To help
centralise discussions and keep related topics together, talk pages for some individual database reports are either redirected here, or shown a notice to start future conversations here.
Looking for a template with no transclusions is much easier than just looking for one that happens to be a self-transclusion...I'm thinking of how to restructure the SQL query to accommodate this, if anyone wants to propose a better query that handles this, please do.
Legoktm (
talk)
04:02, 16 February 2024 (UTC)reply
I don't see why it would need it? Just add a clause to the templatelinks join; you already have the template page's page_id.
quarry:query/80586. Also note the backslashes in the LIKEs; underscore is a metacharacter. —
Cryptic06:17, 22 February 2024 (UTC)reply
We have another issue which could be related to this change.
Template:Anarchism US shows a transclusion at its talk page but it's not used there. So the updated code should also check if the self transclusion is from its own talk page.
Gonnym (
talk)
07:49, 13 March 2024 (UTC)reply
That's a strange one. I'm guessing that one of the "new pages" lists causes this check somehow. I wonder if the problem will resolve itself after the new template page (created March 11) falls off of the list eight days after its creation. –
Jonesey95 (
talk)
16:29, 14 March 2024 (UTC)reply
HaleBot has not edited for a couple of days
I'm not panicking yet, but HaleBot has not edited for a couple of days. Over 48 hours, if my math is right. It averages about 45 edits per day, so a two-day break is unusual. –
Jonesey95 (
talk)
05:01, 22 February 2024 (UTC)reply
See
T358175. It's trivial to restart, but I've left it in a broken state in case it makes it easier for Toolforge admins to diagnose the underlying root cause.
Legoktm (
talk)
05:23, 22 February 2024 (UTC)reply
Polluted categories
I wanted to ask if it's possible to generate an earlier-than-usual update on a report. I hadn't personally done a runthrough on
Wikipedia:Database reports/Polluted categories in about a month or two while assuming that other people were staying on top of it, but it turns out they weren't — so when I went back to it this morning there were 1,000 categories on it, which is its generation limit, and that limit had only gotten it to the letter P, meaning that there are potentially dozens or hundreds more categories hiding on the other side of the wall.
So I've trudged my way through cleaning up what was there (pity me), but wanted to ask if it's possible to run an early update to catch the post-1,000 stuff instead of having to wait three more days for the regular weekly update.
Bearcat (
talk)
16:32, 26 March 2024 (UTC)reply
Weekly potential U5s database report not updated for nearly a year
I'm looking at the query and I don't see how it ever worked. (Besides doing things very inefficiently, it can't see user pages created after late July 2018 - intentionally, though I can't fathom why - nor users who have any deleted edits, which is probably accidental.) I'll see if I can't come up with something that does what I think it was trying to. —
Cryptic19:50, 14 April 2024 (UTC)reply
Now sorting by whether it's a redirect first then by page length, which, while not as good as user creation time, is more likely to be useful than alphabetical. (Sorting by redirect is needed to make it reasonably fast, and there's only a handful of user page redirects that meet the other criteria. And they're likely all problematic anyway.) @
Legoktm: This query can be dropped into /dbreps2/src/enwiki/webhostpages.rs without other changes, or I can take over this report with SDZeroBot's {{database report}} if you prefer. —
Cryptic20:40, 14 April 2024 (UTC)reply
Based on what I encountered, the deleted edit thing was probably supposed to make it not list user pages of users whose only contributions are creations of deleted pages and who have warnings and stuff on their User: instead of User talk: for some reason (example:
User:ISpeakTruth). Doesn't seem to be that many of those, and these situations can no longer occur, so it's probably not needed.
Flounder fillet (
talk)
21:40, 14 April 2024 (UTC)reply
That's possible, but I really don't think it is. The way it was programmed makes it look like it was accidental - it checks the current total of non-deleted edits in user: and user talk: and compares them against the user_editcount field. If it were intentional, directly checking for deleted edits in the archive table would be a more natural way to do it, more accurate, (much) faster, and could be made to only exclude users with deleted edits in non-user/usertalk namespaces besides. —
Cryptic21:52, 14 April 2024 (UTC)reply
I wrote some versions of this query. Of course the query worked previously, the proof is in the page history. My off-hand guess for why it broke is that some query planner got worse or some index got changed and the query is now taking too much time or CPU to generate, but who knows. Sometimes it's a database field that's been renamed, sometimes it's something else entirely.
Cryptic, you should have access to look at the logs yourself, but if you don't for some reason, that seems like the real issue here. I don't know why you'd need to ping Legoktm and others, that seem very silly.
This query made some heuristic choices for finding these types of potentially problematic user pages. These choices obviously have trade-offs. In particular, I happened to be focused on older and longer pages, which is why I added restrictions on page.page_len and page.page_id. I personally also wanted to only find cases where the user had only edited in two specific namespaces, at least to start. However, there are lots of cases that won't be included as a result of making these choices. If a user made a single spam edit to a real article as well as spamming their user page, they wouldn't be included in this report as-written. In cases where user.user_editcount is wrong, this report could omit some pages. In cases where the page length is 498 bytes and still promotional spam that should be deleted, it wouldn't be included here. And so on.
Improvements to this and any other database report are always welcome. I thought the archive table was no longer available in database replicas, but I may be mistaken. Let's see you all do better. Please. :-) --
MZMcBride (
talk)
07:48, 15 April 2024 (UTC)reply
Quarry says my show tables; query against enwiki_p has been queued for 21 minutes now, but I was able to run this query against a database in a different cluster and archive and friends are still available. I guess I was thinking of something else. I'm doubtful it will be efficient or quick to use the archive_userindex table or similar, but I'm very interested to see what you all come up with to uncover more pages to be reviewed and potentially deleted. --
MZMcBride (
talk)
08:08, 15 April 2024 (UTC)reply
Much of what shows up in
Wikipedia:Database reports/Linked miscapitalizations is due to piped links that have no affect on the article appearance, and I spend a lot of time fixing them so that I can get down to what matters in the report. And I take a certain amount of flak for fixing things that don't affect the article appearance. If those piped links were simply skipped, the report might be a more useful list of what to fix.
On the other hand, quite a few of those piped links also have miscapitalized link text in the article, so are still worth looking at sometimes. Maybe we could have reports both ways? Or separate counts of piped and not? Other ideas?
Dicklyon (
talk)
17:07, 4 May 2024 (UTC)reply
A
short description is usually seen in the search bar, and gets cut off after around 40 characters. But it's not hard to find SD's about twice that long,
[1][2] and perhaps even longer than that. I wonder what is the longest short description.
Wizmut (
talk)
06:53, 9 May 2024 (UTC)reply
After some digging and learning I found that it is possible to do a Quarry search for these.
[3] But it might still be nice to have a page dedicated to these cases.
Wizmut (
talk)
10:14, 9 May 2024 (UTC)reply
Thanks for the links. There's a bunch of people who are passionate about short descriptions, and I like it :)
Wizmut (
talk)
13:51, 9 May 2024 (UTC)reply
Hi, I just wanted to note that the report linked above didn't run on Thursday like it usually does. Just thought I'd bring it to the attention of whomever needs to know. Thanks for all your work maintaining these reports.
LEPRICAVARK (
talk)
13:46, 18 May 2024 (UTC)reply
There's a link to source on
its user page. Most reports are a single query, which makes picking them out easy even if you don't speak Rust. A few have significant post-processing or followup queries, though I haven't found one yet that couldn't be done - perhaps a bit less easily, granted - in a single query. —
Cryptic15:27, 22 May 2024 (UTC)reply
So the discussion is here because User talk:HaleBot was retargeted to here.
User:HaleBot lists its two maintainers. The owner, who only reluctantly takes these on after their creators abandon them and nobody else steps up,
hasn't edited in 2 months and his recruited assistant
isn't healthy. The problem is probably a hiccup on Toolforge of some sort. You can't just put Toolforge jobs on autopilot and expect them to run forever.
"See
T358175. It's trivial to restart, but I've left it in a broken state in case it makes it easier for Toolforge admins to diagnose the underlying root cause.
Legoktm (
talk) 05:23, 22 February 2024 (UTC)"
Done, though it didn't make any difference to the results. Asking at
WP:RAQ might get more sets of experienced eyes on such questions; I don't know how many of the regulars there also watch this page. —
Cryptic08:39, 23 May 2024 (UTC)reply
Usurping these reports in-place still needs cooperation from the bot's maintainers, or else it'll eventually overwrite the migrated (and possibly improved) queries. —
Cryptic08:51, 23 May 2024 (UTC)reply
Thanks. The bot seems happy for now. Want to describe the bug for us? Are willing to take on some mods as I was suggesting above?
Dicklyon (
talk)
04:32, 25 May 2024 (UTC)reply
Thanks! Yes,
WP:Piped links display different text, so sometimes a link to a miscapitalized redirect isn't in need of a fix. Often, though, the displayed text is also over-capitalized. It would be awesome to have different reports or different counts of these things.
Dicklyon (
talk)
23:37, 26 May 2024 (UTC)reply
The untagged stubs report is backlogged with a lot of soft redirects to Wikiquote, Wikispecies, etc and lists of lists. It would be useful if these were ignored. The majority of the list is currently false positives which prevents new entries from being added. CFA💬19:30, 13 July 2024 (UTC) (please mention me on reply)reply