[13:02:31] [[Tech]]; Elitre (WMF); +message; https://meta.wikimedia.org/w/index.php?diff=24460710&oldid=24439555&rcid=26199003 [14:02:29] [[Tech]]; Elitre (WMF); /* Upcoming event */; https://meta.wikimedia.org/w/index.php?diff=24460918&oldid=24460710&rcid=26199336 [20:22:40] Hello, I am Pradnya Shimpi, a sophomore at IIT Roorkee. I want to contribute to Wikimedia, how can I get some issues assigned? [20:38:16] Pradnya: Generally we don't assign people tasks, they can claim them themselves though [20:53:13] Reedy oh okay thank you, I'll check out some issues [20:58:22] I'm doing my final year university project on automatically recommending code reviewers. As part of this I'm looking at making mediawiki the example open source project that I would be basing this on. I know about https://wikimedia.biterg.io but I'm wondering whether I can get a copy of public data that it contains. I want to avoid over-querying the service and having it locally would help speed everything up for me. [20:58:47] If I need to contact someone at the WMF for this, is there someone I could go ask? [20:59:48] When I last found data dumps of phabricator and gerrit, these were from 2015. Ideally I'd want to have a up-to-date dataset. [21:03:29] Pradnya: Do you have any particular areas of interest that you would like to contribute to? [21:04:27] phuzion was looking forward to contribute in MediaWiki and pywikibot [21:04:39] Dreamy_Jazz: I don't know for sure, but my go to person to ask would probably be aklapper [21:04:42] Dreamy_Jazz: AKlapper would probably be your best bet... If you email him :) [21:04:48] Thanks both [21:04:48] (Also, cool!) [21:04:52] JINX! [21:05:16] Certainly be able to help facilitate if he can't help directly [21:06:49] Thanks. I will email. [21:06:55] you might also get an answer on https://www.mediawiki.org/wiki/Talk:Community_metrics [21:09:41] Pradnya: I'd suggest checking out Phabricator to see where you might be able to help. I'd also recommend signing up for a Developer account. https://www.mediawiki.org/wiki/Developer_account [21:23:46] Dreamy_Jazz: do you know about the gerrit reviewer-bot already? [21:23:58] it does that (automatically assign reviewers) [21:24:20] (and we turned it off because it was hated right?) [21:24:45] but it's not it's some magic AI that learns or something.. it just has a list of repos mapped to humans.. and tbh I think that's the best option there is [21:25:15] I never really got the point of it when gerrit has a built in feature that does basically the same thing in your settings [21:25:17] bd808: did not know that. that's sad. as if getting reviewers would not already be too hard [21:26:05] Yes I'm aware (have added myself to it for the CheckUser extension). My project would be based around automatically making a list of people who could be added. Importantly this wasn't from a list defined somewhere where people add themselves. [21:26:31] i.e. a user would be prompted with people who are best suited for a particular change [21:26:52] mutante: I think I mixed things up. You're talking about the bot for https://www.mediawiki.org/wiki/Git/Reviewers and I was thinking of the gerrit plugin that picked a reviewer based on blame history. [21:26:53] Dreamy_Jazz: how would it guess? based on who did reviews in the past? [21:27:10] bd808: ah! ACK, yea [21:27:20] Some kind of machine learning algorithm. [21:27:51] s/machine learning algorithm/if-else statements/ ;) [21:28:05] I generally think the problem with getting reviewers on mediawiki is not really a matching problem [21:28:06] silicon does not learn [21:28:10] there is this effect that people are considered an owner of something just because they touched it last. this sometimes make people NOT want to review.. or give an incentive to NOT touch things.. because you know you will get punished for it by being asked again and again untilt he end of time [21:28:25] More a cultural problem in how we recognize reviewers and treat the work of reviewing [21:28:32] bawolff: I'm pretty sure its a risk vs reward problem [21:28:55] yea, for MW specifically I also think the reason is not a matching problem but simply.. there are no people [21:28:57] Probably some reinforcement learning (used ridge regression for something similar already), but wouldn't be primarily based round who touched what [21:29:07] If I +2 a crap patch I get the blame. If I +2 a perfect patch nobody cares. All risk, no reward [21:29:17] indeed [21:29:40] But i think there's also elements of unclear expectations around what level of responsibility you are taking when you +2 something [21:30:44] And ambiguity breeds fear [21:31:26] Obviously my project will not involve trying it out on mediawiki (as that would likely need ethical approval based on the checklist given to me) but is more of a proof of concept. Thanks all for the advice. [21:32:23] So my evaluation would be performing matching analysis on who actually reviewed the change over who was recommended to review it. [21:37:38] Dreamy_Jazz: I would treat "review" and "merge/ submit" as different [21:37:47] Of course [21:37:51] :) [21:38:13] A +1 is useful, but at the end of the day it doesn't merge the patch.... [21:38:40] technically you can also +2 but not submit it..it's just rare [21:38:43] How else would we know if they have a working mouse and keyboard! [21:39:08] :) [21:39:20] At previous $JOB i had, they separated review and submit. Someone else had to review it, but you always pressed merge on your own patches [21:39:42] I kind of liked it, as it re-enforced that you have primary responsibility to make sure your patches are not broken, and the reviewer is just a second check [21:40:30] Yup. My internship was like that. [21:40:34] I guess in operations/* repos it's like that [21:41:25] I do think part of the problem is our review model is very modeled around the idea of working within a team [21:41:34] You work on something, and your team-mate reviews it [21:41:41] a +1 is not usually useless to me though. It is nice to get +1 and then you submit after that. it's not the same as self-merge to me at all [21:41:42] open source volunteers generally don't work that way [21:42:59] I think +1 is mostly useless in mediawiki code due to various political factors, where 50% of +1's are where someone is mad that their patch isn't getting merged, so they convince their friend who isn't qualified to review it, to sign up for an account just to +1 it [21:43:12] Or random community members are +1'ing something because they really want the bug fixed [21:44:10] I could imagine its really different in an operations workflow [21:45:12] Yeah. The one thing about operations/* is that often times it comes with other responsibilities (such as restarting a service) that are best done by the user submitting the patch if they have +2. [21:55:34] bawolff: ACK, I think this is just different between repos. I am not on mw/core, other politics for operations// repos [23:20:45] well, I'm quite the opposite [23:21:00] really wary of +1/+2 patches I haven't looked carefully [23:22:43] plus, I would usually not have enough time to "properly" test things [23:24:49] Dreamy_Jazz: I don't really think there would be ethical issues with applying it to mediawiki [23:25:13] (as long as you don't expect those tagged by your system to "obey" it) [23:25:56] but your University may use a more strict definition of what it considers acceptable, ofc [23:34:34] Obligatory https://bash.toolforge.org/quip/AU7VVae36snAnmqnK_xL [23:35:34] I feel like Chad was the original person to say that. Am I misremembering? [23:39:18] Oh fun fact! I have a patch which could do with some +1/+2s >:D [23:40:11] well, you can try advertising it [23:40:36] rumor is, it might attract some -1/-2 as well :D [23:41:02] See I made someone leave.. /j [23:41:18] https://gerrit.wikimedia.org/r/c/mediawiki/core/+/762102, ba.wolff knowws [23:50:56] it says what it means on Gerrit [23:51:12] and I think the words are actually meaningful in that case [23:51:23] "looks good to me BUT somebody else needs to approve" [23:51:27] Dreamy_Jazz: in case it's useful, all the gerrit review data is stored *in git*. A git clone --bare of a repo should pull down everything, then looking at git for-each-ref refs/changes/*/*/meta the "meta" ref stores review data. [23:51:40] which is what it often really is [23:53:11] thcipriani: Even things like abandoned / yet to be merged changes? [23:53:29] Sidenote, I am so looking forward to being able to use the GitLab VS Code integrations once everything moves there 😌 [23:54:07] Dreamy_Jazz: yeah, as far as I understand it, gerrit has no database, so everything on screen is stored in these magic refs [23:54:25] did you guys notice this? https://wikitia.com/wiki/Main_Page [23:54:46] it's like a fork with only "verified editors" whatever that means [23:54:55] TheresNoTime: But how can we transition to gitlab if we haven't first transitioned to diffusion :P [23:55:28] can we stop using github now? [23:55:29] mutante: I see a certain pattern in the articles on their main page [23:55:31] we don't talk about diffusion bawolff :D [23:55:48] zabe: thank you for the review <3 [23:55:51] As in its all weird obscure companies [23:56:11] * bawolff actually likes gerrit now. The stolkhom syndrome has fully set in [23:56:12] bawolff: SEO and Crypto ?:o [23:56:27] Does anything even canonically use diffusion? [23:56:43] as of a few minutes ago...only one thing :) [23:56:43] And real-estate [23:56:44] we just deleted one of the last remaining things that does [23:56:53] :D [23:56:54] TheresNoTime: An obscure website called facebook [23:57:02] please help me get the steward scripts elsewhere now :) [23:57:20] bawolff: please tell me they don't.. [23:57:26] last I checked twitter still used it, too. But who knows what the status is there now. [23:57:44] Twitter only really has one git now so.. [23:57:51] I will try to install "Phorge" in wmcs [23:57:54] the Phab fork [23:58:13] but then it would only be about replacing Maniphest..not the other Phab apps [23:58:34] the naming in phab is half the problem [23:58:39] well.. Maniphest and the Pastebin :p [23:58:40] So when are we moving away from phab to GitLab for everything? >:D [23:58:49] no [23:58:51] :p [23:58:51] see if you can spot the phab favicon: https://twittoons.com/29 [23:58:53] mutante: wtf is that using WLM logo? [23:59:15] Platonides: haha, probably because they copied banner code [23:59:19] TheresNoTime: As far as i know they still use it at facebook, but i don't have any friends anymore that work there, so i really don't know [23:59:49] heh I'm surprised phab died the way it did then, if Facebook were using it ..