[02:02:26] is HakanIST here? [12:25:26] WikidataFacts : 0.999… is the same as 1. Just like sqrt(4) is the same as 2. they're just two different representations for the same number :p [12:29:24] DanielK_WMDE: Do you have a moment? [12:30:32] we have a meeting [12:30:40] ah right, it's Tuesday [12:30:53] nevermind, then [12:54:03] sjoerddebruin: FYI: https://phabricator.wikimedia.org/T132839#2556926 [12:54:15] Yes, I was reading my e-mail during your ping. [12:54:18] Let me check. [12:54:19] Qualifier and reference suggestions on external ids should now work again (to some degree) [12:55:11] They are back on NTA at least. [12:55:11] I've updated https://www.wikidata.org/wiki/Wikidata:True_duplicates just in case anyone wants to do some clean up [12:56:00] We still have the general problem of the probability there sometimes being to small [12:56:14] but at least we don't delete them for the workaround anymore. [12:56:19] But the workaround still sucks :( [13:02:28] hoo: any wonder why ISNI dropped so much? It used to be part of almost all suggestions for people. [13:04:35] did it? let me check [13:08:27] I can't see anything obvious [13:08:39] but there's *a lot of* data about this [13:08:48] and I don't really have a lot of time right now [13:08:49] Could be that other properties became more popular, but even if I add other identifiers it doesn't show up. [13:12:54] identifiers are taken out of the equation [13:13:02] that's the workaround [13:13:29] Ugh. [13:13:49] Yeah… identifiers are often used across various subject areas [13:13:56] thus they skew suggestions [13:14:18] and because our suggester model can't cope with that, we took the easy way to remove them entirely [13:14:46] I documented the exact problem here: https://phabricator.wikimedia.org/T132839#2270026 [13:15:51] anyway, I got to go [13:15:55] no problem [13:22:20] SMalyshev: any reason why https://www.wikidata.org/w/index.php?title=Q61606&type=revision&diff=363356262&oldid=355521298 isn't processed in SPARQL yet? [13:22:25] (and a few others) [13:58:55] Hi there. At Léa's suggestion I moved the Authority control help page over to the Help section, middle column, "More experienced" area. Does anyone know if the column is alphabetized or in some sort of order? I put it near Source because it seemed like a logical adjacency, but would like to position the item correctly. [13:59:25] Could you post a link? [13:59:52] Ugh. Apologies. https://www.wikidata.org/wiki/Help:Contents [13:59:56] https://www.wikidata.org/wiki/Help:Authority_control [14:00:26] BrillLyle: i think how you did this is good [14:01:14] Thanks @Aude! (Hi!) Also: I moved the page a couple times because it is before my morning coffee, so there are some non-ideal re-directs. Not sure if I need to clean those up somehow? [14:01:35] https://www.wikidata.org/w/index.php?title=Help:Help:Authority_control&redirect=no [14:01:53] https://www.wikidata.org/w/index.php?title=Help_talk:Help:Authority_control&redirect=no [14:02:39] i'm not admin so can't help [14:02:55] someone else here probably can [14:02:57] Ah, okay. Maybe I will try to mark those pages with {{Delete}} and see if that helps [14:02:59] :-) [14:03:01] yeah [14:06:22] I think I fixed redirect and cleaned up after the bad moves. thanks Aude! :-) [14:08:07] Thanks BrillLyle :) [14:19:25] https://gerrit.wikimedia.org/r/#/c/299284/ [14:19:42] This failure doesn't seem to be related to the patch [14:52:51] Jonas_WMDE: Thiemo_WMDE DanielK_WMDE are you aware of any blockers or issues for deployment? [14:53:36] we are getting segmentation fault in the build: https://gerrit.wikimedia.org/r/#/c/305000/ [15:04:54] aude: i'm not aware of any, but that does not mean much. i have not reviewed the set of changes that will be deployed. [15:05:06] the segfault issue is really annoying :( [15:05:35] it seems to consistently happen for some change sets, but not for others. see for instance https://gerrit.wikimedia.org/r/#/c/303844/ vs https://gerrit.wikimedia.org/r/#/c/303616/6 [15:05:45] would like to figure this out, obviously [15:06:02] looks like a zend bug... [15:06:06] hhvm is unaffected. [15:06:09] yeah [15:06:46] i still rather not overrule jenkins [15:06:57] *sigh* yea [15:07:00] oh, look at this https://gerrit.wikimedia.org/r/#/c/305025/ [15:07:08] the deployment build removes mediainfo [15:07:10] and it passes [15:07:16] * aude suspected this might happen [15:07:48] yes, there is probably specific code that triggers this [15:08:04] yep [15:08:25] though in case of the two changes i linked earlier, it's strange - one seems to introduce the issue, and the second wone seems to fix it again... [15:08:43] hmmm [15:08:59] we have a test case that has an optional dependency on mediainfo. might be related to that [15:09:17] that is already in master, and works. but my patches touch that [15:09:19] might be related [15:09:23] could be [15:09:28] i can try to look into this [15:09:45] that would be awesome. [15:10:04] there's already a ticket about this... hoo filed it, i think [15:10:16] would like review of https://gerrit.wikimedia.org/r/#/c/303838/ also [15:10:22] i think the problem first occurred on travis [15:10:33] i want to add some more tests there, but maybe the approach is okay now [15:10:50] i'll have a look [15:10:52] thanks [15:11:05] have talked a bit with stas about this [15:11:24] yea, it's a bit of a conceptual issue [15:11:27] https://gerrit.wikimedia.org/r/#/c/299284/ [15:11:36] in my mind, the issue of extracting the actual data is pretty clear [15:11:45] hoo +2'ed it but got a warning so he -1'ed it. I fixed it :D [15:11:49] but the issue of declaring the fields... not so much. ugh [15:12:05] yeah [15:13:13] there also are fields that ContentHandler provides data for [15:13:19] but no mapping (the mapping is still in cirrus) [15:14:09] oh :( [15:14:15] * aude thinks the mapping and indexing should match [15:15:01] and fields like title that are not definied at all in core [15:15:13] but suppose okay for now and we are going in the right direction [15:16:43] aude from UI viewpoint I don't see any blockers for deployment [15:17:04] Jonas_WMDE: thanks [17:05:44] hey sjoerddebruin, heb je gezien dat het aantal pagina's zonder item flink omlaag is gegaan? Ben jij dat? :-) [17:06:13] Volgens mij niet, ik heb een kleine wikibreak. [19:20:20] multichill there is no link to the embed.html except inside of the embed code, but we could use the 2nd short url button to link to the result [19:40:45] Jonas_WMDE: That might be nice. I like the view :-) [19:41:22] The sparql engine has quite a few nice tricks. We should make those more accessible to get more usage [19:42:19] Would be cool to be able to embed this in wiki articles. Shouldn't be too hard with all the work Yurik has been doing [19:50:18] working on that right now :) [19:53:13] yurik : i think the subtext was "what's taking so long ?!" ;-p [19:53:40] Alphos, ah, welcome new volunteer, happy to see you :) [19:54:00] let me know how you can help, and i will gladly help you get started :-P [19:54:04] :D [19:54:29] i'd love to if i could. but right now i'd be having trouble implementing "Hello world!" in plain text [19:54:45] hopefully things will get better after the summer [19:56:31] hehe [19:56:39] Alphos: No, the subtext would be: He might already have implemented it, but I just didn't notice it yet ;-) [19:56:45] multichill ^^ [19:56:50] good luck! and btw, any help is help - you can even help with organizing our help pages :) [19:57:02] bad idea, as of now [19:57:24] without volunteer help, we have to do all of it, hence the slowdown :) [19:57:28] having trouble concentrating, anything too complex or for too long and i simply shut down for 36 hours - tried it, didn't like it [19:57:51] and "too complex" is incredibly not complex enough by my old standards :-( [19:58:03] cluster headaches take some getting used to >_> [20:00:35] yurik: Did you see the self updating graph based on sparql? [20:01:09] multichill, https://www.mediawiki.org/wiki/Extension:Graph/Demo/Sparql ? [20:01:57] Alphos, sucks :( [20:01:59] yurik: https://www.wikidata.org/wiki/Property_talk:P2803/chart_P569_P570 [20:01:59] P2803 (An Untitled Masterwork) - https://phabricator.wikimedia.org/P2803 [20:02:55] yurik not nearly as bad as the recurring pain, but yeah, i got the whole-package deal :-( [20:03:00] multichill, nice! but i'm confused, how is the data updated? [20:03:54] multichill, or is it done by a script?? [20:04:09] Do you know what Listeriabot is yurik? [20:04:26] no, but i think i see what its doing [20:04:35] i would recommend doing it slightly differently though :) [20:04:38] You should wander outside of the WMF cave more often :P [20:05:02] hehe, i think i spend most of my time outside of WMF caves :) [20:05:24] multichill, i think there is a fairly minor change you can do to make that work completelly automatic [20:06:01] currently you update values by hand (by bot) - and keep the hardcoded values in wiki markup [20:06:12] Listeriabot is a bot that automagically updates pages based on SPARQL and WDQ. It's been our for quite some time now [20:06:32] It's good to figure out what will work with queries embedded in wiki pages and what won't work [20:07:09] yurik: Bot to just try what is possible, later part of MediaWiki [20:07:23] i understand, but i think it would be better to use Wikidata list template to generate the vega markup directly: see https://www.mediawiki.org/wiki/Extension:Graph/Demo/Sparql/Number_of_museums_per_country [20:07:42] instead of hardcoding the values in the vega markup, you can use the query :) [20:07:52] "url": "wikidatasparql:///?query=PREFIX%20wikibase%... [20:08:14] this way graph becomes completelly automatic, and you never need to update pages [20:09:02] I guess that extension wasn't around a year ago? [20:09:51] Maybe you can update https://www.wikidata.org/w/index.php?title=Property_talk:P2803/chart_P569_P570&action=edit ? [20:09:52] P2803 (An Untitled Masterwork) - https://phabricator.wikimedia.org/P2803 [20:10:00] this became part of graph extension when Jonas_WMDE poked me a while ago i think. Enabled in april [20:11:52] multichill, i'm afraid it will require a rewrite of https://www.wikidata.org/w/index.php?title=Template:Basic_pie_chart&action=edit [20:11:55] not a big one [20:12:13] basically we need a new graph template, specific to the properties [20:12:23] i mean - specific to use external data [20:12:32] shouldn't be too hard i guess [20:13:06] i'm not sure how easy it would be to change the graph:chart for this [20:27:11] multichill, is it always pie chart? or are there some other charts in use this way? [20:28:14] Probably more [20:28:35] Just have this one on my watchlist because I still want to figure out what kind of reporting charts to make [20:31:40] multichill, ah, i see. So how about this - i can show you how you can change that pie chart to use sparql, and you adapt it for other pages? [20:32:13] the biggest part of it - the original sparql query needs to be converted to URL (which i think is already being done somewhere) [20:37:39] multichill, btw, this pie chart can be used as is - simply change the URL: https://www.mediawiki.org/wiki/Extension:Graph/Demo/CategoryPie [20:38:42] well, maybe minor changes, like the labels [20:43:27] multichill, another thing - I actually am not very happy with this query: http://tinyurl.com/z29jx3o -- it returns everything concatenated, which is much harder to process. Instead, the query should return two columns: label and value [20:44:18] everything else, like the legend title and item id do not belong in the query result - they should be separate params for the graph template [20:51:30] I think Jura just did that to get things to work [20:51:36] But I agree that it is a bit hackish [21:14:08] multichill, is it hard to change the query into a more regular form? i'm pretty bad at sparql [21:18:17] yurik: not really, see http://tinyurl.com/zmr8uq8 [21:18:45] (wd:Q5 as ?item) is just the sparql way of assigning wd-Q5 to the variable ?item [21:21:06] multichill, cool, i will make a sample pie chart in a bit. Is it a common graph? [21:21:25] i mean - something that shows a pie chart with label-count using wikidata query? [21:23:06] I haven't seen much graphs on Wikidata [21:23:41] yurik: https://www.wikidata.org/wiki/Special:WhatLinksHere/Module:Graph <- hmm [21:24:40] multichill, no, i mean will it be useful to you to have a generic pie chart graph that uses wikidata sparql as data source? [21:24:42] yurik: I wonder if polestar can do pie charts? [21:26:09] SMalyshev, i just looked through all the vega-lite samples, and no, doesn't look like it [21:26:09] https://vega.github.io/vega-editor/?mode=vega-lite&spec=trellis_barley [21:28:49] that's weird - all kinds of charts *except* the one used in every second presentation [21:29:20] no wait vega has this: https://vega.github.io/vega-editor/?mode=vega&spec=arc [21:30:27] so vega can do it, just polestar can't [21:31:32] That would be useful yes [21:53:36] SMalyshev, correct, vega has all the possible graphing primitives. My guess is that pie charts are one-dimensional, as oppose to all other graphs that are x-y [21:54:19] true... probably they're too simple to need polestar I guess [21:54:27] still would be nice... [21:54:59] my dream is to have full visual graph designer hooked to wdqs :) [22:00:29] SMalyshev, that would not be polestar, but lyra :) [22:00:38] sadly, its not really ready yet imo [22:00:56] yurik: yeah I looked into lyra and it's kind of complex. I guess baby steps for now :) [22:01:05] yep :) [22:01:32] SMalyshev, subscribe to vega mailing list (google group), and ask what is the status of vega lite, and if it will support pie charts [22:01:42] i plan to enable it fairly soon, but i saw there was a lot of changes recently [22:58:26] SMalyshev, https://www.wikidata.org/wiki/User:Yurik/sandbox - we can obviously improve it further [22:58:59] yurik: coolio! [22:59:20] SMalyshev, i will hack at it a bit more to get it to show stuff in a better way [22:59:30] the radial labels don't look good for it [22:59:55] I think for pies you can just display them as a column aside [23:49:09] SMalyshev, https://www.wikidata.org/wiki/User:Yurik/sandbox [23:49:34] we really ought to fix the server-side fonts :( [23:49:37] yeah looks good [23:49:44] the fonts are broken ;( [23:49:53] fonts are kind of weird [23:49:55] true [23:50:11] Also, I'd take the legend a bit closer to the pie [23:50:22] right now there's tons of whitespace there [23:50:39] yurik technically pie-charts are 2-dimensional, with colors as a second, discrete, dimension ;p [23:51:18] Alphos, i guess so. Ok, i really don't know why they didn't add vega-lite support for it :) [23:51:28] yurik: also, I wonder if you can do tooltips? It's be nice if the numbers were tooltips (and maybe some text too)? [23:51:44] SMalyshev, i can totally do the tooltip, but then you have to click to activate the graph [23:51:58] hmm... maybe not worth it then [23:52:05] yep [23:52:13] not until we support vega as client-side by default [23:52:52] yurik because pie-charts are a nightmare to draw perhaps ? :p [23:53:18] (really, not kidding, nightmare, compared to right-angle charts) [23:54:59] Alphos, not really - they are actually almost identical to draw as rectangular ones [23:55:27] you need to scale the data, but instead of scaling it to the height, you scale it to 360 degrees [23:55:36] yurik absolutely [23:55:41] but that's a nightmare ^^ [23:56:01] nah, drawing layered (stacked) charts is much worse [23:56:16] you need to figure out where each layer ends, and the next one begins [23:56:23] for each data point [23:56:26] granted [23:56:33] and totals have to still fit within total graph height [23:56:40] d3 takes care of all that [23:56:41] maybe they're allergic to camembert ? [23:56:53] i like that theory better [23:56:58] (yah, we call "pie charts" "camemberts", in french) [23:57:02] SMalyshev, made it slightly smaller [23:57:12] i see another possibility [23:57:14] the whitespace [23:57:48] maybe they specifically didn't want pie-charts because of their potential use for ugly, lying data representation [23:58:39] 2D pie charts (aka "disks") are fine. 2.5D pie charts ("disks, but flipped backwards") are frequently used to misrepresent data [23:58:58] maybe they didn't want people to ask for pie charts with perspective ? [23:59:57] having met Vega people, i would say they simply didn't care for pie charts that much, simply because pie charts are not used in science that much