Wikipedia:Request a query

This is a page for requesting one-off database queries for certain criteria. Users who are interested and able to perform SQL queries on the projects can provide results from the Quarry website.

You may also be interested in the following:

• If you are interested in writing SQL queries or helping out here, visit our tips page.
• If you need to obtain a list of article titles that meet certain criteria, consider using PetScan (user manual) or the default search. Petscan can generate list of articles in subcategories, articles which transclude some template, etc.
• If you need to make changes to a number of articles based on a particular query, you can post to the bot requests page, depending on how many changes are needed.
• For long-term review and checking, database reports are available.

Quarry does not have access to page content, so queries which require checking wikitext cannot be answered with Quarry. However, someone may be able to assist by using Quarry in another way (e.g. checking the table of category links rather than the "Category:" text) or suggest an alternative tool.

Longest ref names

In my travels I have come across some very long ref names in ref tags, sometimes automatically generated by incorporating the title of the work that is being referenced. Occasionally I will shorten excessively long ref names just to improve readability of the wikitext in that section. This has me curious as to whether it is possible to generate a list of the longest ref names being used in articles, as there are probably some likely targets for this kind of cleanup. Is it possible to either generate a list of the longest ref names in order of length, or barring that, a list of ref names that are more than, say, 50, or perhaps 75, characters? BD2412 T 02:49, 29 December 2024 (UTC)

References aren't in the database except in page source, which isn't available in the replicas. You can find some with search, like insource:ref insource:/\< *ref *name *= *[^ <>][^<>]{76}/. That's going to miss a bunch of them, most obviously any refs crazy enough to include angle brackets in their names (though those mostly seem to be errors anyway) or ref syntax using non-standard spaces, but ElasticSearch's gratuitously awful regex engine can't do a whole lot better. Also won't find ref names populated through templates - I understand some infoboxes do this. —Cryptic 05:26, 29 December 2024 (UTC)

This is definitely plenty to start with. It is crazy that there are so many lengthy ref tags. The Wikipedia article was the most references has under 1000 of them, and if every ref name was made of an arbitrary combination of letters and numbers, they could all be handled with two character ref names. BD2412 T 18:44, 29 December 2024 (UTC)

Australia Project

I am interested to know how the Australian project is progressing. Number of

• articles created
• articles deleted
• edits by editors with/without Australia user boxes.

Wakelamp d[@-@]b (talk) 12:42, 2 January 2025 (UTC)

Which specific user boxes? What time frame? Articles those users have deleted, or articles those users created that anyone deleted? Counting edits by editors without specific user boxes is Right Out; it's going to be well over a billion, would take days to count if the query didn't die (it would), and would be useless for any purpose. —Cryptic 16:46, 2 January 2025 (UTC)

@Wakelamp, would a list of the top editors of tagged articles be useful, and then you could compare the membership list by hand? See https://quarry.wmcloud.org/query/78918 for WPMED's list. WhatamIdoing (talk) 05:53, 5 January 2025 (UTC)

Number of editors per year

I bring https://quarry.wmcloud.org/query/89411 back to haunt you. I'm trying to assemble a table of registered editors per year. I have figured out how to modify the query to pick a different year. But what I want now is the number of registered editors in each year who made 10+ edits during that year (so, 10 edits in 2024 counts, but 5 edits in 2023 plus another 5 edits in 2024 does not), 100+ edits, and 1,000+ edits.

What do you think? ("The query will die" is hopefully not the answer.) WhatamIdoing (talk) 05:52, 5 January 2025 (UTC)

I can't think of a way to do this that doesn't look at every edit in the time range. I think it's likely the query will die. But then, it managed to complete for one month (January 2024), so maybe quarry:query/89557 for all of 2024 will eventually too. No counts of currently-deleted edits this time. —Cryptic 12:04, 5 January 2025 (UTC)

quarry:query/89569 includes deleted edits. —Cryptic 23:38, 5 January 2025 (UTC)

Maybe this is something that would have to be done by the WMF's Analytics team. I can get "one edit this year" from the original query that you wrote for me, and I'm currently slowly walking it back. It takes ~40 minutes to run, plus several hours for me to remember to check it.

@Jdforrester (WMF) may be interested in knowing that the peak for number of registered editors who made 1+ edit appears to be 2014–2015, aka when the visual editor became available again. WhatamIdoing (talk) 03:34, 6 January 2025 (UTC)

The 2024 query has completed. The one including deleted edits took an hour to run.

Is the right choice to fork it and run each year separately, or can it be expanded to do all/several years at once? WhatamIdoing (talk) 04:57, 6 January 2025 (UTC)

Better to fork it. The time for deleted edits isn't going to change much - it has to do a full table scan, since there isn't an appropriate index - but it's still the live edits that take the bulk of the time, and that is improved by narrowing the timespan looked at. —Cryptic 05:13, 6 January 2025 (UTC)

Okay. I forked it to quarry:query/89581, changed the years from 2024 to 2023, and set it to run again. If this works, then I can repeat that step a dozen times.

BTW, the earlier query got a slightly smaller total number of editors for 2024. WhatamIdoing (talk) 05:18, 6 January 2025 (UTC)

That's expected, regardless of which earlier query you mean. Query 89557 will have fewer than 89569 because there's plenty of users who have deleted edits in 2024 but no currently-live ones. One based on 89411 will have very slightly fewer because the views of the revision and archive tables it's looking at are slightly more heavily redacted than the ones 89557/89569 use. —Cryptic 05:35, 6 January 2025 (UTC)

89411 tells me there were 812,635 editors in 2023.

89581 (forked from your new 89569) tells me there were 11 fewer editors in 2023.

But the same scripts for 2024 vary in the opposite direction. The second script finds 29 more editors in 2024. Mostly the second script seems to be finding a small handful more editors (2 to 50) in each year. WhatamIdoing (talk) 02:46, 7 January 2025 (UTC)

I don't see a run in 89411's history for 2023. quarry:query/80211 shows that figure, but it was run almost a year ago. I've just rerun it, and it's giving me 812621, which is both more consistent with the 812624 from the new query and offset in the right direction. Or less wrong direction, anyway.

I'm reasonably confident that the reason we're getting fewer numbers from the same queries now compared to a year ago is because there's been more revdeletions and suppressions in the meantime; neither query can see such edits. It doesn't surprise me a bit to find out there's been revdeletions of 2023 edits made after early February 2024. I'll run a comparison of users that would be seen by the two queries so we can have a better idea why they show up in the second and not the first; it's going to take a while though. —Cryptic 04:25, 7 January 2025 (UTC)

OK, the three users that show up in the second query for 2023 but not the earlier one are Hhhj24 (talk · contribs · deleted contribs · page moves · block user · block log), IFAG dreifive (talk · contribs · deleted contribs · page moves · block user · block log), and Christian Granbacher (talk · contribs · deleted contribs · page moves · block user · block log). The first two have one live edit each in 2023; the third has two deleted edits in 2023. All three have user_editcount = 0 despite those edits, as can be seen in their contributions links. The way the earlier query works is it first fetches all users with user_editcount at least 1, then checks each of those users to see whether they have any live or deleted edits in the requested timeframe. So it doesn't ever check for edits by those three users because of the bad data in user_editcount. —Cryptic 05:22, 7 January 2025 (UTC)

Thank you. I'm going to file this under "deletions happen" and not worry about it. WhatamIdoing (talk) 03:25, 8 January 2025 (UTC)

Talk pages

Hey, I'm looking for help with creating two SQL queries

1. Find all talk pages (excluding archive talk pages) that do not use {{WikiProject banner shell}}.
2. Find all articles with no talk page (red link).

Gonnym (talk) 13:11, 15 January 2025 (UTC)

There are at least hundreds of thousands, likely millions, of each. quarry:query/89907 and 89908 have the first 10k. —Cryptic 14:18, 15 January 2025 (UTC)

Thanks! The first query is not producing correct results. The second result on the list is Talk:Destroy-Oh-Boy! which does use the template. Gonnym (talk) 16:50, 15 January 2025 (UTC)

No, the second result is Talk:!!Destroy-Oh-Boy!!. The page it redirects to transcludes the template. —Cryptic 17:40, 15 January 2025 (UTC)

(Which, of course, isn't that helpful an answer, so I've updated the query in-place to exclude talk pages that are redirects. —Cryptic 17:43, 15 January 2025 (UTC))

Serial commas in page titles

I posted the following request at WP:BOTR and was advised to come here.

Extended content

Hello, I'm not sure that this request can be completed automatically; please accept my apology if it can't. I just want some lists, without edits to anything except the page where you put the lists, so it's not a CONTEXTBOT issue: just a "good use of time" issue. Could you compile some lists of pages in which serial commas are present or are omitted? I just discovered List of cities, towns and villages in Cyprus and created List of cities, towns, and villages in Cyprus as a redirect to support serial commas. Ideally, whenever a page could have a serial comma in the title, we'd have a redirect for the form not used by the current title, but I assume this isn't always the case. First off, I'd like a list of all mainspace pages (whether articles, lists, disambiguation pages, anything else except redirects) that use a serial comma. I think the criteria might be: [one or more words] comma [one or more words] comma ["and" or "or"] [one or more words] I'm unsure whether they're rigid enough, or whether they might return a lot of false positives. Secondly, I'd like a list of all pages whose titles are identical to the first list, except lacking a serial comma. Redirects would be acceptable here, since if I'm creating serial-comma redirects, it helps to know if it already exists. Thirdly, I'd like a list of all mainspace pages (whether articles, lists, disambiguation pages, anything else except redirects) that could use a serial comma but don't. I think the criteria would be: [Page is not on first or second list] [one or more words] comma one or more words] ["and" or "or", but no comma immediately beforehand] [one or more words] Once the list is complete, the bot checks each page with the following process: "if I inserted a comma immediately before 'and' or 'or', would it appear on the first list?" If the answer is "no", the bot removes it from the list. Fourthly, I'd like a list of all pages whose titles are identical to the third list, except they have a serial comma. Again, redirects are acceptable. Is this a reasonable request? Please let me know if it's not, so I don't waste your time.

After my request, User:Qwerfjkl suggested that I come here and offered some comments: intitle:/[A-Za-z ]+, [A-Za-z ]+, (and|or) [A-Za-z ]+/ would work for the first request and intitle:/[A-Za-z ]+, [A-Za-z ]+ (and|or) [A-Za-z ]+/ would work for the second.
The latter two lists are trickier. To this, I replied Is there a way to download a list of results from a particular search? As far as I know, the only way to get a list of results is to copy/paste the whole thing somewhere and delete everything that's not a page title. (With 11,544 results for the first search, this isn't something I want to do manually.) Also, the first search includes redirects, e.g. Orders, decorations, and medals of the United Nations is result #1.

Please ping me if you reply; I've never heard of this page before, and I might not remember to come back if you don't ping me. Thanks for your help. Nyttend (talk) 03:22, 20 January 2025 (UTC)

@Nyttend: Doing the whole thing would be very awkward; we don't have access to create temporary tables on the replicas, so it would have to all be in one query, and I haven't been able to cajole the query pessimizer into doing it efficiently. What I can do quite easily is give you lists of all mainspace titles that match each of those regexes, along with whether each is a redirect, and what the titles would be if the comma is added/removed (as appropriate). quarry:query/90019 and 90020. Those should be sufficient to construct your four lists, though perhaps not trivially. I can do that too if you're unable, though not tonight and maybe not tomorrow either; let me know. —Cryptic 04:32, 20 January 2025 (UTC)

Could you give me two tables with four columns of data? First column is the link, second column is existing status (red, blue redirect, blue non-redirect), third column is alternate-comma variant (with-comma for the without-comma table; without-comma for the with-comma table), and fourth column is status of variant. If you do this in a simple table, I should be able to copy/paste it into Excel, sort so I can easily remove everything I don't want, and move the remaining content back to MediaWiki. Nyttend (talk) 04:44, 20 January 2025 (UTC)

Ah, never mind, I've explored the links and just realised that there's a way to download a CSV with almost all this information. This should suffice. Thanks! Nyttend (talk) 04:47, 20 January 2025 (UTC)

@Nyttend: I've updated both queries to include the status of the variant titles. —Cryptic 05:00, 20 January 2025 (UTC)