[00:01:01] Okay, but the Rational Number type proposal is for a type that is congruent with the set of rational numbers, including all floating point numbers (excluding infinities, signed zeroes and NaN) [00:01:29] Yeah [00:01:37] Floats have *most* [00:02:00] Which is why they are *ok* for the mathematical representation [00:02:11] But for DevEx [00:02:16] For UserEx [00:02:21] For simplicity [00:02:28] It all is very bad [00:02:37] But float does well in those categories [00:13:08] So long as float isn’t used to represent rationals, I’m happy to let that type look after itself 😎 (re @Feeglgeef: But float does well in those categories) [00:13:27] Yeah (re @Al: So long as float isn’t used to represent rationals, I’m happy to let that type look after itself 😎) [00:13:29] Obviously [00:13:36] I just think this should come first [00:13:55] Because of its use in programming languages and simplicity [00:18:07] Rationals are really simple in Python. As are decimals. I’m guessing we’ll get one float type before we get a third programming language… so is it just JavaScript you’re worried about? (re @Feeglgeef: Because of its use in programming languages and simplicity) [00:18:37] If by float, you mean the Float64 proposal as it currently stands, I suggest you go ahead and support that. But I suppose that the WF representation of floats in general has more decisions to make than the decision in Rational about where to put the sign. Which string representation of floats would you store? Do we need to consider qNan and sNan? It seems there are lots [00:18:37] of subtle [00:18:37] ties in https://en.wikipedia.org/wiki/IEEE_754 but I am not an expert. (re @Feeglgeef: I just think this should come first) [00:19:18] Whatever both JS and Python support [00:19:31] Less devs are familiar with this (re @Al: Rationals are really simple in Python. As are decimals. I’m guessing we’ll get one float type before we get a third programming ...) [00:19:47] I expect we will get both Rational and Float* (and maybe more) before a new programming language! (re @Al: Rationals are really simple in Python. As are decimals. I’m guessing we’ll get one float type before we get a third programming ...) [00:20:12] My goal with floats is to approach the programming languages where they are (re @Toby: If by float, you mean the Float64 proposal as it currently stands, I suggest you go ahead and support that. But I suppose that t...) [00:20:25] Yeah, typical British understatement 😎 (re @Toby: I expect we will get both Rational and Float* (and maybe more) before a new programming language!) [00:20:33] And where devs will be familiar with them [00:23:02] Sure, I'm fine with that goal for floats. But does the Float64 proposal get this right yet? I'm confident that Rational is completely ready to go, no matter which way the sign issue goes. (re @Feeglgeef: My goal with floats is to approach the programming languages where they are) [00:23:46] My proposal may be different from the Float64 proposal (re @Toby: Sure, I'm fine with that goal for floats. But does the Float64 proposal get this right yet? I'm confident that Rational is compl...) [00:23:56] But it's literally a string [00:24:02] Nothing else [00:24:11] Kinda like Natural Number [00:24:12] what are the options for that string? [00:24:27] Whatever JS will allow you to input and process (re @Toby: what are the options for that string?) [00:24:44] If we eval it in JS and it works it's valid [00:24:54] We shouldn't actually eval it [00:24:57] Of couree [00:25:00] But [00:25:10] That's what the conditions will be [00:25:43] so, "2+2" is a valid FeeglgreefFloat? (re @Feeglgeef: If we eval it in JS and it works it's valid) [00:26:04] What? [00:26:05] No [00:26:13] Not an expression in JS [00:26:18] A *value* in JS [00:26:27] 2+2 is an expression [00:26:32] Not a value [00:26:49] NaN is a value [00:28:12] How do I tell whether a string is a valid value? (re @Feeglgeef: A *value* in JS) [00:28:31] [-+]?(?:(?:NaN|Infinity)|(?:\d+\.?\d*|\.\d+)(?:[eE][-+]?\d+)?) [00:28:43] Looks about right [00:31:49] ^-?(NaN|Infinity|((\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?))$ [00:31:52] Is better [00:32:13] Should cover everything [00:32:22] Lemme know if it's missing something [00:33:33] I'll try, but I really have no idea, which is one reason why I think we need to ponder this proposal longer than Rational. (re @Feeglgeef: Lemme know if it's missing something) [00:34:02] I mean the Regex is what I personally would propose (re @Toby: I'll try, but I really have no idea, which is one reason why I think we need to ponder this proposal longer than Rational.) [00:34:15] If anyone has any objections we should sort those out [00:34:23] But otherwise the current state is fine [00:34:37] And if we later find out it's wrong it's a really really easy fix [00:34:53] We just have to change the Regex [00:35:10] Which I think everyone has the rights for [00:37:58] I also think that functioneers should get Type creation rights ASAP so we don't need to have this problem again [00:41:06] How many Types would you have already created by now? ;-) (re @Feeglgeef: I also think that functioneers should get Type creation rights ASAP so we don't need to have this problem again) [00:41:18] At least 10 (re @Toby: How many Types would you have already created by now? ;-)) [00:41:26] Since I got the right [00:41:30] 5 days afot [00:41:32] Ago [00:41:49] haha, yes that is what I thought... [00:42:09] and how many would I have objected to? [00:42:18] Well [00:42:23] Like a few? [00:42:24] 11? (re @Toby: and how many would I have objected to?) [00:42:32] Lmfao (re @Al: 11?) [00:42:54] Most would be on the proposals page (re @Toby: and how many would I have objected to?) [00:43:06] A few that are not on there that I personally support [00:43:29] Actually mainly the ones on there [00:44:01] The thing is most of them I would have created impulsively and without discussion [00:44:30] So we should probably require discussion by policy before Functioneers are allowed to create them [00:45:12] I just had a play with this in regex101. So this is unlimited precision? When you try to process in a specific language does that toggle between *32 *64 *128 etc? Or does it just force it into *64 and fail when given something like "1e1000"? (re @Feeglgeef: ^-?(NaN|Infinity|((\d+(\.\d*)?|\.\d+)([eE][+-]?\d+)?))$) [00:47:57] I didn't account for that (re @Toby: I just had a play with this in regex101. So this is unlimited precision? When you try to process in a specific language does tha...) [00:48:05] Yes, the proposal process is supposed to be used, but the developers have been known to create some we didn't even know were on the horizon. If you haven't looked under the carpet yet, check out things like: Z14294 (re @Feeglgeef: So we should probably require discussion by policy before Functioneers are allowed to create them) [00:48:41] I was thinking of a more immediate discussion (re @Toby: Yes, the proposal process is supposed to be used, but the developers have been known to create some we didn't even know were on ...) [00:49:04] Like 2 days for creation imminent and these are the exact details that will not change [00:49:46] Sort of like a request for rights but instead for the creation of a type [00:50:12] While the type proposals are pretty specific they are not exact [00:50:41] These discussions would ideally have the exact code that would be used in the type [00:52:42] Or we could just not have discussions but instead limit the amount created per user and sitewide [00:53:05] Like 1 sitewide/day and 1/user/week [00:53:32] Or we could just have no limits [00:53:50] @vrandecic what do you think? [01:11:08] When I click on the links in [[Wikifunctions:Catalogue/List_operations]] they all take me to the Deutsch version. [01:12:38] Same for me [01:12:45] I don't know what would cause that [01:12:52] They are normal Z+s [01:13:36] Actually seems to be happening everywhere [01:14:52] Wh [01:15:08] It's inconsistent for me [01:15:16] Sometimes it happens [01:18:35] I'm ok with Z19551 being deleted [01:18:50] I just couldn't find it while searching [01:19:00] Z13155 should have more aliases [02:25:46] @vrandecic summary of the above conversation: Toby wants the rational number type to be created first, because of its mathematical accuracy. I want the float64 type to be created first because of its simplicity and developer experience. My hope is that both of these can be created in the very near future (hopefully, the type creation process is not that [02:25:46] complicated). You said tha [02:25:48] t one could be implemented early next week. Could we get both by the end of November? I'd be happy with that. [02:27:26] I'm going to create my own float proposal, heavily based on the current one, but with a few tweaks. [02:29:42] Actually the existing one seems to be fine! I'll add a support vote [02:35:18] @vrandecic also if you've got bandwidth to improve types, it might help to finalise and officially release Typed pairs. Once they are in place, along with other uses, I think I could build a faster implementation of Z19510 which may not timeout. (re @Al: Their support needs more work. I haven’t dug into it but, you know… architecturally it seems right 😏 We just convert [02:35:18] each elemen...) [02:41:01] I made Z19571 as a sample validator for floats [02:41:20] It's tested and has one implementation [03:15:50] I submitted an RfA. If I'm being honest it probably will fail, but, I'm trying anyway :) [05:14:38] Z195783 I made as a sample validator for rational numbers [05:14:56] Z19583* [07:18:11] I could add an aggregated list, where every page is included. (re @Nicolas: the only downside is that we can't do a ctrl+f on all the content anymore) [07:32:59] I've started with an equality function Z19586. I needed ScienceD90's Z17534 just to construct the pairs in the Tests! I couldn't specify them from the interface. (re @Al: Their support needs more work. I haven’t dug into it but, you know… architecturally it seems right 😏 We just convert each elemen...) [08:25:44] 😏 I was assuming that was intentional 😎 (re @Toby: I've started with an equality function Z19586. I needed ScienceD90's Z17534 just to construct the pairs in the Tests! I couldn't...) [09:07:32] I'm also confused at why Z19588 is failing. When I put what I think is the exact same input into the function page, it works... (re @Al: 😏 I was assuming that was intentional 😎) [10:05:09] Yeah… eventually. 8684 ms for me 😮 It’s an odd error but it may be a timeout in the make pair function 🤷‍♂️ [12:07:02] Typed Pair is a generic type, and it will take a while before we properly get to those. We are working right now on a document to understand and plan things out, but don't expect typed pair to show up soon. (re @Toby: @vrandecic also if you've got bandwidth to improve types, it might help to finalise and officially release Typed pairs. Once the...) [12:09:13] Thanks for the summary! I understand the (precise) rational type, and still plan to pick that one up for next week, given the conversation (otherwise I would have gone for the data-related ones). Regarding float64 proposal, I am convinced of its use and need, but personally not yet of the current proposal. I will put my questions on the [[Wikifunctions:Type [12:09:13] proposals/float64]] pa [12:09:13] ge. (re @Feeglgeef: @vrandecic summary of the above conversation: Toby wants the rational number type to be created first, because of its mathematic...) [12:17:17] Wikidata-related ones are of course also very interesting. (re @vrandecic: Thanks for the summary! I understand the (precise) rational type, and still plan to pick that one up for next week, given the co...) [12:21:55] Did you mean “date-related”? I was thinking of dates as my second priority, after rationals… (re @vrandecic: Thanks for the summary! I understand the (precise) rational type, and still plan to pick that one up for next week, given the co...) [12:22:37] Yes, date, not data [12:22:51] We are working the the Wikidata-related ones separately as well. [12:58:11] Newsletter 179 [12:58:12] * The Dream of a Universal Language [12:58:13] * Catalogue has been refactored [12:58:15] * Recent Changes in the software [12:58:16] * Presentation series on lexicographic data and its usage [12:58:18] * Recording of November Volunteers’ Corner [12:58:19] * Lighting talk at Wikiconvention francophone [12:58:21] * Upcoming types [12:58:22] * Function of the Week: simple tense of verb [12:58:24] https://www.wikifunctions.org/wiki/Wikifunctions:Status_updates/2024-11-07 [12:58:25] That was a long one! [16:12:17] I am in favor of a future tense type (re @vrandecic: Newsletter 179 [16:12:18] * The Dream of a Universal Language [16:12:19] * Catalogue has been refactored [16:12:21] * Recent Changes in the software [16:12:22] * Presentati...) [16:12:28] As was mentioned in the Status Update [16:12:57] It should have every tense that appears in any Wikilambda language [17:25:07] I have a rough idea of what you mean, a type of all tenses -- but what are the use cases? Which functions would you write with that? [17:25:28] Well [17:25:47] I obviously the featured one would be improved [17:25:56] And [17:26:07] There are other functions that will inevitably require this [17:26:16] For abstract wikipedia [17:26:23] I'm not saying in the near future [17:26:28] In the distant future [17:26:54] In the near future we have much higher priorities in typed to be created [17:27:15] I am also in favor of some kind of typed list dropdown type [17:27:24] If reasonably possible [17:27:33] To allow dropdown inputs [17:27:39] Like in simple tense verb [17:30:07] maybe. I'll keep pondering this. [17:30:19] Alright [17:30:22] is there any good example of functions using other functions inside? I want to show some to te students in a couple of weeks [17:30:33] All: I could use some input on the rational number type, if we want to have this set up next week: https://www.wikifunctions.org/wiki/Wikifunctions:Type_proposals/Rational_number#Comments [17:30:35] The metric function one (re @Galder: is there any good example of functions using other functions inside? I want to show some to te students in a couple of weeks) [17:30:55] Toby made it if IIRC (re @Feeglgeef: The metric function one) [17:31:00] It times out tho [17:31:09] And run unary endofunction [17:32:22] Z17958 (re @Galder: is there any good example of functions using other functions inside? I want to show some to te students in a couple of weeks) [17:33:15] I've already commented (re @vrandecic: All: I could use some input on the rational number type, if we want to have this set up next week: https://www.wikifunctions.org...) [17:35:44] I was looking more on a Python function that loads another function. Because there may be some commons things (like shared exceptions) (re @Feeglgeef: And run unary endofunction) [17:36:08] This currently is not possible (re @Galder: I was looking more on a Python function that loads another function. Because there may be some commons things (like shared excep...) [17:36:19] Or I don't get what your saying [17:36:27] you're* [17:36:30] Any function using another function would be a composition, otherwise it is not possible [17:36:57] I thought you meant a function that requires a function as input (re @Galder: I was looking more on a Python function that loads another function. Because there may be some commons things (like shared excep...) [17:37:17] In which case Z17958 is my favorite [17:38:18] https://www.wikifunctions.org/view/en/Z18579 has a bunch of exceptions that are the same for most of the declensions [17:38:43] we could have it centralized and coherent in someway, but I don't know if it is possible to nest functions (I don't know how to explain that better) [17:39:24] It is not possible to call another Wikifunctions function from a programming language implementation (re @Galder: we could have it centralized and coherent in someway, but I don't know if it is possible to nest functions (I don't know how to ...) [17:39:35] If you are using compositions you can do so [17:40:23] If you'd like to have something standard across all basque functions you would need to exclusively use compositions [17:40:29] or a function that first checks if the lexeme is an adjective before making a superlative [17:40:58] Yes we can’t mix composition and codes together (re @Feeglgeef: It is not possible to call another Wikifunctions function from a programming language implementation) [17:41:11] Yeah (re @cvictorovich: Yes we can’t mix composition and codes together) [17:41:12] Galder: here's one Z12735 -- something like that? [17:41:17] But indeed we may need that [17:41:24] because I can make "dark" > "darker" but I shouldn't be able to make "table" > "tabler", which is not a valid word [17:41:44] Check against word types (re @Galder: because I can make "dark" > "darker" but I shouldn't be able to make "table" > "tabler", which is not a valid word) [17:41:55] You can check that with a lexeme object IIRC (re @Galder: because I can make "dark" > "darker" but I shouldn't be able to make "table" > "tabler", which is not a valid word) [17:42:13] Z6005? [17:42:22] Check this out [17:42:42] This cannot be wrong (re @cvictorovich: Check against word types) [17:45:00] a bit lost with that, the students will make python functions, and the idea would be not to mantain a list of exceptions across ~50 functions (re @wikilinksbot: Z12735 – all meet criteria) [17:45:27] You'd need to make compositions here :( (re @Galder: a bit lost with that, the students will make python functions, and the idea would be not to mantain a list of exceptions across ...) [17:45:34] Right inside implementations (re @cvictorovich: Check against word types) [17:46:47] very complex for what I can explain to them... thanks for the tips! [17:47:44] If you want your students to select lexemes as inputs there is a way to check from that [17:48:47] Galder training new users? [17:49:23] @cvictorovich (re @Galder: is there any good example of functions using other functions inside? I want to show some to te students in a couple of weeks) [17:49:37] Yes? (re @Feeglgeef: @cvictorovich) [17:49:46] There are some [17:50:00] Many examples involving lexemes are [17:50:05] Look at what I am replying to (re @cvictorovich: Yes?) [17:50:32] This is my answer (re @cvictorovich: Many examples involving lexemes are) [17:50:35] You can use something like Z19298 (re @Galder: because I can make "dark" > "darker" but I shouldn't be able to make "table" > "tabler", which is not a valid word) [17:50:49]