they were non Euclidean minutes [21:37:11] Ironholds: ready now, if you’re still around [21:37:22] DarTar, yeah, I've been in the hangout for 20 mins ;p [21:37:28] aarg [21:37:29] ok [21:38:10] DarTar can bend time!? Now this I'd like to have seen. [21:38:37] you have to believe, halfak [21:41:46] Oh I'm definitely a believe in non-euclidean space-time. All that gravity does fun things to space. :)