Wikipedia:Village pump (technical)

Jump to navigation Jump to search
 Policy Technical Proposals Idea lab WMF Miscellaneous 
The technical section of the village pump is used to discuss technical issues about Wikipedia. Bug reports and feature requests should be made in Phabricator (see how to report a bug). Bugs with security implications should be reported differently (see how to report security bugs).

Newcomers to the technical village pump are encouraged to read these guidelines prior to posting here. If you want to report a JavaScript error, please follow this guideline. Questions about MediaWiki in general should be posted at the MediaWiki support desk. Discussions are automatically archived after remaining inactive for five days.

Frequently asked questions (FAQ) (see also: Wikipedia:FAQ/Technical)
Click "[show]" next to each point to see more details.
If something looks wrong, purge the server's cache, then bypass your browser's cache.
This tends to solve most issues, including improper display of images, user-preferences not loading, and old versions of pages being shown.
No, we will not use JavaScript to set focus on the search box.
This would interfere with usability, accessibility, keyboard navigation and standard forms. See task 3864. There is an accesskey property on it (default to accesskey="f" in English). Logged-in users can set a gadget in their preferences.
No, we will not add a spell-checker, or spell-checking bot.
You can use a web browser such as Firefox, which has a spell checker.
If you have problems making your fancy signature work, check Wikipedia:How to fix your signature.
If you changed to another skin and cannot change back, use this link.
Alternatively, you can press Tab until the "Save" button is highlighted, and press Enter. Using Mozilla Firefox also seems to solve the problem.
If an image thumbnail is not showing, try purging its image description page.
If the image is from Wikimedia Commons, you might have to purge there too. If it doesn't work, try again before doing anything else. Some ad blockers, proxies, or firewalls block URLs containing /ad/ or ending in common executable suffixes. This can cause some images or articles to not appear.
For server or network status, please see Wikimedia Metrics.
« Archives, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185

Talkpage technical error at Talk:Cultural Marxism conspiracy theory[edit]


Can someone please have a look at this? I was having a talkpage technical error at Talk:Cultural Marxism conspiracy theory? Relevant version. I have fixed the issue in the current version, but I think it will mess-up the archives when we get to that time. (please {{ping|}} on reply; thanks!) --Emir of Wikipedia (talk) 21:08, 2 November 2020 (UTC)

If someone could help that would be appreciated. Emir of Wikipedia (talk) 18:33, 3 November 2020 (UTC)
Bumping thread. Emir of Wikipedia (talk) 17:20, 4 November 2020 (UTC) (please {{ping|}} on reply; thanks!)
@Emir of Wikipedia: I'm not sure I'm seeing what you're looking at or what you might have done to fix it, but it looks like there's a quote box with an unclosed ref tag in it. Look for <ref name="braune2">. I think this code would produce a "ref invoked but not defined" error on the talk page, but I'm not sure if that error would render inside the quote box, or if it would just mess things up unpredictably. You need to close the tag, otherwise Wikipedia thinks everything after it is part of the reference. You can add a / to the end of the tag (<ref name="braune2" />), but then you also need to define the reference and add a {{reflist-talk}} somewhere in the section. It might be better to pull the URL of the reference from the article and replace the ref tag with an external link on the talk page instead. Ivanvector (Talk/Edits) 17:36, 4 November 2020 (UTC)
Ivanvector. Thanks. I just used an external link, and it looks like that has fixed it. --Emir of Wikipedia (talk) 19:50, 4 November 2020 (UTC)
This is the edit. Emir of Wikipedia, on a side note, please don't manually archive VPT threads (or in general, threads on any page which uses bot archiving), you used an out-of-sequence archive number with the upshot that when Lowercase sigmabot III (talk · contribs) next visited this page, it moved threads to a different archive. If a thread is resolved, just use {{resolved}}. --Redrose64 🌹 (talk) 08:18, 6 November 2020 (UTC)

Announcing xCite: Templates eXported from Wikipedia[edit]

xCite is weekly dump of CS1|2 templates on enwiki and other select languages. I made it for another project and thought it might have general use so am making it public. It will run from cron indefinitely. AFAIK there are no other dumps like it. It has many potential uses but was mainly designed as an intermediary step to feed into a database of citations. It could run much faster depending how many concurrent Toolforge slots are configured. -- GreenC 15:07, 16 November 2020 (UTC)

Interesting stuff, GreenC. Thanks for doing this. – SD0001 (talk) 04:37, 18 November 2020 (UTC)
Sure thing, User:SD0001! It seemed trivial to scrape CS1|2 but the challenge is to do it fast, consistent and reliable. So far it is working well. -- GreenC 05:20, 22 November 2020 (UTC)

Category Navigation is messed up[edit]

I posted this elsewhere and was asked to report it here. I am not sure if this is the right place to post this, but I need to post it somewhere. The system is not letting me go through categories in the normal way. When I open a category it has the first 200 articles. However when I click on next it goes straight to a page starting with the 200 articles beginning with the first one categorized under B. If I go back it gives me a page with the previous 200 articles, but will only allow me to go back 1 page. This means in some categories some articles in the category cannot be navigated from. This is a very frustrating situation.John Pack Lambert (talk) 20:28, 16 November 2020 (UTC)

The above is not the full extent of the problem. When I click on the A tab to navitage in categories it takes me to B, and when I click on B it takes me to C. The more specific tabs like Ae or Aj take me to B, and always at the start of B. So that method of navigation is a problem as well. I have seen this in the 1927, 1929 and 1990 birth year categories. The same problem occured when I went to the category 20th-century American journalists. It seems to be a general navigation problem. I actually turned off my computer and turned it back on things it might be a function of something on my computer. I do not think it is. I have only noticed this problem in today.John Pack Lambert (talk) 20:35, 16 November 2020 (UTC)
I started noticing this problem today as well, within Living people and specific birth year categories, after about 20:00 UTC. --Ken Gallager (talk) 20:49, 16 November 2020 (UTC)

I would guess this is related to the Unicode upgrade. (More specific thread.) Try again toward the end of the week. --Izno (talk) 20:50, 16 November 2020 (UTC)

(reposting from above): It took about nine days in 2016 and was estimated to take about eight days in 2018. I don't know if the process is similar this time, but I would not be surprised to see it take a week to get sorted (I couldn't resist). – Jonesey95 (talk) 22:04, 16 November 2020 (UTC)
@Johnpacklambert: Examples are always good, per the box at the top. Without a link to a category, we can't see what is happening, so can't judge if there is a problem or not. --Redrose64 🌹 (talk) 21:24, 17 November 2020 (UTC)
The problem seems to have been resolved.John Pack Lambert (talk) 21:26, 17 November 2020 (UTC)
Maybe on that one page, but see my example below. --Guy Macon (talk) 20:40, 19 November 2020 (UTC)

Re: "Examples are good":

On the page Category:Taylor & Francis academic journals when I click on "A" I get B through J and when I click on "B" I get C through J. "C" gives me D through J, and "D" gives me E through L.

"0-9" gives me A but the "previous page " link works and brings me to Écoscience for some strange reason.

"W" gets me X and "X" gets me "There are no pages or files in this category".

--Guy Macon (talk) 20:09, 19 November 2020 (UTC)

Category: Living People[edit]

Hi people, the list jumps from Amal Azzudin to Shahida Abbasi, I suspect that the recent people have been added. I haven't noticed before...GrahamHardy (talk) 18:41, 18 November 2020 (UTC)

@GrahamHardy: All categories will be out-of-sorts for the next few days due to a recent change in the Wikipedia software. If it's still a problem in a week and a half, raise the issue again. davidwr/(talk)/(contribs) 18:44, 18 November 2020 (UTC) (true, but irrelevant) davidwr/(talk)/(contribs) 18:46, 18 November 2020 (UTC)
Oh, the problem you saw may be unrelated, it has to do with the {{DEFAULTSORT}} on the respective pages. Most, but not all, people are "sorted" last-name-first. davidwr/(talk)/(contribs) 18:46, 18 November 2020 (UTC) Update Yup, Shahida Abbasi has {{DEFAULTSORT:Abbasi, Shahida}} in it. davidwr/(talk)/(contribs) 18:48, 18 November 2020 (UTC)
No, those sort keys don't explain the incorrect sorting. The actual issue is the one you struck, also described above in #Tech News: 2020-47. MANdARAX  XAЯAbИAM 19:21, 18 November 2020 (UTC)
Doh, yeah, Abbassi comes before either Amal or Azzudin. Both should come after "Caffeine, ingesting." davidwr/(talk)/(contribs) 19:24, 18 November 2020 (UTC)

Categories not alphabetizing new articles properly[edit]

This may be a problem that has already been noticed by other people, but as I can't find evidence that it's been raised here recently I wanted to mention it just in case.

Some recently created articles — specifically articles that were either newly created today, or already existed but got added today to a category that they weren't previously in — don't seem to be getting alphabetized correctly in their relevant categories. For example, in Category:Members of the Delaware House of Representatives, the articles Sarah McBride, Eric Morrison and Marie Pinkney are all flopping out of alphabetical order, and just sitting at the end of their respective letters. And by the same token, Category:Members of the Tennessee House of Representatives features end-of-letter misfiles of Torrey Harris and Eddie Mannis, both newly created today, and Leonidas "Leon" Howard, which already existed but had its sortkey corrected today because the creator had copypasted the defaultsort template from another person without correcting it to reflect Leon Howard's name instead of the other person's.

I know this problem has occurred in the past; however, I don't recall whether it resolved itself naturally, or whether somebody had to tweak something at the server level to fix it. Bearcat (talk) 19:53, 18 November 2020 (UTC)

Expect this to take a week or so to be resolved. See above. – Jonesey95 (talk) 21:41, 18 November 2020 (UTC)

References and punctuation[edit]


What is going on here? I originally was told "references after punctuation," and now it seems that WikiCleanerBot is going around switching them to before. Is this a policy change? It seems like something that affects a lot of articles (certainly any that I have worked on). If it is a policy change, how was it communicated? Peter Flass (talk) 15:53, 17 November 2020 (UTC)

Peter Flass, it shouldn't be a policy change; H:FOOT still considers references immediately after punctuation to be correct. It seems to be a code error. —Tenryuu 🐲 ( 💬 • 📝 ) 16:24, 17 November 2020 (UTC)
If you think a bot is malfunctioning, please contact its operator first - if that doesn't lead to a resolution you can start a discussion at WP:BOTN. — xaosflux Talk 16:35, 17 November 2020 (UTC)
Peter Flass Do you have any example of a malfunction? My bot is applying MOS by putting references after punctuation... The edit comment is "Reference before punctuation" because it's the problem that is fixed... --NicoV (Talk on frwiki) 20:23, 17 November 2020 (UTC)
Peter Flass: Examples are always good, per the box at the top. Without a diff (or similar), we can't see what happened, so can't judge if there is a problem or not. --Redrose64 🌹 (talk) 21:18, 17 November 2020 (UTC)
As I expected, it was a false alarm. --NicoV (Talk on frwiki) 08:02, 19 November 2020 (UTC)
@NicoV: It's unclear to me whether the ambiguous edit summary was a factor here, but it certainly would not hurt to remove the ambiguity. Nor would it be unduly difficult to insert the word "Fix" at the beginning. One or two minutes' effort and you're done for eternity. ―Mandruss  08:52, 19 November 2020 (UTC)
@Mandruss: Yes, I was thinking about that too. I changed the configuration, here's an example of the result. --NicoV (Talk on frwiki) 11:10, 19 November 2020 (UTC)
@NicoV: Improvement. Thanks. ―Mandruss  15:45, 19 November 2020 (UTC)

Searching for pound signs[edit]

I recently came cross n article that had used the string UK£ where GB£ should have been used. I changed it, and decided to search for similar cases.

But it seems that the £ character cannot be searched for; a search for UK£ seems to find every page including UK, even with quote marks, and even using insource.

Can someone suggest a work-around, please? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 13:59, 18 November 2020 (UTC)

Did this not work:
insource:/UK£/ – 150ish results
Trappist the monk (talk) 14:07, 18 November 2020 (UTC)
@Pigsonthewing: (ec) A search for UK insource:/UK£/ works. (insource: with slashes uses a regex rather than the fuzzier matching of quotes. The initial UK improves efficiency.) Certes (talk) 14:08, 18 November 2020 (UTC)
The ISO 4217 currency code is GBP. GB£ is just as invalid as UK£. --Redrose64 🌹 (talk) 23:53, 18 November 2020 (UTC)
Were these attempts at using an ISO code, you would be correct. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 09:21, 19 November 2020 (UTC)

Is it possible to change the edit summary field in Twinkle when welcoming users?[edit]

I am using Twinkle to leave customised welcome notes on the talk pages of users who have signed up for an online edit-a-thon on climate change. See for example here: User talk:Rickeyre. Someone has complained on my talk page (see here) that the edit summary field says "Welcome to Wikipedia!" instead of saying "Welcome to online edit-a-thon". Therefore my question: is it possible to change the edit summary content when using Twinkle with a customised welcome note? Personally, I don't think this is a big deal but the other person said I should ask her for advice. EMsmile (talk) 14:47, 18 November 2020 (UTC)

It is presently not possible. Changes to twinkle are discussed at WT:TW, though this doesn't sound important enough to be worth changing. – SD0001 (talk) 11:35, 19 November 2020 (UTC)
Thanks, User:SD0001, I'll copy my question to there. I agree it's not an important change but I'll ask there anyhow just in case it's super easy to do. EMsmile (talk) 13:10, 19 November 2020 (UTC)

Can admins see other people's watchlists?[edit]


Do we admins have the ability to see other editors' watchlists? If not, is there a reason why not? I can see it being additionally useful when comparing suspected socks, for instance, if I note that the two socks are interested in the same articles, or if a brand new editor watchlisted a bunch of drafts that I suspect of having been created via UPE. Just curious. Thanks, Cyphoidbomb (talk) 17:22, 18 November 2020 (UTC)

I don't know whether they can but the natural counterargument would be privacy. Maybe there should be a special class or tool like with what's that 'see the IP behind a nick' tool. --Palosirkka (talk) 17:38, 18 November 2020 (UTC)
They cannot. * Pppery * it has begun... 17:43, 18 November 2020 (UTC)
No we can't, and it's not going to happen. "Grossly unethical" doesn't begin to cover it; at absolute minimum we'd need express consent from every user and need to give all existing users time to edit their watchlist before it went live, given how much sensitive personal information could be inferred from a watchlist. It would be orders of magnitude more intrusive than CU. ‑ Iridescent 17:44, 18 November 2020 (UTC)
(e/c) @Cyphoidbomb: If you were to go to WP:WATCHLIST you would read, under "Privacy" that "No user, not even administrators, can tell what is in your watchlist, or who is watching any particular page. Publicly available database dumps do not include this information either. Only developers who have access to the servers that hold the Wikipedia database could obtain this kind of information." There's a link to the Privacy Policy, which might be the sort of thing of which an Admin might be reasonably expected to be vaguely aware. DuncanHill (talk) 17:46, 18 November 2020 (UTC)
They cannot. There is no degree of access (admin, 'crat, steward, staff) that provides this ability. Providing access to it would be a gross violation of the privacy policy. I did some theoretical design work about users allowing others (like a mentor) to see their watchlist, but it was bound up in some principles behind how Flow worked. It was never implemented, nor is it likely to be.--Jorm (talk) 17:47, 18 November 2020 (UTC)
You can't and I doubt it will be accessible in the future. My own opinion is that I don't think granting access would be a good idea from a privacy stance. Dreamy Jazz talk to me | my contributions 17:54, 18 November 2020 (UTC)
@Jorm: FWIW, a user can generate a key for sharing their watchlist feed, and if they share that secret key it will allow others to read their WL. Also, a user can create a BotPassword that has access to view their own watchlist and if shared would allow others to access their WL (via API). I don't recommend anyone actually doing either of those things though for this use case (but there are use-cases where you may share it with "yourself" for special purposes. — xaosflux Talk 17:56, 18 November 2020 (UTC)
Xaosflux, Wow! I actually knew about the key, but the bot password thing is new to me. MediaWiki is filled with weird esoterica like that. Jorm (talk) 18:01, 18 November 2020 (UTC)
I can see a case for giving checkusers access, so they can give us a simple similar/different verdict (like comparing IP addresses without giving us the numbers), but that's a policy rather than technical matter. Certes (talk) 18:09, 18 November 2020 (UTC)

Followup question:[edit]

Can the police see editors' watchlists if they have a warrant? Their browsing history? Do we know what information is retained on the servers and thus available with a court order and what is deleted after a certain amount of time? Doed the W?F have a Warrant canary? Asking for a friend. :) --Guy Macon (talk) 23:43, 21 November 2020 (UTC)

@Guy Macon serious question or joking? If serious, the source of MediaWiki is public, as is the database structure. But my spidery sixth sense tells me this is joking. ProcrastinatingReader (talk) 23:59, 21 November 2020 (UTC)
Not a joke, and I think you might not understand what open source means. Go ahead and download the source for MediaWiki. It's at [ ] Try to find User:ProcrastinatingReader's watchlist or even IP address. Now download the entire English Wikipedia database. Instructions are at Wikipedia:Database download, Also see [ ] Again, you won't find User:ProcrastinatingReader's watchlist or even IP address. So yes, that was a serious question, and the fact that one can download Wikipedia does not answer it.
BTW, the source of Tails (operating system) is also public, and I use TAILS regularly, but good luck finding any information on me that way, even if you have a warrant. (Or, more likely in my case, even if you are the government of the PRC where I use TAILS.) Even the developers who created TAILS don't have access to any information about their users. --Guy Macon (talk) 02:17, 22 November 2020 (UTC)
Yes but you don't need to know the contents of someone's watchlist to know what data the software stores, you just need to look at the database structure for the watchlist table, ie mw:Manual:watchlist table. I would imagine the WMF, like any other entity, would comply with a legal request for data stored on their servers. The data on your computer is not stored unencrypted on the Tails developers' servers. ProcrastinatingReader (talk) 03:01, 22 November 2020 (UTC)
@Guy Macon:, how could Special:Watchlist possibly work if the full watchlist wasn't retained on the servers? The only question is to what extent WMF-Legal is willing to resist such a demand. OTOH I am curious about the browsing history. There's some value to keeping that in the very short term. ("Hmmm, server load has increased 10x in the past few hours. Let's see which requests are the problem.") But if it's being kept for a long time, I'd like to know why. Suffusion of Yellow (talk) 02:34, 22 November 2020 (UTC)
What have you been browsing, Suffusion of Yellow? :D ProcrastinatingReader (talk) 03:09, 22 November 2020 (UTC)
As soon as I saw "This user has made over 4 edits to Wikipedia" I knew I was dealing with a heavy hitter. --Guy Macon (talk) 03:16, 22 November 2020 (UTC)
Rubber hose, ski mask, bear trap, The Pentagon, lubricant, cat scratch disease. Suffusion of Yellow (talk) 04:08, 22 November 2020 (UTC)
Re: "how could Special:Watchlist possibly work if the full watchlist wasn't retained on the servers?", it (and anything else that is only visible to me when I am logged in) could work by retaining an encrypted copy that requires my password to read. That's what the W?F can do, but I doubt that it is what they actually do. I do know that the W?F doesn't store my password, just a hash used to verify it.
Knowing the database structure for the watchlist table does not change this. The W?F could populate that table from an encrypted copy when I log on and delete the unencrypted watchlist when I log out. But probably not. -Guy Macon (talk) 03:11, 22 November 2020 (UTC)
That would still only protect users who never log in. The moment you log in, "they" have your plaintext password, and anything encrypted with it. Suffusion of Yellow (talk) 04:08, 22 November 2020 (UTC)
Yup. Or they could simply do this. :( -Guy Macon (talk) 05:18, 22 November 2020 (UTC)
Here is the privacy policy and here is the most-recent transparency report. --Izno (talk) 03:40, 22 November 2020 (UTC)

Category:Wikipedia non-empty soft redirected categories[edit]

This is a minor problem but these redirect categories are all empty but appear in this category to have contents. I have followed the instructions and did null edits to several of them but that had no effect. Many of these categories are several years old so I guess this situation has existed for a while without anyone noticing. Any ideas? Liz Read! Talk!

Category counting has some bugs. Category:CS1 errors has incorrectly shown Category:CS1 errors: dates‎ as having two subcategories for more than two years. – Jonesey95 (talk) 21:45, 18 November 2020 (UTC)

Why are empty monthly maintenance categories missing from Category:Candidates for uncontroversial speedy deletion?[edit]

Is this also why empty maintenance categories, which automatically get marked as CSD G6 when they empty, never show up in the CSD categories? See Category:Uncategorized from July 2020 which is categorized under Category:Candidates for uncontroversial speedy deletion but doesn't appear in that category. I know I brought this problem to the Village Pump years ago and was told that it was a low priority problem. Liz Read! Talk! 22:39, 18 November 2020 (UTC)
Darned if I know. There are multiple phabricator tasks that are beyond my comprehension, including T221795, T85696. – Jonesey95 (talk) 23:14, 18 November 2020 (UTC)
So, of course, my example of a tagged category that doesn't appear in the parent category got deleted as soon as I mentioned it here so I won't find another example. But it's very common for empty maintenance categories tagged for deletion to exist for days, weeks or months before an admin stumbles upon them and deletes them. Clearly not a top priority but an example of pages not appearing in categories that they are in.
Category:G13 eligible AfC submissions is another category that usually has drafts tagged to appear in it but the category, when you look at it, is almost always empty. Thanks for the phab tasks. Liz Read! Talk! 20:44, 20 November 2020 (UTC)
I think that the reason this happens is that the appearance of the speedy deletion template is caused by the emptying of the category instead of the editing of the category page. See the history of Category:Use Harvard referencing from October 2019, for example, which I emptied about 24 hours ago. As soon as it was emptied, the speedy deletion template appeared when the page was purged, but without an edit to the category page, the page's category memberships will not be updated. A bot that null-edited {{Monthly clean-up category}} (13K transclusions, so don't do it too often) once every n days would make the "cats to delete" categories no more than n days out of date while we wait for a better solution or workaround. Pinging ProcrastinatingReader to see if this is something for User:ProcBot/PurgeList. – Jonesey95 (talk) 23:46, 20 November 2020 (UTC)
In theory the software should refresh those at some point itself, and for these categories if there's a delay of "only" one or two weeks I think it's okay as it is, if it's longer then perhaps worth null editing. Do you know if there's any others like this (a) to have an idea of how widespread this issue is and (b) I want to test something on one (before it's null edited)? ProcrastinatingReader (talk) 00:06, 21 November 2020 (UTC)
In theory, sure, but in practice, see the phab links I provided above. It can take months for pages to be categorized correctly if they are not edited (or null-edited). As for other categories that aren't properly categorized yet, see the empty cats in Category:Use Harvard referencing and compare them to the population of Category:Candidates for uncontroversial speedy deletion. All of the empty cats in the former should be in the latter. Just for fun, I have null-edited about 500 pages transcluding {{Monthly clean-up category}}, and four categories appeared in the speedy deletion category; if that 1% hit rate continues, there might be 100 empty cats out there waiting to be deleted. – Jonesey95 (talk) 00:27, 21 November 2020 (UTC)
Those phab tasks are interesting but that seems related to the first issue you guys discussed (the cat counts). The issue of CSD tagging seems to be the typical job queue / null edit one. The key thing to note here is how long is being waited for the job queue to naturally get around to the pages after the cat becomes empty. If it's just a week or so then I don't think we need to force a null edit, since this is a (relatively, compared to eg AfC expiring drafts) low-priority issue I think. If it's longer than a few weeks it may be appropriate to force a null edit every now and then. I'll run a one-time on the transclusions of Monthly clean-up category and see what hits for now, it'll take a bit to complete. ProcrastinatingReader (talk) 01:43, 21 November 2020 (UTC)
As I tried to explain, these monthly maint cat pages are not being "tagged" or otherwise modified in any usual sense, so the job queue is not involved; a null edit or link refresh is required to update category membership. The relevant phab tasks in this case are probably T135964 and T159512. The tag is displayed only when the category becomes empty.
You don't have to run a null edit on the monthly cats; it is already in progress, and as far as I can tell, it has unearthed about 15 empty cats in 2,000 edits. I don't know of a way to find out how "stale" the empty cats were, but if someone looks at monthly category deletions in the last 110 minutes or so at Special:Log/delete, and can figure out how long those categories had been empty, that will be the answer to the question about staleness. – Jonesey95 (talk) 02:18, 21 November 2020 (UTC)
Okay, I'm not following all of this because I'm not adept at coding but I appreciate you thinking through what the problems might be. A couple of things to mention:
  • In Category:Wikipedia non-empty soft redirected categories, there were empty categories appearing, somewhere between 10-20. I did a null edit on those categories and they disappeared from the soft redirect category. I tried a null edit on the categories in this category that appear to have contents (mostly saying 1P), and it didn't work with them. It had no effect.
  • You're right that with the maintenance categories, which appear in Category:Clean-up categories, when they empty, the bot that created them has an automatic CSD tag that appears which is dated. You don't really know when you stumble upon them how long they have been empty but I've seen ones that had CSD tags dated months and even years ago. I don't know how they would be affected by a null edit because even though they are CSD tagged, they don't appear in CSD categories, so you actually stumble upon them accidentally or go look for them. When they are found, they are simply deleted.
  • I guess my bigger question is how come with most categorization, as soon as you categorize an article, talk, user or other page, it instantly appears in that category. But with these few irregular categories, or Category:G13 eligible AfC submissions, that doesn't happen. I don't understand why they would be exceptions. The only difference is with the typical case, there is a physical edit that changes the page but with the maintenance and G13 eligible pages, there is a shift in status that is not based on physical edits but in bot instructions to move a page/category to a different status. Even though they are CSD tagged or, in the case of G13s recategorized, they just don't appear in the parent categories. As for the soft redirected categories, that's a fluke I don't get. Liz Read! Talk! 03:04, 21 November 2020 (UTC)
Maybe the explanation at WP:NULLEDIT will help. Yes, the CSD message is visible, but only an edit (null or otherwise) will change the page's actual category membership. Category names may appear on the page after a simple browser refresh or a WP:PURGE, but the page will not actually show up in the category until it is edited. It's confusing, to be sure. This difference is why I suggested help from a null-edit bot above. As we have seen so far today, there were about 30 empty categories that resulted from null-editing about half of the monthly maintenance categories, so the problem is real. – Jonesey95 (talk) 05:09, 21 November 2020 (UTC)

Lua error: not enough memory on COVID-19 pandemic in India[edit]

Hi, on COVID-19 pandemic in India in place of most of the References and all External links it shows Lua error: not enough memory, i don't have any idea about it, and thought someone here may know about it. Thanks - Mayankj429 (talk) 14:47, 19 November 2020 (UTC)

@Mayankj429: short diagnosis - this page has too many expensive templates on it, especially inside of references; short answer - split the article or reduce some templates. Long answer - it is hard to profile the best place to optimize some of these templates, see phab:T188492 and related tasks for a lot more on this. — xaosflux Talk 15:11, 19 November 2020 (UTC)
Perhaps splitting the COVID-19_pandemic_in_India#Testing section to Testing_of_the_COVID-19_pandemic_in_India would be a good section to split, it is heavily referenced. — xaosflux Talk 15:14, 19 November 2020 (UTC)
N.B. many of the COVID articles are having similar issues with tons of reference templates making them fail in other ways, usually in to Category:Pages_where_template_include_size_is_exceeded. — xaosflux Talk 15:17, 19 November 2020 (UTC)
The India page is not as bad, but many many of the COVID pages suffer from PROSELINE problems that could also do with culling and summarizing, cutting out references to minor or interim reports in favor of summarizing and final reports. --Masem (t) 15:23, 19 November 2020 (UTC)
In this case the post-expand include size is well within limits (1003636/2097152 bytes), unlike the other covid pages where the post-expand include size exceeds 2MB and causes the citation templates to fail. Doesn't this means that the Lua memory used for the citations is not being released on each citation use and accumulates? Iirc, Lua's memory management is automatic and there no way of running the garbage collector programmatically. —  Jts1882 | talk  15:57, 19 November 2020 (UTC)
@Jts1882: this isn't a template-expand byte problem, but it is very similar in that there are many many templates that are all calling modules and exceeding the memory thresholds for page processing; iirc all of these are using the same memory pool during page rendering, not a separate pool per instance. — xaosflux Talk 16:01, 19 November 2020 (UTC)
Removing the infobox reduces memory from over the 50 MB limit to 29 MB. It also reduces the rendering time by 67%. Could the infobox be simplified? Certes (talk) 16:00, 19 November 2020 (UTC)
@Certes: Module:WikidataIB may be "expensive" there. — xaosflux Talk 16:03, 19 November 2020 (UTC)
@RexxS: any thoughts on that? — xaosflux Talk 16:09, 19 November 2020 (UTC)
In Module:WikidataIB using less of local variables for wikidata functions would help.
Module:Wd is also used on Covid-19 in India, I shaved a bit of the time and memory the sandbox version uses in this diff.--Snaevar (talk) 19:57, 19 November 2020 (UTC)
@Xaosflux: yes, WikidataIB is quite heavyweight for doing jobs like returning a single value out of 300 or so values because it reads and processes each value before passing the list of values to Template:P-1. The infobox itself uses 29.63 MB. Disabling the calls to WikidataIB reduces that to 15.74 MB, I checked by trying just the eight WikidataIB calls alone and they used 12.64 MB. The fastest and cheapest way of getting values is to use {{#property: }}, so I've written a string function to return the last value in a list. I suggest replacing the eight WikidataIB calls like this: Change
  • {{#invoke:WikidataIB|getValue|fwd=ALL|noicon=true|P1603|list=p-1}} which produces 9,177,840 (4.22 MB and 0.440 seconds)
to: {{#invoke:String2 |findlast |{{#property:P1603}} }} which produces 9,177,840 (548 KB and 0.264 seconds)
@Snaevar: I agree that stuff like local snak = propval.mainsnak or propval local dtype = snak.datatype local dv = snak.datavalue increases the memory footprint. I can only say I wasn't the instigator of that sort of "tidying". When the module was first written, we usually saw only a few values for any property. Now that Wikidata has expanded dramatically, it's not uncommon for some properties to run to several hundred values. Working on Template:Cite Q, I came across an entry for a document with about 500 signatories (which exceeds the 400 entries that Lua can load to link their names), so I'm looking for optimisations at present. --RexxS (talk) 21:17, 19 November 2020 (UTC)
WikidataIB provides functions other than simply fetching a value, and it should not be replaced without careful consideration. I'm pretty sure that WikidataIB, by default, fetches only sourced statements from Wikidata, per the rough consensus at this 2018 RFC. The proposed #property code could change that behavior. – Jonesey95 (talk) 23:18, 19 November 2020 (UTC)
@Jonesey95: when we're talking about coding into templates, I agree with you 100%, but the use here is to retrieve the last figure from one of the properties of COVID-19 pandemic in India (Q84055514) within a locally-supplied parameter value for use in a single article. The values are updated daily on Wikidata and are sourced to so I'm not too worried about importing unsourced data in this particular case. The problem reported here will only get worse as the daily updates grow relentlessly on Wikidata and we'll eventually get to the stage where Lua won't cut it, which is why I'm considering using the built-in functions that I assume are coded in either C or php. --RexxS (talk) 23:51, 19 November 2020 (UTC)

Need to remove broken links to EBSCOhost Connection[edit]

At Wikipedia:Help desk/Archives/2020 November 8 § Broken EBSCOhost Connection links, I noted that there are many broken links to EBSCOhost Connection in cited sources in Wikipedia articles. Samwalton9 (WMF) did some research and reported that EBSCO says that EBSCOhost Connection is deprecated and there is no way to remap the broken links to new URLs. Would someone be willing to write and run a bot to remove the broken links to EBSCOhost Connection (URLs beginning with from cited sources in Wikipedia articles? Thanks, Biogeographist (talk) 16:48, 19 November 2020 (UTC)

This something for WP:URLREQ. I'd like to do more research and verify there is another way to save them. Worst case are archive URLs so they wouldn't need to be removed. Example. -- GreenC 17:13, 19 November 2020 (UTC)
Copied to Wikipedia:Link rot/URL change requests#EBSCOhost Connection. -- GreenC 17:19, 19 November 2020 (UTC)
I responded at Wikipedia:Link rot/URL change requests § EBSCOhost Connection. Biogeographist (talk) 17:33, 19 November 2020 (UTC)

BAG nomination notice[edit]

Hi! This is a procedural notification that I've requested to join the Bot Approvals Group. Any comments would be appreciated at Wikipedia:Bot Approvals Group/nominations/SD0001. Thanks, – SD0001 (talk) 17:55, 19 November 2020 (UTC)

Contributions says user exists, xtools says no?[edit]

If I go to Special:Contributions/BasketballGOD12, I see four edits for the user. The "Edit count" link at the bottom links to, but when I click that, xtools says, "The requested user does not exist". My brain hurts. -- RoySmith (talk) 19:51, 19 November 2020 (UTC)

@RoySmith: Special:CentralAuth/BasketballGOD12 exists, xtools works off of a replica that is likely behind. — xaosflux Talk 19:54, 19 November 2020 (UTC)

Rendering graphs using {{Template:Graph:Lines}}[edit]

The rendering of the graphs using Template:Graph:Lines is not working properly here. When we click to preview the graph, it looks good, then when we click to save, it changes and looks ugly. Why is this? I have left two example graphs in my user sandbox subpages, one here (User:Bob247/sandbox/graph:line) and one at commons (C:User:Bob247/sandbox/graph:line) for comparison. As we can see it works perfectly over at commons, and the templates are apparently the same at both locations. Note that if you click edit on the examples on this wiki, they instantly look like the ones over at commons. Can someone with knowhow look into this? Thanks. Bob247 (talk) 22:43, 19 November 2020 (UTC)

In preview your browser is generating the graph, but once it is saved the server takes over that task. Some features are supported by the browsers but not the server, hence the uglyness.--Snaevar (talk) 15:31, 20 November 2020 (UTC)
It will be fixed once phab:T236892 is done (it is allready underway).--Snaevar (talk) 15:39, 20 November 2020 (UTC)

Google not rendering Template:sfrac fractions properly from the article Integer[edit]

When Google crawls Wikipedia and displays the first couple of lines of a page, they do not render the sfrac template properly. For example, in the Google search [1], you, on Google, come across the line "For example, 21, 4, 0, and −2048 are integers, while 9.75, 5+, and √2 are not", which is copy-pasted from the Wikipedia article. However, the 5+, with the wiki code of {{sfrac|5|1|2}}, should be rendered by Google as 5 1/2. Can this be fixed by Wikipedia, or it is just Google's fault?. I’ve confirmed this on Google via iOS 14 Safari. Benica11 (talk) 04:38, 20 November 2020 (UTC)

I think Google is doing the best it can with the bizarre code that is output by {{sfrac}}, which looks like this:
<span role="math" class="sfrac  nowrap"><span class="int">5</span><span class="plus visualhide">+</span><span class=" tion" style="display:inline-block; vertical-align:-0.5em; font-size:85%; text-align:center;"><span class="num" style="display:block; line-height:1em; margin:0 0.1em;">1</span><span class="slash visualhide">/</span><span class="den" style="display:block; line-height:1em; margin:0 0.1em; border-top:1px solid;">2</span></span></span>
For some reason, the template includes a hidden plus sign after the first number. Google shows the plus sign. – Jonesey95 (talk) 05:05, 20 November 2020 (UTC)
Yeah. Google would probably be smart enough to ignore it if it was display: none, but it isn't, it's hidden by moving it way off the page. See Template_talk:Frac/Archive_1#Major_accessibility_issue for the apaprent reasoning on why this is done. Apparently for accessibility reasons, I presume for some kind of reader which can parse this as text to speech, some kind of "5 plus one slash two" or something. Personally, I think there's probably more damage than good done here. ProcrastinatingReader (talk) 05:18, 20 November 2020 (UTC)
Maybe the better way to do this is to use aria-hidden ProcrastinatingReader (talk) 05:22, 20 November 2020 (UTC)
The actual section of the talk page you want is Template talk:Frac/Archive 1#Sfrac: accessibility with screen readers. I don't know much about ARIA ... I'm willing to test it out, but I can almost guarantee that screen readers won't do what you expect with it. Graham87 06:01, 20 November 2020 (UTC)
@Graham87 perhaps, but I think a better solution needs to be found, if there is to be one at all. I did a quick search for mixed numbers html/accessibility/screen readers/etc and got no results, which leads me to think we're probably overdoing it. ProcrastinatingReader (talk) 02:16, 21 November 2020 (UTC)
I've adjusted the usage on the article to use TeX instead. Let's see if Google likes this, instead, I guess. ProcrastinatingReader (talk) 02:19, 21 November 2020 (UTC)
Okay, never mind ;p -- reverted by @Anita5192 ProcrastinatingReader (talk) 02:48, 21 November 2020 (UTC)
Accessibility of mathematics content for screen reader users like me is notoriously bad, leading to a lack of blind people taking up mathematical subjects, causing a vicious cycle. I'm not surprised there isn't much discussion of this problem on the Internet. Using LaTeX would have worked here ... Graham87 04:32, 21 November 2020 (UTC)
If the plus wasn't there, the 5 and 1 would be run together and probably read out as "fifty-one". --Redrose64 🌹 (talk) 14:09, 20 November 2020 (UTC)
51/2 is still an example of a non-integer - but of course it can create problems elsewhere. I don't like that style for numbers in general (it can be read as 5*1/2), but that's a different discussion. --mfb (talk) 01:08, 21 November 2020 (UTC)

Accessing special index of WikiData qualifier's value[edit]

Hello friends. I need to access for example publication date of this page. for example I need to print "publication date" of PHP version 7.0.9. I know how to use Wikidata but I can just access the first one and the latest one. In fact I can use "normal" and "preferred" and don't know how to access this version. I used this code:


But this code will return 2020-10-29 that is belongs version 7.4.12 and as you know it should print "22 July 2016" because the publication date of version 7.0.9 is "22 July 2016". I also tried these codes:


but these will show the first (5.3.29) and latest (7.4.12) index. I also don't know how to access version 1.0. so would you give me a hand please? I'm totally new in WP and please speak in simple way and even guide me with links.

Thanks in advance. GameO7er (talk) 14:34, 20 November 2020 (UTC)

You can specify the version via the optional "raw_value" parameter: {{wikidata|qualifier|mdy|raw|P348|7.0.9|P577}} -> 2016-07-22. Note that "raw" and "mdy" together are contradictory, "raw" has preference here. --mfb (talk) 22:41, 20 November 2020 (UTC)


I want to change TimedText:Example so that it displays the same basic format as the commons page at Commons:File:1958-03-17 3rd Vanguard Successful.webm with the three File/Talk/TimedText tabs, the "Available closed captioning." box, and the CC0 "Licensing" box.

Is that possible? I can't figure out how to format the page to show the extra tab, and of course cutting and pasting the Wikitext was useless. Are there any example pages on the English Wikipedia with the all three tabs that I can use as an example? --Guy Macon (talk) 19:59, 20 November 2020 (UTC)

@Guy Macon: the TimedText namespace is for holding structed text that corresponds to a media file (it is "captioning" for it). An example from AllPages that is marked up is at, you can see that this is associated to the media file File:Baby Got Back sample.ogg. These pages should not be marked up with licensing templates in the TimedText namespace. — xaosflux Talk 20:25, 20 November 2020 (UTC)
If TimedText:Example can't be modified to contain an example of timed text, should I put it up for deletion? --Guy Macon (talk) 20:34, 20 November 2020 (UTC)
You should rather put the format from Commons:File:1958-03-17 3rd Vanguard Successful.webm to an example file page. TimedText:Example should contain an example of timecodes instead, which will work. There is no need to delete the timedtext page.--Snaevar (talk) 12:36, 21 November 2020 (UTC)
@Guy Macon: it certainly can contain timed text, I think you are confused about what you are looking at. TimedText namespace holds the actual captioning, you are comparing it to a page in the File namespace (which can be linked to a timed text). — xaosflux Talk 16:20, 21 November 2020 (UTC)
Thanks for the clarification. It does look like I was looking at the wrong thing. That being said, if someone goes to TimedText:Example they should see an example of timed text. Would making the page a redirect to a video with timed text accomplish that? Or maybe just a link to such a video?
Thinking about it, I would like the reader to see 4.2
With timed text "This is" at one second and "a test." at two seconds. --Guy Macon (talk) 17:39, 21 November 2020 (UTC)
Click on the TimedText tab, select a language and press Go, and you'll see instructions there: [2]. Nardog (talk) 18:27, 21 November 2020 (UTC)
Thanks! now has timed text. Now I just need to make it so that someone who goes to TimedText:Example sees that video. Should I create a redirect? Or maybe just a link? --Guy Macon (talk) 19:12, 21 November 2020 (UTC)
I doubt you can create a redirect in that namespace. Besides—again—what the hell is the point of making TimedText:Example a bluelink? Nardog (talk) 21:13, 21 November 2020 (UTC)
Page in the TimedText namespace should contain....timed text in srt format! Not wikilinks, not redirects, etc. If this page isn't actually timedtext for something (such as File:Example.mp3) it should probably just be deleted. If you just want it to be an example of what timed text looks like, but not actually link it to an actual media file, then do that - just put some srt marked up example text in it. — xaosflux Talk 02:20, 22 November 2020 (UTC)
In what circumstances would that be useful? If one needs an example why not just link to an existing .srt page for an actual video? Nardog (talk) 18:23, 21 November 2020 (UTC)

Bug in CharInsert[edit]

Please join the discussion at MediaWiki talk:Edittools § Spaces not possible after cursor operator. —⁠andrybak (talk) 10:43, 21 November 2020 (UTC)


I added Special:Diff/989942531 (CSS class: "autoconfirmed-show") to the sandbox version of protection banners (used by {{pp-move/sandbox}}). This template is present on my sandbox at User:ProcrastinatingReader/sandbox. It caused the green move lock in the top right to disappear. Last I checked, I was autoconfirmed, so I should still see this lock, but I don't? What am I doing wrong here? (paging @Izno and MusikAnimal:) ProcrastinatingReader (talk) 22:07, 21 November 2020 (UTC)

Because MediaWiki:Group-autoconfirmed.css only div, p, span, small, table and li tags with the class, and images are none of these. * Pppery * it has begun... 22:23, 21 November 2020 (UTC)
Ah, right... Is there any particular reason why it's limited to certain tags, and if so is there any issue with adding img and a tags to that list? ProcrastinatingReader (talk) 22:58, 21 November 2020 (UTC)
@ProcrastinatingReader: Hmm we could try putting img as display: inline (with span and small), but I'm not sure if that's a safe assumption. Frankly I don't think the other rules are safe assumptions either, but I didn't invent this system :) For instance, I would say add the CSS class to the #mw-indicator-pp-default element instead, but as a div that would force it to be display:block when it should be display:inline-block. Why are we conditionally showing the padlock, anyway? Shouldn't it be visible to everyone? MusikAnimal talk 23:19, 21 November 2020 (UTC)
@MusikAnimal re your question, see this TfD. TLDR logic is that IPs and non-autoconfirmed can't move anyway, so shouldn't see move protection icon. That's what I was trying to achieve here with the autoconfirmed-show. The diff is just for a quick demo, I was actually thinking of moving it up to the a parent class, but that would suffer from the same issue it seems. Looking into the div... ProcrastinatingReader (talk) 23:25, 21 November 2020 (UTC)
Hmm... The div is controlled by use of the 'indicator' tag. Does this tag support a class attribute? eg like Special:Diff/989953973 ProcrastinatingReader (talk) 23:36, 21 November 2020 (UTC)
@ProcrastinatingReader: Apparently not. I guess there's no harm in adding img to the list, as no one is probably using this CSS class on img tags yet. However, when I tested hiding an img on my userpage (which has many indicators), there's some leftover spacing. Hiding the parent div (.mw-indicator) was the only way I was able to remove it. In most cases this probably doesn't matter, but things could look a bit off for instance on articles that are move-protected and have a GA/FA classification. Would wrapping the entire indicator in a span tag work? MusikAnimal talk 05:19, 22 November 2020 (UTC)
@MusikAnimal Does a span tag around an indicator carry over? eg I did Special:Diff/990015122 but the span just stayed in the body as the indicator was moved to the top by the software (indicator itself isn't wrapped by the span). ProcrastinatingReader (talk) 08:56, 22 November 2020 (UTC)

New versions of Template:merge and Template:split[edit]

I have created new versions of these templates in {{merge/sandbox}} and {{split/sandbox}} which are more similar to each other. Both of these templates have additional reason parameters. Additionally, the new split template is suitable for other namespaces. I tried to avoid making breaking changes, but it is not guaranteed. Also, some related templates may need to be modified. What are your thoughts on these templates? JsfasdF252 (talk) 20:28, 22 November 2020 (UTC)

@JsfasdF252: Try WP:VPT for feedback on ideas, not this page. Headbomb {t · c · p · b} 20:44, 22 November 2020 (UTC)
WP:VPR? – Jonesey95 (talk) 22:38, 22 November 2020 (UTC)
@JsfasdF252: Why are you jumping straight to RfC without, so far as I can tell, having followed WP:RFCBEFORE? Also, please don't put templates in section headings, they make inward linking much more difficult. --Redrose64 🌹 (talk) 22:59, 22 November 2020 (UTC)

placement of dead link template in ref?[edit]

If there is a <ref> that consists of an external link, does {{dead link}} go before the </ref> (inside) or after </ref> (outside)? RJFJR (talk) 22:45, 22 November 2020 (UTC)

inside is preferred as there is no need to know the link is dead in prose text of the article. If someone hovers or clicks the ref then they can learn of its status. No need to inform about the dead link unless the readers shows interest in the source. That said you'll see them on both inside and outside.--Moxy 🍁 22:58, 22 November 2020 (UTC)

Images misaligned on mobile website in infoboxes[edit]

Hi. I don't know why, but on some pages with {{infobox election}} — e.g. 2019 Bolivian general election and 2020–21 United States Senate special election in Georgia — the images are misaligned on the mobile website. I cannot figure out what causes this. It happens on my phone, on my computer in the mobile site, and in screenshots I've seen of at least these two pages from other people. How can this be fixed? One thing I did is I made sure the aspect ratios of the photos in the Bolivian one are all the same, but that changed nothing apparently. DemonDays64 (talk) 07:20, 23 November 2020 (UTC) (please ping on reply)

Categories not showing[edit]

Something weird's going on with Category:Indian computer scientists at the moment - none of the categories are showing. You can add one via Hotcat and that works, but anything that goes via the main edit screen doesn't show - none of the usual tricks like null edits, purging etc seem to work. My only thought is that it's had some revert battles with an IP of late, so that it's maybe been locked somehow? Could someone take a look? TIA. Le Deluge (talk) 14:58, 23 November 2020 (UTC)

Fixed The comment at the top of the page was incorrectly closed. * Pppery * it has begun... 15:02, 23 November 2020 (UTC)
Gah - that was my first thought, I looked at it and didn't see that extra !, so annoying when you know something must be right in front of you and you don't see it. Thanks. Le Deluge (talk) 16:27, 23 November 2020 (UTC)

Nowiki not working?[edit]

Template:Linked is showing some failures of the < nowiki > HTML tag and this is having knock-on consequences by displaying items and populating categories inadvertently. Can anyone work out what's causing this? Timrollpickering (talk) 16:50, 23 November 2020 (UTC)

Not sure what the bug was, but replacing the brackets with HTML entities has fixed it. ‑ Iridescent 17:08, 23 November 2020 (UTC)

Tech News: 2020-48[edit]

17:17, 23 November 2020 (UTC)

Horizontal rule with text in the middle[edit]

I'm doing a redesign of the languages list that appears at the bottom of the main page. I'd like to get it looking a little more like the "Read Wikipedia in your language" list on the global landing page, where the "More than 250,000 articles" etc. headers are in the middle of a horizontal rule (which produces a cleaner visual hierarchy). Could anyone help me code that at the sandbox? {{u|Sdkb}}talk 22:00, 23 November 2020 (UTC)

This Stackoverflow thread looks promising. I think the one on the main page uses JS, but the thread linked here uses CSS, so you might be able to use template styles. – Jonesey95 (talk) 22:43, 23 November 2020 (UTC)
Proof of concept follows, using similar technique to the global landing page; it could use some refinement but it works. {{Nihiltres |talk |edits}} 22:51, 23 November 2020 (UTC)

Welcome to Wikipedia


Discussion at Wikipedia talk:RefToolbar § Autofill access-date button not there[edit]

 You are invited to join the discussion at Wikipedia talk:RefToolbar § Autofill access-date button not there. Any technical help will be greatly appreciated! Funandtrvl (talk) 01:23, 24 November 2020 (UTC)

I think I've identified the issue, but we need an interface admin to make the edit. Ideally sooner rather than later. {{u|Sdkb}}talk 02:04, 24 November 2020 (UTC)
Thank you!! Funandtrvl (talk) 04:02, 24 November 2020 (UTC)