[00:00:02] (03CR) 10Awight: Store content quality as an integer index (031 comment) [extensions/JADE] - 10https://gerrit.wikimedia.org/r/476994 (owner: 10Awight) [00:08:02] (03CR) 10jenkins-bot: Store content quality as an integer index [extensions/JADE] - 10https://gerrit.wikimedia.org/r/476994 (owner: 10Awight) [00:17:44] awight: Does it make sense to block JADE reviews on the open questions about DB schemas? [00:18:53] hoo: AFAICT, marostegui gave it a good drive-by [00:19:40] * awight tries to compile DBA comments [00:22:15] hoo: https://phabricator.wikimedia.org/T200297#4834319 [00:22:31] It ends with an encouraging "Other than a possible misbehaviour of the optimizer, they look ok to me." [00:22:34] :) [00:22:40] I read that as "hell yes" [00:25:08] awight: Sounds good… but what about the covering index? AFAICT this is not fully needed in that query [00:28:27] looking [00:29:51] It does seem to be used [00:30:39] We filter on jader_revision (has its own index) and fetch the jader_judgment, which is indexed by revision_id and available in the covering index. I believe that saves us a full row lookup. [00:31:20] hoo: ^ [00:31:48] oops wrong table but the point should stand [00:31:56] If we're talking about the query https://phabricator.wikimedia.org/P7927 [00:32:22] So (jaded_revision, jaded_judgment, jaded_damaging, jaded_goodfaith) still has two "superfluous" fields [00:32:51] The filter could have also included goodfaith, is what I'm thinking. [00:33:40] True that… ok, I'll maybe play with it some more and but I don't see anything blocking it still :) [00:33:46] jaded_judgment would also be included if we want to fetch the linked judgment page [00:34:12] and the covering index would save an intermediate jade_diff_judgment lookup when joining revision to page as judgment_page [00:34:25] true that this query doesn't light up all those columns, though... [00:34:40] hoo: Right on, thanks for making the time. [00:45:20] halAFK: sorry, that took longer than i expected. Big fan of the built-in diff view. I think this is a good iteration; it's worth having this one and the other one you made to show to users. If you're up to it, it may be worth mocking up some other things so that we can see how the different parts connect together. [00:47:54] halAFK: intellectual challenge for you, may not be possible to accomplish practically: what is the least number of page elements we can get away with? [01:16:45] 10ORES, 10Scoring-platform-team, 10translatewiki.net, 10Security: New ORES model relies on translatewiki.net API, which is not hosted on WMF production - https://phabricator.wikimedia.org/T213131 (10awight) Here's a possible variation on the "new ORES cluster" proposal. If @Nikerabbit and translatewiki.ne... [01:18:20] 10ORES, 10Scoring-platform-team, 10translatewiki.net, 10Security: New ORES model relies on translatewiki.net API, which is not hosted on WMF production - https://phabricator.wikimedia.org/T213131 (10awight) [06:23:58] (03CR) 10Krinkle: Introduce ext.ores.api (031 comment) [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [06:43:17] (03PS8) 10Krinkle: Introduce ext.ores.api [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [06:43:19] (03PS1) 10Krinkle: [WIP] Single function for ext.ores.api [extensions/ORES] - 10https://gerrit.wikimedia.org/r/483338 [06:45:00] (03CR) 10Krinkle: [C: 03+2] build: Update 'stylelint-config-wikimedia' to v0.5.0 [extensions/ORES] - 10https://gerrit.wikimedia.org/r/483292 (owner: 10VolkerE) [06:46:52] (03CR) 10jerkins-bot: [V: 04-1] Introduce ext.ores.api [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [06:47:04] (03CR) 10jenkins-bot: build: Update 'stylelint-config-wikimedia' to v0.5.0 [extensions/ORES] - 10https://gerrit.wikimedia.org/r/483292 (owner: 10VolkerE) [06:47:39] (03PS9) 10Krinkle: Introduce ext.ores.api [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [06:47:41] (03PS2) 10Krinkle: ext.ores.api: Single function [extensions/ORES] - 10https://gerrit.wikimedia.org/r/483338 [06:51:22] (03CR) 10jerkins-bot: [V: 04-1] Introduce ext.ores.api [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [06:52:01] (03CR) 10jerkins-bot: [V: 04-1] Introduce ext.ores.api [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [09:11:13] o/ [09:25:42] He's back! [09:26:36] Having fun with Oozie at the moment... https://gerrit.wikimedia.org/r/#/c/analytics/refinery/+/482753/ [09:32:50] haha :))) [09:33:06] I haven't done anything with oozie :( [09:40:54] whew, that last patchset... [09:40:56] https://gerrit.wikimedia.org/r/#/c/analytics/refinery/+/482753/6/oozie/ores/revision/recent_scores/import_recent_ores_revision_scores.hql [09:41:03] line 30 for a good time [09:41:39] I explode and union only to split out again into two tables [09:41:46] saved a full scan, lost my wits. [09:41:55] 'night [10:21:20] (03CR) 10Ladsgroup: [C: 03+2] ext.ores.api: Single function [extensions/ORES] - 10https://gerrit.wikimedia.org/r/483338 (owner: 10Krinkle) [10:27:51] 10ORES, 10Scoring-platform-team, 10Operations, 10Release Pipeline (Blubber), 10Release-Engineering-Team (Backlog): The continuous release pipeline should support more than one service per repo - https://phabricator.wikimedia.org/T210267 (10Ladsgroup) >>! In T210267#4860670, @Ottomata wrote: > Q: would bl... [11:40:59] 10ORES, 10Scoring-platform-team, 10translatewiki.net, 10Security: New ORES model relies on translatewiki.net API, which is not hosted on WMF production - https://phabricator.wikimedia.org/T213131 (10Joe) >>! In T213131#4868178, @awight wrote: > Here's a possible variation on the "new ORES cluster" proposal... [11:58:55] 10ORES, 10Scoring-platform-team, 10translatewiki.net, 10Security: New ORES model relies on translatewiki.net API, which is not hosted on WMF production - https://phabricator.wikimedia.org/T213131 (10Joe) Another detail I kinda assumed was a given, but it's better to reiterate it: We'd need ORES to send to... [12:01:42] 10ORES, 10Scoring-platform-team, 10translatewiki.net, 10Security: New ORES model relies on translatewiki.net API, which is not hosted on WMF production - https://phabricator.wikimedia.org/T213131 (10Ladsgroup) In general I agree that we should not have a wildcard proxy. Given that ores uses pickle a lot, I... [13:11:37] 10ORES, 10Scoring-platform-team, 10translatewiki.net, 10Security: New ORES model relies on translatewiki.net API, which is not hosted on WMF production - https://phabricator.wikimedia.org/T213131 (10akosiaris) For what is worth, let me say that both approaches of a dedicated ORES installation (one perhaps... [13:20:10] I'm gone for the day, I don't feel alright [13:20:11] o/ [15:57:26] 10Scoring-platform-team, 10DBA, 10MediaWiki-Database, 10Blocked-on-schema-change, and 2 others: Schema change for rc_this_oldid index - https://phabricator.wikimedia.org/T202167 (10Marostegui) [16:05:00] Hi everyone, happy new year! I was wondering if someone could help me with two things concerning metrics combinations in ORES queries and the Recent Changes filters. [16:12:57] o/ ThresholdT [16:13:04] What are you hoping to do? [16:13:37] 1) Referring to queries like "https://ores.wikimedia.org/v3/scores/enwiki/?models=damaging&model_info=statistics.thresholds.true.%27maximum%20recall%20@%20precision%20%3C=%200.9%27", is there some kind of documentation on which combinations of metrics are useful and which are nonsensical? [16:15:14] Oh. Um. I think that depends on your purpose. [16:15:27] What would be an example of a nonsensical combination? [16:15:45] Maybe minimum recall @ precision >= 0.9. :) [16:16:03] But hey, if you want to minimize recall, maybe you have some purpose for that. [16:16:25] Like you're trying to find the highest empirical probability. [16:18:21] I was thinking that combining precision and !precision for example would not make a lot of sense at first [16:19:04] ThresholdT, we could make a table of metric pairs. [16:19:35] But I'm wondering if this is a real problem. I.e. under what circumstances do you expect someone to be choosing pairs that don't make sense? [16:20:31] Oh, no, I didn't mean it as in it's a problem. It's more for documentation purposes that I'm trying to find out which combinations are the most useful [16:22:07] Ahh. I see. We've approached this from a case study perspective. [16:22:29] E.g. The worked example in https://www.mediawiki.org/wiki/ORES/Thresholds [16:23:05] It would be better to have a set of examples rather than just the one. [16:32:50] Sorry I had a connection problem. The other thing I wanted to ask: [16:32:58] 2) I know about the Recent Changes quality prediction filters being a helpful tool in varying the precision and recall of catching damaging edits and I was wondering if there is a documentation on the exact numbers concerning threshold, precision, recall and other involved metrics for the filters from "Very likely good" to "Very likely have problems". [16:35:25] ThresholdT, I might not have gotten the 1) that goes with that 2). [16:35:33] Oh wait. I see it above :) [16:37:03] https://en.wikipedia.org/wiki/Special:ORESModels [16:37:07] ThresholdT, ^ [16:37:19] :) [16:37:29] Oh, perfect! [16:37:30] Looks like the threshold is getting rounded for "User intent" [16:39:16] * awight puts on coal lamp and dives into https://gerrit.wikimedia.org/r/#/c/analytics/refinery/+/482753/ [16:40:01] And those are the only metrics playing a role in the filters? [16:44:41] ThresholdT, Not sure exactly what you mean. The threshold optimizations that they target are based on precision and recall. [16:45:32] Ok, yes, I think I understand now [16:45:57] Thank you! :) [16:49:25] Happy to help. :) [17:19:57] 10Scoring-platform-team, 10DBA, 10MediaWiki-Database, 10Blocked-on-schema-change, and 2 others: Schema change for rc_this_oldid index - https://phabricator.wikimedia.org/T202167 (10Marostegui) 05Open→03Resolved This is all done [17:20:00] 10Scoring-platform-team (Current), 10MediaWiki-Database, 10MW-1.32-notes (WMF-deploy-2018-08-21 (1.32.0-wmf.18)), 10Patch-For-Review, and 3 others: Use index on rc_this_oldid - https://phabricator.wikimedia.org/T139012 (10Marostegui) [19:37:41] (03CR) 10Krinkle: "@Ladsgroup: Looks like you may have to backport a few of the fixes from my next commit for the tests to pass (a few very minor eslint warn" [extensions/ORES] - 10https://gerrit.wikimedia.org/r/459549 (https://phabricator.wikimedia.org/T201691) (owner: 10Ladsgroup) [20:07:13] 10MediaWiki-extensions-ORES, 10Scoring-platform-team (Current), 10User-Ladsgroup: Use rc_timestamp index when joining to ores_classification - https://phabricator.wikimedia.org/T138444 (10Ladsgroup) [20:07:23] 10Scoring-platform-team (Current), 10MediaWiki-Database, 10MW-1.32-notes (WMF-deploy-2018-08-21 (1.32.0-wmf.18)), 10Patch-For-Review, and 3 others: Use index on rc_this_oldid - https://phabricator.wikimedia.org/T139012 (10Ladsgroup) 05Open→03Resolved [20:16:50] PROBLEM - puppet on ORES-redis02.experimental is CRITICAL: CRITICAL: Catalog fetch fail. Either compilation failed or puppetmaster has issues [20:42:50] RECOVERY - puppet on ORES-redis02.experimental is OK: OK: Puppet is currently enabled, last run 37 seconds ago with 0 failures [21:16:03] Feeling a bit funky. I'm going to go AFK for a bit to lay down. Will report back in 60 minutes. [22:04:51] back [22:04:58] Working on wireframes again [22:23:12] harej, see https://docs.google.com/drawings/d/1nHuCsDbYCAbGM0F-RdfGjAAx_mA2nddXOerOLg5P4-Y/edit [22:23:26] That's the single column one, but more fleshed out. [22:24:02] (03CR) 10jenkins-bot: Localisation updates from https://translatewiki.net. [extensions/JADE] - 10https://gerrit.wikimedia.org/r/483571 (owner: 10L10n-bot) [22:24:23] The buttons are too eye catching. They are defeating the purpose of the wireframe so I'm going to deal with that. [22:25:12] 10ORES, 10Scoring-platform-team (Current), 10Analytics: Backfill ORES Hadoop scores with historical data - https://phabricator.wikimedia.org/T209737 (10awight) [22:25:18] Solves for the duplicate data field problem. [22:25:19] Yeah; wireframes should focus on the structural relationship between interface elements [22:25:34] Balsamiq mockups look terrible, but that's on purpose [22:27:14] Right. I've been considering that. I only wanted to use a bit of color to communicate "Imagine the use of colors here" [22:27:20] But it ended up being distracting :| [22:27:49] It's amusing how google draw thinks "EpochFail" is misspelt but "awight" is not. [22:28:04] alight, awight [22:28:06] https://www.urbandictionary.com/define.php?term=aw%27ight [22:28:56] 10ORES, 10Scoring-platform-team (Current), 10Analytics, 10Patch-For-Review: Wire ORES scoring events into Hadoop - https://phabricator.wikimedia.org/T209732 (10awight) I've left the working draft of the two coordinator jobs running in my user database. The hardcoded datacenter is probably a blocker, any s... [22:29:42] ! I wasn't awawe of that usage [22:29:42] 04Error: Command “i” not recognized. Please review and correct what you’ve written. [22:29:59] * awight pats AsimovBot on the pattern-matcher [22:30:13] halfak: one idea i have is that there's a switch on the one-column diff view that lets you switch to a two-column diff, at which point the diff takes up the full width and the other elements are pushed down. I presume this wouldn't be a feature on mobile. Also, in general we need to think about how to make this thing play nice with MobileFrontend, at least to the extent that MF doesn't fuck with us too much. [22:31:04] Right. Agreed on both points. I'm certain people will have strong opinions. [22:31:09] It might be possible to tell MobileFrontend "don't worry about Jade, we got this" and have Jade be a screen-responsive interface from the start. [22:31:40] Jon Robson is sitting behind me. I'll ask him. [22:32:03] We'll want full width for displaying a page rendering for revision judgments too. [22:32:19] Maybe we can see this as a hybrid of (table) and (two column) [22:32:31] I'll work that out quick,. [22:34:13] I asked Jon and he said (a) make sure that when loading ResourceLoader modules to target both desktop *and* mobile (only desktop is targeted by default) and (b) some RL modules are intentionally not loaded on mobile so we need to keep that in mind [22:34:22] Other than those things we shouldn't have too many issues [22:41:03] Cool. I'm sure prateek will have some ideas for how to make sure things float sanely too. We'll need to ask him. In the meantime, I'll make sure we have things to point at./ [22:42:51] awight ^ for your information [22:47:33] :) I've seen this ResourceLoader paradigm so can help with that. [22:47:56] Other good news--I have rough drafts of both of our ORES score import jobs running in Oozie (hadoop). [22:48:10] There are streams of data backfilled since Dec 2018 [22:48:29] Lots of additional bits to work through, but this is a good foundation. [22:52:05] Great! We'll have to remember that this was one of our ultra-secret-adam-needs-codereview-and-we're-listening-to-our-users goals. [22:52:12] :) [22:52:26] harej, I added a second view to https://docs.google.com/drawings/d/1nHuCsDbYCAbGM0F-RdfGjAAx_mA2nddXOerOLg5P4-Y/edit [22:52:37] you did! it's a good explication of the idea i had [22:52:40] it includes a full size diff and two columns of judgement/schema/things. [22:52:45] Nice [22:53:27] o/ hoo [22:53:30] Just saw your email. [22:53:35] Will you be around for a bit? [22:56:21] yes [22:57:03] halfak: ^ [22:58:13] Great! Want to pair program some features for wikidata? [22:59:30] halfak: Sure: 9 [22:59:32] I have 30 minutes right now :) [22:59:36] Sure :) [22:59:48] Can we do it on your machine, I screwed my setup :S [23:00:39] :) [23:00:47] Yeah. Just messaged you on hangouts [23:00:53] THrough your WMDE account [23:01:00] one sec [23:04:32] 10Scoring-platform-team, 10Analytics, 10Release-Engineering-Team: Investigate formal test framework for Oozie jobs - https://phabricator.wikimedia.org/T213496 (10awight) [23:09:04] 10Scoring-platform-team, 10Analytics, 10Release-Engineering-Team: Investigate formal test framework for Oozie jobs - https://phabricator.wikimedia.org/T213496 (10awight) One alternative is a JUnit class meant for workflow and coordinator testing: http://oozie.apache.org/docs/5.1.0/ENG_MiniOozie.html An exa... [23:42:01] hoo: https://github.com/wikimedia/articlequality/pull/72 [23:44:25] wikimedia/articlequality#123 (wikidatawiki_completeness - c69b64f : halfak): The build passed. https://travis-ci.org/wikimedia/articlequality/builds/478098966 [23:44:34] \o/ PROGRESS [23:44:39] And with that, I'm out. [23:44:44] Time to pedal [23:56:04] o/