[00:55:59] Has someone looked into this (or is it a known bug?): https://www.wikifunctions.org/w/index.php?title=Z32879&action=edit&uselang=en passes the test in edit mode. But once you leave edit mode, it says it fails, even after a nudge on the function itself. [03:57:34] Help! In my talk I plan to show off the wikidata-supported-power of one of my favourite functions Z28509. But I think V2 has caused a problem for Z25785 which now fails all tests. I've reported at T421666. Any diagnosis help for Cory would be much appreciated, because the talk is in about 10 days... [04:13:36] Why is Z32890 returning void for all cases? [04:24:59] You are not supposed to pass a Gregorian calendar month to Z13522 (re @nya_1F616EMO: Why is Z32890 returning void for all cases?) [04:26:28] If you want to check if a given month is February (or any other month), then use Z16137 (re @Earldridge Jazzed: You are not supposed to pass a Gregorian calendar month to Z13522) [04:36:52] I will fix the implementation [05:36:00] Then I wonder why the JavaScript seems to receive a month in the form of an integer indexed from 0. (re @Earldridge Jazzed: You are not supposed to pass a Gregorian calendar month to Z13522) [07:26:56] I think it’s *T406784* but no longer limited in the way it seemed to be? 🤔 (re @u99of9: Has someone looked into this (or is it a known bug?): https://www.wikifunctions.org/w/index.php?title=Z32879&action=edit&uselang...) [07:33:00] I think there is also something funny happening with and show the rest as a text string. (re @Al: I think it’s T406784 but no longer limited in the way it seemed to be? 🤔) [07:35:05] That is what we call conversion to code. Each Type has a specific function to turn it into an in-code object for each supported language. (re @nya_1F616EMO: Then I wonder why the JavaScript seems to receive a month in the form of an integer indexed from 0.) [07:41:34] Seems so, that’s why the href had to change to include www… I don’t know whether the loss is just a rendering effect or a consequence of how the link is deactivated. But when the current version of the composition gets executed, it’s fine. It may have got worse with v2 because of the change of end points 🤔 It’s hard to imagine it’s anything but a split-brain proble [07:41:34] [07:41:35] m 🤷‍♂️ (re @u99of9: I think there is also something funny happening with In this case it is Z16132. They are usually discussed before or during creation of the Type. (re @u99of9: That is what we call conversion to code. Each Type has a specific function to turn it into an in-code object for each supported ...) [07:47:14] @Sannita Could you ask the team to take a look at this and comment on the ticket, please. (re @wikilinksbot: T406784 – When editing a test case, an outdated version of the code for amended implementations may be used for several hours [o...) [07:59:57] I’ve commented on the ticket and amended its title [07:59:58] *T406784* (re @Al: Seems so, that’s why the href had to change to include www… I don’t know whether the loss is just a rendering effect or a conseq...) [08:03:50] Any thoughts on this one? (re @u99of9: Help! In my talk I plan to show off the wikidata-supported-power of one of my favourite functions Z28509. But I think V2 has ca...) [08:05:24] Not yet, but I’m looking at it. (re @u99of9: Any thoughts on this one?) [09:06:18] yep, done (re @Al: @Sannita Could you ask the team to take a look at this and comment on the ticket, please.) [09:10:06] I’m looking at Z19949, which is failing to resolve the denominator, but I can’t see why it would resolve the numerator and then fail with the denominator 🤷‍♂️ (re @u99of9: Any thoughts on this one?) [09:13:05] But in Z20122 it’s the other way around 🤔 (re @Al: I’m looking at Z19949, which is failing to resolve the denominator, but I can’t see why it would resolve the numerator and then ...) [09:43:20] I made an echo-free implementation, which fails less: Z32898. It fails with values like “10.39°”. I haven’t connected it yet in case it derails your investigation. (re @Al: Not yet, but I’m looking at it.) [09:55:03] Error type: Error in evaluation (https://www.wikifunctions.org/view/en/Z507) [09:55:04] Error data: [09:55:05] function call: "invalid literal for int() with base 10: '1047°'" [09:55:07] Something in the Python? [09:55:08] …then maybe the argument reference fails to resolve but we don’t get the error from the function call, only a function call with an unresolved argument reference that successfully propagates into the final result, where it happens to be meaningless. 🤔 (re @Al: I made an echo-free implementation, which fails less: Z32898. It fails with values like “10.39°”. I haven’t [09:55:08] [09:55:10] connected it yet in ...) [10:04:41] I think that one is an edge case that isn't covered by the python yet. I've added it as a test Z32899. But I don't think it is responsible for my earlier blanket failures. (re @Al: Error type: Error in evaluation [10:04:43] Error data: [10:04:44] function call: "invalid literal for int() with base 10: '1047°'" [10:04:46] Something in the...) [10:07:30] Ah, good (as it were). I don’t know why stripping out the outer Echo helps fix things, but shall we connect and nudge? (re @u99of9: I think that one is an edge case that isn't covered by the python yet. I've added it as a test Z32899. But I don't think it is r...) [10:09:10] I think leave it for now until Cory has seen it. (re @Al: Ah, good (as it were). I don’t know why stripping out the outer Echo helps fix things, but shall we connect and nudge?) [13:34:36] I should add that the presence of debugs here makes me think they may be interfering with the eval somehow. (re @u99of9: I think leave it for now until Cory has seen it.) [13:43:12] Or it’s connected with *T421531* somehow: effectively, an error in a function call evaluation within an argument reference leaves the argument reference referring to the function call rather than its resultant error. 🤷‍♂️ (re @u99of9: I should add that the presence of debugs here makes me think they may be interfering with the eval somehow.) [16:06:55] @Sannita would you be willing to ask the team to prioritize this because it has a high negative impact on usability when old versions persists invisibly? [16:06:56] Also I highly recommend you to put a clear warning on the front page listing the most important/ux affecting bugs. [16:06:58] And also it would be nice for us to know which sprint you plan to work on them (bonus) (re @wikilinksbot: T406784 – When editing a test case, an outdated version of the code for amended implementations may be used for several hours [o...) [16:08:10] it has been picked by someone in the team, but they say it's erratic and difficult to replicate, but they're on it (re @Npriskorn: @Sannita would you be willing to ask the team to prioritize this because it has a high negative impact on usability when old ver...) [16:08:17] Would you be willing to help warn other users in some prominent way on-wiki? (re @Jan_ainali: Why no guarantees? It is known to be inconsistent?) [16:09:33] Great to hear, erratic bugs are the most fun to debug 😅 (joke) (re @Sannita: it has been picked by someone in the team, but they say it's erratic and difficult to replicate, but they're on it) [16:09:55] yeah... (re @Npriskorn: Great to hear, erratic bugs are the most fun to debug 😅 (joke)) [16:21:43] i added Z32904 to the helper function and it fails with the same error so something in that Python function is not right or you found a bug 🪲 (re @Al: I made an echo-free implementation, which fails less: Z32898. It fails with values like “10.39°”. I haven’t connected it yet in ...) [16:32:24] I think I fixed it. But then there was a more complicated problem in its companion, so I thought it better to unfix it again 😏 It would be better to pull the composed functions together into a single implementation, otherwise there are three separate traversals into code for each Rational number read… on top of the dispatch by language, which was preferring JavaScript [16:32:25] last t [16:32:25] ime I looked. But, you know, with physical code replication, it’s usually better to get the original right before you start copying it! (re @Npriskorn: i added Z32904 to the helper function and it fails with the same error so something in that Python function is not right or you ...) [16:34:23] I debugged uring chatgpt and it found the error i think. It was caused by the S. [16:34:25] https://chatgpt.com/s/t_69caa5e0777c81918c846e3c842504d6 (re @Al: I think I fixed it. But then there was a more complicated problem in its companion, so I thought it better to unfix it again 😏 I...) [16:35:53] maybe a series of composition helper functions would be better? [16:35:53] clean brackets from string, etc (re @Al: I think I fixed it. But then there was a more complicated problem in its companion, so I thought it better to unfix it again 😏 I...) [16:40:32] Not if all the helpers are ultimately implemented in code, which all string operations currently are, of necessity. I’ve raised this on *T420333*. (re @Npriskorn: maybe a series of composition helper functions would be better? [16:40:34] clean brackets from string, etc) [16:44:59] (Its title was changed from “Z886 and Z868 have no connected implementations”; my hope being that they would work again.) (re @wikilinksbot: T420333 – Z886 and Z868 have no connected implementations; should we just drop them instead of leaving them bereft? [open]) [17:19:26] Q25169 - Abstract Wikipedia [17:19:28] https://abstract.wikipedia.org/wiki/Q25169 [17:20:25] I just found this! I have missed the launch, a whole new playground to perouse 🤩♥️ [17:23:15] @Jan_ainali it seems we have work to do 😅 : https://tools-static.wmflabs.org/bridgebot/98ad6fb7/file_79012.jpg [17:30:35] Are you already on it, or should I give it a go? (re @Npriskorn: @Jan_ainali it seems we have work to do 😅) [18:57:39] go ahead 😀 (re @Jan_ainali: Are you already on it, or should I give it a go?) [18:59:09] @Sannita I just discovered the improvement to the html output where rendering is now togglable 🤩 [18:59:10] Nice! Please send my thanks to the team 😀 [19:04:36] Will do! But the team is also in the channel, so they read you :) (re @Npriskorn: @Sannita I just discovered the improvement to the html output where rendering is now togglable 🤩 [19:04:37] Nice! Please send my thanks to ...) [19:04:45] And sometimes they also interact [19:08:38] @vrandecic I absolutely love the new reference functions 😍 [19:16:02] With the current capabilities and the new copy paste UI for compositions I think we now have everything to start making more complex cite web|journal|etc functions. [19:16:04] I'm so exited to get started 🤩 [19:16:05] Imagine having a complete feature rich set of cite functions for AW 🤩 [19:16:07] Maybe we should start with CiteQ? [19:16:08] It has been criticized by enwiki people, but I can't remember the main pain points. [19:16:10] Does anyone want to collaborate on citation functions? [19:16:19] Hmm. I was hoping to just copy-paste Z32858 and changing one of the text strings. But in Z32910 I get some type of error that I don't get at the original. although I didn't change anything else than specifying the arguments and that text string. What happens here? (re @Jan_ainali: Are you already on it, or should I give it a go?) [19:25:32] Ah, this time it was missing a lexeme, but I couldn't figure that out from the error message. Now it works! 🎉 (re @Jan_ainali: Hmm. I was hoping to just copy-paste Z32858 and changing one of the text strings. But in Z32910 I get some type of error that I ...) [19:34:41] Could the (mul) be presented in another way so one might not think it is part of the rendered text? : https://tools-static.wmflabs.org/bridgebot/13330109/file_79034.jpg [19:35:52] I'm afraid that this might also make people think they should go add a Swedish label on the item... [20:47:17] That's a slightly incorrect version of Z32899. These numerators are all positive. The sign is handled separately. So the "S" is not the problem. It's the degree sign being considered part of the number because it is before the space. But again, this python edge case only applies to rare examples. The main problem applies to all quantity reads. (re @Npriskorn: i added [20:47:17] Z32904 to th [20:47:17] e helper function and it fails with the same error so something in that Python function is not right or you ...) [21:15:49] Yes. There are limited options while the construction is in text, rather than HTML. We could just return it as text, or with any constant string before and/or after the foreign text. What we can’t do is decide whether or not an item should have a label in the specified language, because all we have is a list of labels. We should consider providing a formatting option that [21:15:49] enume [21:15:50] rates all the available ways that referenced items can be styled: bold, with AW links, WD links for missing labels and so forth. We can start that now with parallel functions, but then every sentence-fragment is multiplied by all the styling options (and, dare I add, permutations thereof, but not separately for each language, at least). (re @Jan_ainali: Could the (mul) be [21:15:50] present [21:15:52] ed in another way so one might not think it is part of the rendered text?) [21:21:17] It only appears where previously an empty string would have been returned. It’s an extreme fallback; more sophisticated alternatives should be applied before, or instead of, calling this function. (re @Jan_ainali: I'm afraid that this might also make people think they should go add a Swedish label on the item...) [21:23:29] Hmm, I don't think we should consider it an extreme fallback. Rather, it should be the normal use case for a lot of concepts (for example, for a lot of human name/language combinations). [21:43:41] Mul is rather special. I would support leaving off the prefix just for mul. (re @Al: Yes. There are limited options while the construction is in text, rather than HTML. We could just return it as text, or with any...) [21:46:16] I think `Z23468` was always intended to return the label for a particular language. [21:46:17] *About [21:46:19] English [21:46:20] *text from Wikidata item label, for given language [21:46:22] returns the label string of the Wikidata item in the specified language [21:46:23] Returning an empty string rather than an error was a later enhancement, and returning a (mul) label or, failing that, an (en) label was the latest enhancement. [21:46:25] Z24766 is a function you can use if you want tailored fallbacks. If (as seems perfectly reasonable) you want to see values from name or title statements, for some classes of item, you would have to use a different function again, like Z29057 does, for example. (https [21:46:25] //www.wikifunctions.org/view/en/Z29057?call=%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z29057%22%2C%22Z29057K1%22% [21:46:26] 3A%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z6821%22%2C%22Z6821K1%22%3A%7B%22Z1K1%22%3A%22Z6091%22%2C%22Z6091K1%22%3A%22Q3107329%22%7D%7D%2C%22Z29057K2%22%3A%22Z1002%22%7D) (re @Jan_ainali: Hmm, I don't think we should consider it an extreme fallback. Rather, it should be the normal use case for a lot of concepts (fo...) [21:56:25] I’m not entirely opposed to that; it just means we’ll need a new function that does what this one currently does (or something similar). (re @u99of9: Mul is rather special. I would support leaving off the prefix just for mul. Or at least having a formatting version which works ...) [22:07:04] There is a spectrum of ways we can deal with errors and missing data. 1. Throwing exceptions will rarely help an AW user. 2. Returning blank text is stark, but may sometimes go unnoticed. 3. Returning annotated labels from other languages may prompt WD attention. 4. Returning un-annotated labels from close variants will usually just flow, and is necessary when WD [22:07:05] doesn't even su [22:07:05] pport that variant. 5. Returning dodgy sentences in the right language may prompt WF attention. [22:07:54] I think "mul" may often work as #4 (re @u99of9: There is a spectrum of ways we can deal with errors and missing data. 1. Throwing exceptions will rarely help an AW user. 2. Re...) [22:09:57] When we set up language configs, I think we should aim for #5 in the default function. (re @u99of9: There is a spectrum of ways we can deal with errors and missing data. 1. Throwing exceptions will rarely help an AW user. 2. Re...) [22:32:07] 1. Agreed [22:32:08] 2. Always best avoided in HTML fragments [22:32:10] 3. Yes. [22:32:11] 4. I think it depends, but that’s what the fallback route is for. [22:32:13] 5. Generally, yes, but then individual languages should be able to opt out explicitly in the article call (this applies to every fragment, though). [22:32:14] Default functions should provide linked content following their global function’s pattern. This is currently blocked by functions’ being specified as returning Z11 rather than Z89. These links would always require some kind of label, even if it’s a QID. (re @u99of9: There is a spectrum of ways we can deal with errors and missing data. 1. Throwing exceptions will rarely [22:32:14] help [22:32:16] an AW user. 2. Re...) [23:56:51] (copying from the abstractwiki main page talk bc i havent gotten an answer there yet) On en:Main Page, there's a counter for active users and articles. Would anyone object if I were to add the counter to this page and the accompanying css to Template:Main page/styles.css? [23:59:15] I think that'd be good