[00:49:43] I've created https://pypi.org/project/wikifunctions because I noticed the name was available [01:01:31] There's also an on-wiki explainer page ([[User:Feeglgeef/wikifunctions-pip]]) [04:55:30] /me [04:55:37] /me [04:55:41] /me [05:49:37] i agree, had the same issue yesterday (re @Csisc1994: A huge problem) [05:59:57] I renamed the function and added aliases (re @u99of9: You can use Z32145. This one has fallbacks if the label is not available in the language they switch to.) [08:17:59] I re-labelled Z32145 again. For improved discoverability and ease of substitution, it seems like a good idea for alternative functions to start with the same string. Whether they should also have a common “tag” alias, like Z32839 does, I’m not sure. Because the alias “section title” matches the label for the default function, it ranks more highly in the selector’s sea [08:17:59] [08:17:59] rch results. Conversely, functions that are less suitable alternatives, especially new ones, should avoid following this pattern (opinion). (re @Npriskorn: I renamed the function and added aliases) [08:56:34] Abstract Wikipedia and the dream of a Universal Language [08:56:35] https://www.youtube.com/watch?v=de-JVxJlVgM&t=954 [09:10:21] @vrandecic what is your thoughts on finishing the technical system all these people have dreamt on and worked on for so many years? [09:10:22] Is this a first man on the moon-scale event for humanity? [09:14:00] If we were to calculate all contributors that worked on any part of the finished system, would that number be as large as the number necessary for the first moon landing? [09:14:01] I'm thinking everyone who contributed to computers science, hardware, software stack including Linux, programming languages (PHP, C, javascript, etc) and the infrastructure it runs on (kubernetes, etc) and finally the application layer (MediaWiki, AW extension, etc) [09:15:36] How many standardized man-years is this combined? [09:15:37] 1M? 100M? [09:20:00] chatgpt estimate the total man years for the moon landing to be 25-60M man-years (https://chatgpt.com/s/t_69cce32911e88191b4f33b88f9f6de72) [09:26:51] chatgpt estimate the total man-years for abstract Wikipedia and everything below it to be 68 million to 215 million man-years. [09:26:52] That's way more than the Apollo mission! [09:42:28] See https://abstract.wikipedia.org/wiki/User:So9q/Impact_assessment [09:48:24] What should we call systems like WD+AW? [09:48:25] Multi-user knowledge storage, representation and rendering system? [09:48:26] Are there plans to release it as a software suite? [11:37:48] btw there are no special demonyms in many languages (re @Gears: Just thinking about Z32822, and realising that while in English there is only one form of the demonym, in other languages there ...) [12:10:04] So it should fall back to the country’s label… or, as is generally the case, dispatch by language first and fall back through the default function specified in the configuration 🤔 (re @OverflowCat: btw there are no special demonyms in many languages) [12:13:12] Basically I mean for an articles "X is a New Yorker" OK, "In English its demonym is New Yorker" OK, but not "New York people are called New Yorkers" (re @Al: So it should fall back to the country’s label… or, as is generally the case, dispatch by language first and fall back through th...) [12:21:05] Right… because it would read as “New Yorkers are called New Yorkers”… or whatever? It’s a good point 🤔 I think that might still be controlled by the language, it’s just that for some languages the result should be empty 🤷‍♂️ In any event, it should be in the basic architecture that any HTML fragment can be suppressed per language on a particular page (or e message> [12:21:07] nabled only for certain languages). I’m inclined to file a ticket. (re @OverflowCat: Basically I mean for an article "X is a New Yorker" OK, "In English its demonym is New Yorker" OK, but not "New York people are ...) [12:45:18] A problem with Z32053. [12:46:00] It renders well on Wikifunctions. But, on Abstract Wikipedia, tag is missing. [12:57:42] This is a good example of something that I think might have caused issues for me too. It would be a good bug report to file. (re @Csisc1994: It renders well on Wikifunctions. But, on Abstract Wikipedia, tag is missing.) [13:12:35] Anyone can spot the problem with Z32992. [13:13:06] I created it now to support citing the Wikidata items of research papers and book chapters. [13:16:15] I found it. I used "Wikidata item" instead of "Wikidata item reference". (re @Csisc1994: Anyone can spot the problem with Z32992.) [13:18:43] It works. [13:19:59] Great, would you like a tip? (re @Csisc1994: I created it now to support citing the Wikidata items of research papers and book chapters.) [13:32:07] Yes. [13:32:15] Of course. [14:17:44] I'm planning to make a Tagalog superlative definition function. If an adjective lexeme doesn't have a superlative form, should it fallback to forming the superlative based on the positive form or lemma of the lexeme, or should every Tagalog adjective lexeme have a superlative form? [14:21:00] How Z32053 is rendered [14:21:25] https://tools-static.wmflabs.org/bridgebot/5aa36fc7/file_79103.jpg [14:22:08] It's difficult to read these. Why not a regular screenshot? (re @Csisc1994: ) [14:22:38] I do not have Telegram Desktop. Give me two minutes. [14:23:28] So whats the difference between the two? (re @Csisc1994: I found it. I used "Wikidata item" instead of "Wikidata item reference".) [14:24:58] https://tools-static.wmflabs.org/bridgebot/a22e0ec5/file_79104.jpg [14:25:05] Here it is (re @Jan_ainali: It's difficult to read these. Why not a regular screenshot?) [14:26:12] +1. (re @OverflowCat: So whats the difference between the two?) [14:26:20] Will be very good to clarify. [14:28:29] The reference is just the Qid, the other one is the whole item. (This could certainly be clearer in the glossary.) [14:29:23] Do you have any solution to fix Z32053. [14:29:50] It works on Wikifunctions but not on Abstract Wikipedia, as shown in the screenshot I just shared. [14:30:17] Cite Q should be ported here [14:32:27] We first need to investigate the oain points that has led to very low adoption eg in enwiki IMO (re @OverflowCat: Cite Q should be ported here) [14:33:04] I miss it every time I add references on wiktionary [14:33:21] Take a look at [[abstract:Q15433043]]. I implemented Cite Wikidata. [14:33:40] Let’s not. And not reproduce mistakes made by all wikis (re @Npriskorn: We first need to investigate the pain points with citeq that has led to very low adoption eg in enwiki IMO) [14:33:50] The idea is simple. The reference is just a link to the Wikidata item. [14:34:13] Wiktionary has a bunch of R templates for books, but they are not suitable for journal articles on a small topic (re @OverflowCat: I miss it every time I add references on wiktionary) [14:35:26] Abstract wiki is a young project, not affected by grumpy opiniated people rejecting things just because they start with a Q (re @Npriskorn: We first need to investigate the pain points with citeq that has led to very low adoption eg in enwiki IMO) [14:35:54] The visual editor does not support it, and help pages says little about it (re @Npriskorn: We first need to investigate the pain points with citeq that has led to very low adoption eg in enwiki IMO) [14:37:24] I propose to use "English simple cite web" and "Cite Wikidata" for references. They are pretty simple. [14:39:07] Also, a problem with the name of the Page when changing languages. It does not change. [14:54:51] The English, Malay, and Bangla functions use a superlative-forming function, while the German function gets the lexeme form that has the superlative (Q1817208) grammatical feature. Currently, there aren't even any Tagalog adjective lexemes, so which approach is recommended? (re @Earldridge Jazzed: I'm planning to make a Tagalog superlative definition [14:54:51] function. If an adjective lex [14:54:52] eme doesn't have a superlative form, should i...) [15:09:15] im not happy with any of them at the moment 🙈 (re @Csisc1994: I propose to use "English simple cite web" and "Cite Wikidata" for references. They are pretty simple.) [15:10:17] I suggest we start with creating the necessary helpers and book/jounal/web functions an after that the top level dispatch function. [15:12:15] A possible pain point is that it doesn't help much when most references in WP is still missing in WD. 🤷 (re @Csisc1994: The idea is simple. The reference is just a link to the Wikidata item.) [15:17:15] It is better than nothing. [15:18:48] I just finished working on this at [[abstract:Q15433043]]. The good thing is that the layout is acceptable across languages. No adaptation is required, beyond solving the tag problem in Cite Web. [15:19:14] yes and we can set up Entitybase when/if this community needs more data than Wikibase can house as an alternative or replacement. [15:19:14] The APIs are very simple and way easier to cache than the Wikibase ones. 🤩 (re @Csisc1994: It is better than nothing.) [15:19:48] We can try it. (re @Npriskorn: yes and we can set up Entitybase when/if this community needs more data than Wikibase can house as an alternative or replacement...) [15:21:18] Wow this is by far the most comprehensive page I have seen so far in AW 😍 (re @Csisc1994: I just finished working on this at [[abstract:Q15433043]]. The good thing is that the layout is acceptable across languages. No ...) [15:22:14] For now I suggest uploading on a case by case basis to WD. Entitybase is still alpha software 😊 (re @Csisc1994: We can try it.) [15:24:11] Still need to work a little bit on Cite Web. Then, I will try to turn the text we have into Abstract one. (re @Npriskorn: For now I suggest uploading on a case by case basis to WD. Entitybase is still alpha software 😊) [15:25:28] at least the title worked 😅🙈 : https://tools-static.wmflabs.org/bridgebot/5c85a7fb/file_79108.jpg [15:25:58] But, if you change the language, it does not change. (re @Npriskorn: at least the title worked 🤣) [15:27:56] It's still mostly static text rather than functions? (re @Csisc1994: But, if you change the language, it does not change.) [15:27:59] Jokes aside. The Orchestrator struggles a lot with a long page like this. It only works in the mornings. [15:28:24] References changed. (re @Jan_ainali: It's still mostly static text rather than functions?) [15:29:12] Turned into monolingual text. Needs to see a little bit the functions in Wikifunctions to decide what I can use to generate the statements. [15:29:53] Also, subtitles are multilingual now. [15:36:55] Al GrounderUK there is a need for inputting multiple qids i think. new function or change this one? Z32839 [15:36:56] e.g. Higher Education And Career (sentence case) [15:40:18] I like this as a start, then one can edit to functions that output something like the goal as seen in the article. [15:40:19] But we are maybe a year or more away from supporting these long and complicated sentences. (re @Jan_ainali: It's still mostly static text rather than functions?) [15:44:01] I think it's kind of backwards and not the way we will want to build articles once there are client wikis using them. But I guess the harm right now is only that it confuses everyone who is new to AW and happen to come across it. (Which in my opinion is something that we, especially right now, should try to avoid. It's difficult to understand as it is.) (re@Nprisko [15:44:01] rn: I like thi [15:44:02] s as a start, then one can edit to functions that output something like the goal as seen in the article. [15:44:04] But we are ma...) [15:45:24] A draft namespace, when created, can be used for something like this (re @Npriskorn: I like this as a start, then one can edit to functions that output something like the goal as seen in the article. [15:45:25] But we are ma...) [15:56:38] I think the “problem” is that only certain websites are permitted. It sounds like the same problem as we were having with Z32878 before I added the www. prefix. In general, I think I prefer the Q approach, which in AW —today— limits the active link to the source item in Wikidata. (re @Csisc1994: I just finished working on this at [[abstract:Q15433043]]. The good thing is [15:56:38] [15:56:38] that the layout is acceptable across languages. No ...) [16:32:15] Are you thinking we only want generated sentences in the main namespace? (re @Feeglgeef: A draft namespace, when created, can be used for something like this) [16:32:36] Yes (re @Npriskorn: Are you thinking we only want generated sentences in the main namespace?) [16:41:52] sounds like a policy draft, would you be willing to add it on wiki? (re @Feeglgeef: Yes) [16:42:24] Not right now, because I don't think it should be policy at the moment [16:42:35] Once we have our first viable article then sure [16:49:43] Do we have criteria established for what a viable article needs? I think that may be useful to document to inspire people to work on The Right Task™️. (re @Feeglgeef: Once we have our first viable article then sure) [17:00:49] viable, best practice draft: [17:00:50] * uses only top-level multilingual functions [17:00:52] * generate all sentences [17:00:53] * has references [17:00:55] * has at least 2 sections and 10 sentences [17:01:14] We don’t. It should at least contain a labelled link to its own Wikidata item, if you ask me. If it’s that minimal, perhaps it should declare itself to be a soft re-direct to Wikidata. Perhaps the first “fragment” on the page should be a hatnote (possibly empty) with the lead paragraph following as the mandatory second “fragment”. (re @Jan_ainali: Do we have criteria [17:01:14] [17:01:14] established for what a viable article needs? I think that may be useful to document to inspire people to wor...) [17:01:59] i created a function for hatnotes of I remember correctly 😀 (re @Al: We don’t. It should at least contain a labelled link to its own Wikidata item, if you ask me. If it’s that minimal, perhaps it s...) [17:04:21] Yes. But it’s just text. Can’t we put it in a box? (re @Npriskorn: i created a function for hatnotes if I remember correctly 😀) [17:05:16] Z30148 sure, it's modeled after enwiki where its just italic text if I remember correctly (re @Al: Yes. But it’s just text. Can’t we put it in a box?) [17:06:46] Feel free to create a new one with the formatting you want and we can rename mine plain or unboxed or whatever. [17:09:24] it currently has no formatting except the class (re @Npriskorn: Z30148 sure, it's modeled after enwiki where its just italic text if I remember correctly) [17:10:23] Nah. I’ve always found the italics intrusive myself, but the community should reach a consensus. I don’t think we want different styling for hatnotes (just for everything except hatnotes and h2) (re @Npriskorn: Feel free to create a new one with the formatting you want and we can rename mine plain or unboxed or whatever.) [17:14:51] The last point seems very ambitious. I would be happy with one section, and then perhaps count the number of factual claims which I think 5 or 6 would be enough (regardless of how many sentences needed for them). (re @Npriskorn: viable, best practice draft: [17:14:52] * starts with a hatnote (empty if not yet needed) [17:14:53] * uses only top-level multilingual functions [17:14:55] * g...) [17:18:03] Personally, I think pages with zero sentences are fine. If there is a sentence, it should generate in more than one language, but how few counts as “viable”, I’m not sure. I’m tempted to suggest “at least one non-European language” and possibly “at least one European language in addition to English (if English is viable)”. (re @Npriskorn: viable, best practice draft: [17:18:04] * starts with a hatnote (empty if not yet needed) [17:18:05] * uses only top-level multilingual functions [17:18:07] * g...) [17:45:46] viable, best practice suggestion: [17:45:47] * starts with a hatnote (empty if not yet needed) [17:45:49] * uses only top-level multilingual functions [17:45:50] * generate all sentences [17:45:52] * has references [17:45:53] * has at least 6 sentences [17:45:55] * the sentences render in a non-european language as well as in a European one besides English. [17:47:18] What about https://abstract.wikipedia.org/wiki/Q300385?uselang=sv? (re @Jan_ainali: The last point seems very ambitious. I would be happy with one section, and then perhaps count the number of factual claims whic...) [17:51:29] The last point is not what I suggested. I’m fine with sentences that render in zero European languages; I just think that if it works in English, it should also have to work in at least one other European language. (re @Npriskorn: viable, best practice suggestion: [17:51:31] * starts with a hatnote (empty if not yet needed) [17:51:32] * uses only top-level multilingual functions...) [17:53:19] Al GrounderUK better? (re @Npriskorn: viable, best practice suggestion: [17:53:20] * starts with a hatnote (empty if not yet needed) [17:53:22] * uses only top-level multilingual functions...) [17:53:49] Turned English Simple Cite Web into a Multilingual "Simple Cite Web". [17:54:06] However, the function "display date" does not seem to work properly. [17:54:53] I am thinking that it could be any two two languages with enough* linguistic difference (so it not just happen to work in a pair of languages that are very similar). [17:54:55] * Defining what enough is is to be determined. [17:58:35] Could work… I’d still add “not including English”, though 🤷‍♂️ (re @Jan_ainali: I am thinking that it could be any two two languages with enough* linguistic difference (so it not just happen to work in a pair...) [18:00:09] You wouldn't be happy with just English and Finnish, for example? (re @Al: Could work… I’d still add “not including English”, though 🤷‍♂️) [18:01:12] Yes, but I don’t understand the third point. (re @Npriskorn: Al GrounderUK better?) [18:03:30] Personally, no. But French and Finnish would be fine. (re @Jan_ainali: You wouldn't be happy with just English and Finnish, for example?) [18:44:19] Perhaps we could say “fundamentally different or only loosely related”? It’s not something we should expect to quibble about at an academic level. We might also consider “at least one of Arabic, Spanish, French, Russian and Chinese” (that is: the official United Nations languages, excluding English and generalising Chinese without political intent). (re @Jan_ainali: I a [18:44:19] [18:44:20] m thinking that it could be any two two languages with enough* linguistic difference (so it not just happen to work in a pair...) [18:53:24] I like that. (re @Al: Perhaps we could say “fundamentally different or only loosely related”? It’s not something we should expect to quibble about at ...) [19:43:37] Thanks for pointing that out. It seems to have regressed with the latest deployment. @Sannita, I know it’s holiday time, but could you let the team know that *T419789 *seems to have regressed. I’ve already commented on the ticket, but the impact is quite widespread even if it is localised to dates (and it may not be). 🙏 (re @Csisc1994: However, the function "display date" [19:43:37] [19:43:38] does not seem to work properly.) [20:22:51] It seems to me that all language-configured functions now fall through to the default function, which basically breaks Abstract Wikipedia [20:22:52] @Sannita @vrandecic @genocation (re @Al: Thanks for pointing that out. It seems to have regressed with the latest deployment. @Sannita, I know it’s holiday time, but cou...) [21:19:57] Geno and I are officially out, especially her for two weeks. I'll send a message to the other engs and see about that [21:20:48] Many thanks for responding 🙏 (re @Sannita: Geno and I are officially out, especially her for two weeks. I'll send a message to the other engs and see about that) [21:23:32] No problem, I'll keep an eye on chat anyway, but don't expect patches during Easter :P [21:35:57] A user is asking me about https://wikifunctions.beta.wmflabs.org which is linked from https://meta.wikimedia.org/wiki/Abstract_Wikipedia/Beta_tutorial. That wiki was shutdown about a year ago with https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/1135100. Could someone update/delete [[m:Abstract Wikipedia/Beta tutorial]] as appropriate? [21:41:52] I'm still really hoping for normality by the 10th. I'm very glad to see the quantity reader working again this morning, but now the display seems broken on the first render (but sometimes works after recollapsing?) (re @u99of9: Do you have a rough timeline? I'm giving a talk on Wikifunctions at Wikicon-Australia on 10-12 April. I really hope that everyth...) [21:48:01] quiddity: is the IRC side of this chat +Rr on purpose? I was testing something and tried to join via web.libera.chat as an anon and got "Cannot join channel (+r) - you need to be logged into your NickServ account". [21:50:02] I’d expect all readers to be defaulting, but that shouldn’t be your issue 🤔 (re @u99of9: I'm still really hoping for normality by the 10th of April. I'm very glad to see the quantity reader working again this morning,...) [21:52:07] …and all displays too, of course. Do you have a sample call? (re @u99of9: I'm still really hoping for normality by the 10th of April. I'm very glad to see the quantity reader working again this morning,...) [21:57:10] I just tried to get one (on my phone) and can't reproduce the display delay. So this may be sufficient for this (untranslated) example. (re @Al: …and all displays too, of course. Do you have a sample call?) [22:00:48] A new reason to make sure our defaults are good! (re @Al: It seems to me that all language-configured functions now fall through to the default function, which basically breaks Abstract ...) [22:02:31] Here’s hoping… no harm in your being 100% SI/ISO compliant, is there 🤞 (re @u99of9: I just tried to get one (on my phone) and can't reproduce the display delay. So this may be sufficient for this (untranslated) e...) [22:11:52] No idea. There is a CAPTCHA requirement on the Telegram side. Is that equivalent? (re @wmtelegram_bot: quiddity: is the IRC side of this chat +Rr on purpose? I was testing something and tried to join via web.libera.chat as ...) [22:13:59] In better news this week… links from Abstract Wikipedia to Wikifunctions now function correctly. It seems the uselang argument is respected until you visit the main page. [22:14:35] bd808: Hmm, I don't recall. I *think* we had significant spam/troll problems that necessitated it, at some point in the past? I do see that a bunch of other channels use +r (Wikidata, Wikipedia-en, Wikimedia-dev, etc), so I'm not sure what the standard is. I'd defer to James_F as the main IRC-expert with ops in here. [22:15:44] Good to see you pop up in here! (re @wmtelegram_bot: bd808: Hmm, I don't recall. I *think* we had significant spam/troll problems that necessitated it, at some point in t...) [22:30:15] No numeric separators, by default, though 🤔 (re @u99of9: A new reason to make sure our defaults are good!) [22:32:15] Oh yes, and I can't show off switching to European radix format. (re @Al: No numeric separators, by default, though 🤔) [23:19:02] We should probably default to U+202F separation by threes for groups of more than four digits. The radix character is a challenge, but I think it has to be . 🤷‍♂️ (re @u99of9: Oh yes, and I can't show off switching to European radix format.) [23:22:44] I’ve disconnected Z14403 expecting that to fix the immediate problem. There may be a performance impact but in a few cases it’s actually quicker to avoid JavaScript. [23:27:45] Progress! : https://tools-static.wmflabs.org/bridgebot/3158f319/file_79126.jpg [23:31:32] Excellent work. (re @Al: Progress!) [23:32:43] Probably, we can do better if someone explains to use how to parse what is fetched using "Fetch Wikidata Item", "Fetch Wikidata Property", or "Fetch Wikidata Lexeme".