[01:18:50] When a configured monolingual text function uses a default function from its configuration as a fallback, the output is often not very purely monolingual. In those circumstances, does anyone see problems with returning that result with special language code (perhaps mul, zxx, ...which?)? That would help us to then annotate it better in the HTML wrapping function (e.g. [01:18:50] HenkvD's p [01:18:51] roposal to render it in magenta, or to add an edit pencil pointing to the configuration which needs that language option). [01:49:26] Seems so (re @Jan_ainali: The way we have abstract articles now, we can only have one version per topic, right? So if we want a different version about a ...) [02:43:51] from [[:wikidata:Help:Monolingual_text_languages#Special_language_codes]] it looks like zxx (not applicable) might be best? (re @u99of9: When a configured monolingual text function uses a default function from its configuration as a fallback, the output is often no...) [02:47:51] Can someone help me out and explain why Z37442 returns void on every call? I've been trying for a bit but can't seem to figure it out. [03:55:13] the function is supposed to return an HTML fragment, but most of the if-else options return a string. (re @M: Can someone help me out and explain why Z37442 returns void on every call? I've been trying for a bit but can't seem to figure i...) [04:43:02] It seems as if Z36691 just isn't working in the composition, therefore it goes to the fallback string (where the same issue happens), but when I run Z36691 manually using the same parameters in the testcases it works perfectly. [04:47:37] I investigated briefly but also haven't found the issue yet. (re @M: Can someone help me out and explain why Z37442 returns void on every call? I've been trying for a bit but can't seem to figure i...) [04:54:27] The if calls Z32556 and not Z36691 directly? [04:56:09] "measuring update load, exploring ways to make NLG functions faster and more resilient, and making rendered fragments more durable to improve the experience for readers and editors" this all sounds good, but I fear it may not cover the main problem at the moment which is that AW sends all fragment requests at once, the orchestrator/evaluator/renderer gets flooded, the [04:56:09] cache saves [04:56:09] all the output errors for a day until the next barrage. Should there be some kind of request queue? (re @vrandecic: Abstract Newsletter Nr. 256 [04:56:11] * Moving toward our first Abstract Wikipedia integration milestone [04:56:12] ** Supporting the creation of Ab...) [05:03:19] My assumption is that Z32556 will call the function which exists in its "expected string" class, and if that function returns a malformed string, then it will resort to the fallback string. Am I incorrect? (re @adafede: The if calls Z32556 and not Z36691 directly?) [05:04:11] I say class, I mean parameter [05:08:25] Hmmm true misread it [07:47:35] or a Voyage: namespace on abstract.wikipedia — that might be a bit easier to set up (re @Jan_ainali: The way we have abstract articles now, we can only have one version per topic, right? So if we want a different version about a ...) [08:25:47] That is a clever idea. I agree. (re @vrandecic: or a Voyage: namespace on abstract.wikipedia — that might be a bit easier to set up) [08:26:14] We can also fully automate WikiSpecies and move it to a Species Namespace too. [08:26:59] That's something I would like to see the WikiSpecies community to ask for, and not do unto them [08:46:36] `species` is an interwiki prefix for Wikispecies so it would conflict [10:11:21] gengh has now engaged the key task on Phab. Thanks. (re @u99of9: "measuring update load, exploring ways to make NLG functions faster and more resilient, and making rendered fragments more durab...) [12:14:54] You forgot to fetch the item (re @M: Can someone help me out and explain why Z37442 returns void on every call? I've been trying for a bit but can't seem to figure i...) [13:04:42] I’d like us to be clear about what the HTML result would look like, as an eventual target. If there is more than one entity referenced, the languages for each may be different, like [13:04:44] (mul)Paris • (??)capital • (fr)France [13:04:45] That string isn’t what we want, but it’s either a Z6 or (mul)Z11 according to the outer function’s return type. On the other hand, if the three language tags are identical, it can be a Z11 with that tag. The fuzziness between those cases is where the problems lurk, and I think it would be better to let the default function wrap a function that produces the intended HTML, [13:04:46] co [13:04:47] ercing it (temporarily) to the current return type. The last time I thought about this, I came up with [[Wikifunctions:Type proposals/HTML fragment structure]], which is still only an early draft, I’m afraid. I mention it here because I think we should avoid trying to coerce actual HTML fragments and, in particular, avoid the implicit parsing. (re @u99of9: When a [13:04:47] configured mon [13:04:49] olingual text function uses a default function from its configuration as a fallback, the output is often no...) [13:05:25] This is on the way to working. Here's how a default output now shows in Japanese. https://abstract.wikipedia.org/wiki/Q7343 The spanner (not pencil) links to the configuration table as a call to action to add a Japanese configuration. https://abstract.wikipedia.org/wiki/Q7343https://abstract.wikipedia.org/wiki/Q7343https://abstract.wikipedia.org/wiki/Q7343 : [13:05:25] https://tools-static.wmflabs.org/bridgebot/e08d8bd4/file_81646.jpg [13:22:36] I don't know what we'll eventually want the HTML to look like. If you know, I'm probably happy to follow. For now I test stuff to see what works best. I'm less concerned about the exact language spans, and don't mind if a non-configured fragment ends up looking like a (mul)Z11. I don't really see much point of showing the user what language the bad label came from - [13:22:36] they can alre [13:22:37] ady see it's bad. But it is important to show them where and how to fix it (the pencils to Wikidata). The bad sentence structure comes from the default function, so here (the spanner) shows them where to fix that problem. I'm not up to thinking about lexeme problems yet. I don't think I understand your coercing proposal, can you show an example of what you want (maybe [13:22:37] a test on a [13:22:38] default function somewhere)? (re @Al: I’d like us to be clear about what the HTML result would look like, as an eventual target. If there is more than one entity refe...) [13:30:05] PS, as part of making the spanner link work, at Z37464 I had to use a manually typed and unvalidated string to represent the ZID. I couldn't search for the config I wanted. Before this gets built-in too far, it would be really nice to have [[Wikifunctions:Type_proposals/Wikifunctions_object_reference]] soon, even if it doesn't have an associated selector ?chip yet. It [13:30:05] seems like [13:30:07] a simple structure. Feeglgeef I don't care about the name, I don't think this should hold us back. @vrandecic is there any obstacle to this? (re @u99of9: This is on the way to working. Here's how a default output now shows in Japanese. https://abstract.wikipedia.org/wiki/Q7343 The...) [13:43:12] Or if Z9 is now "working" (per T405837), is that equivalent? Should I be using that Al GrounderUK ? (re @u99of9: PS, as part of making the spanner link work, at Z37464 I had to use a manually typed and unvalidated string to represent the ZID...) [13:43:53] I'm putting it on top of my list of types to look at, thanks! (re @u99of9: PS, as part of making the spanner link work, at Z37464 I had to use a manually typed and unvalidated string to represent the ZID...) [13:49:24] I’m not sure what your problem was, but does Z29102 provide a solution? You can also use Z29267 and unquote the reference only when required. (re @u99of9: Or if Z9 is now "working" (per T405837), is that equivalent? Should I be using that Al GrounderUK ?) [13:53:01] The second argument of Z37464 shouldn't ask for any old string, it should ask for a Wikifunctions object reference. Later when there is a selector, users could search for the name of a configuration object. (re @Al: I’m not sure what your problem was, but does Z29102 provide a solution? You can also use Z29267 and unquote the reference only w...) [13:56:38] Or can I just set the argument's type as Z9 then use Z29102 when constructing the link? What is the difference between a Z9 and your Type proposal? (re @u99of9: The second argument of Z37464 shouldn't ask for any old string, it should ask for a Wikifunctions object reference. Later when t...) [14:03:49] Right… then we’d still have the problem with those objects that lack an identity. That’s a data-gap: the supplied reference is de-referenced to an object from which the supplied reference is irrecoverable. The proposed type wraps the reference so that it doesn’t get de-referenced in the first place. (re @u99of9: The second argument of Z37464 shouldn't ask for any old stri [14:03:49] [14:03:50] ng, it should ask for a Wikifunctions object reference. Later when t...) [14:05:09] I don’t think Z9 is a viable argument type. (re @u99of9: Or can I just set the argument's type as Z9 then use Z29102 when constructing the link? What is the difference between a Z9 and ...) [14:05:14] Yes, that makes sense. I definitely don't want this argument dereferenced. (re @Al: Right… then we’d still have the problem with those objects that lack an identity. That’s a data-gap: the supplied reference is d...) [14:24:04] Yes, it’s more: the supplied ZID (quoted) iff [type of object] is Z14294? That looks doable within the implementation. Shall I give it a go? It could always link to the configuration list if the supplied ZID is not valid, I suppose, or the multilingual outer function? (re @u99of9: Yes, that makes sense. I definitely don't want this argument dereferenced.) [14:33:10] Since Denny is willing to consider the type, maybe we should wait and use that? But feel free to have a play if you like. It's used in about 3 compositions, so any change needs a bit of carry over work. 🎃 time for me. (re @Al: Yes, it’s more: the supplied ZID (quoted) iff [type of object] is Z14294? That looks doable within the implementation. Shall I g...) [14:34:28] I’ll add it to my backlog 😏 (re @u99of9: Since Denny is willing to consider the type, maybe we should wait and use that? But feel free to have a play if you like. It's u...)