[10:29:37] ello [10:29:45] how would I nominate an entry for deletion? [10:34:20] Hi ToAruShiroiNeko :) [10:34:23] You can use https://www.wikidata.org/wiki/Wikidata:Requests_for_deletions [10:59:26] abian thank you [20:44:00] WikidataFacts : https://twitter.com/i/web/status/859080875014860800 order the clusters by alphabetic order, then by alphabetic order of the concatenation of their names ? [20:45:46] Alphos: I’m not sure what you mean [20:45:55] deduplication ;) [20:46:28] if you order each cluster, internally if you will, then all the clusters, alphabetically, you dedupe them [20:47:01] so instead of duplicates ABCDEF and FABCDE, you'd get duplicates ABCDEF and ABCDEF, which you can then DISTINCT [20:47:17] Alphos: there are no duplicate results here [20:47:28] there are four alternatives on the left, and 25 on the right [20:47:31] i was thinking of the LIMIT 100 ^^ [20:47:38] in total that makes 100 different chains, which satisfies the LIMIT 100 [20:48:45] the chains are already ordered because the last name = first name only goes one way [20:49:25] perhaps I’m still misunderstanding you, but I also think that messing too much with how WDQS is allowed to find the results will just result in timeouts [20:49:37] oh i see, you don't have cycles, ?person10 instead of ?person1 [20:50:04] my way would have been useful with cycles ;) [20:50:40] oh, I see, with cycles it is possible to return the same cycle multiple times [20:50:46] yep [20:50:47] but I didn’t see that in the results [20:50:58] I could FILTER(?person1Label < ?person2Label && …) [20:51:12] no, much better, just ?person1 < ?person2 [20:51:14] you could, but then you'd only end up with a subset of results [20:51:30] that could work too, but once again, subset of results ;) [20:51:43] you won't be able to hit all the possibilities that way ;) [20:51:45] no? [20:51:53] in each cycle there will be one smallest element [20:52:05] and one of the possibilities will have that one as ?person1 [20:52:10] so each cycle should be possible to find [20:52:18] yes, but you could have person3 between person2 and person1 [20:52:19] unless the same person occurs multiple times (which could be fixed with <= ) [20:52:36] I’d compare ?person1 with all other persons, not ?person1 < ?person2 < ?person3 [20:52:47] yep, order them that way [20:53:06] then concat their full names, and order the sets/cycles by that [20:54:11] or concat their Qid, even better [20:54:30] wait, why should I order the entire cycles too? [20:54:47] to dedupe while still hitting the LIMIT you set [20:54:58] (DISTINCT) [20:55:14] but that will definitely kill performance, won’t it? [20:55:21] because it can no longer just use the first results it finds [20:55:31] it has to find *all* results, order them all, and then select the first ones [20:56:37] yes, a little, but at least you'd get distinct cycles ; and probably less of a perf draw then the /*/