| 18:13 | <shu> | did we ever talk about Math.fround on bigints? alon (who works on emscripten) reached out to me that as it is currently specced (float64->float32), trying to round BigInts to float32s requires accumulating rounding errors by first going to a float64 |
| 18:14 | <shu> | i'm assuming this rounding behavior actually broke the compilation of some c/c++ code in emscripten |
| 18:16 | <devsnek> | we need to generally get bigint supported in math methods |
| 18:16 | <ljharb> | +1 |
| 18:16 | <devsnek> | its on my list of proposals |
| 18:16 | <ljharb> | i haven't had time to get that together but will be happy to cochampion with whoever does |
| 18:17 | <devsnek> | https://snek.dev/proposal-list |
| 18:18 | <shu> | i am not sure all math methods make sense |
| 18:18 | <devsnek> | well not literally all of them |
| 18:18 | <devsnek> | wouldn't make much sense with Math.sin |
| 18:19 | <shu> | it's a fairly small subset actually |
| 18:20 | <shu> | the bitwise stuff, min/max, uh... sign, abs |
| 18:21 | <devsnek> | abs, cbrt, clz32, fround, imul, the logs, max, min, pow, sign |
| 18:25 | <shu> | sgtm |
| 18:45 | <Bakkot> | the logs are kinda weird because you often want a number out |
| 18:45 | <Bakkot> | most bigints in my code are such that log2(bigint) would be less than 2**53 |
| 18:45 | <shu> | i assumed the logs would continue to output numbers |
| 18:46 | <shu> | it would be unusual to want integral logs, right? |
| 18:49 | <bradleymeck> | except for change of base? |
| 18:49 | <Bakkot> | dunno about that. integral log2 (rounded up) tells you how many bits are are in a thing. |
| 18:50 | <TabAtkins> | Yeah but you can just... round up, if you want that. I'd want the actual fp log of my bigint by default. |
| 18:51 | <Bakkot> | sure, just saying that integral logs are not unusual |
| 19:01 | <ljharb> | shu: it'd definitely be a case by case audit |
| 19:02 | <ljharb> | but anything where it makes sense, it should work |
| 19:49 | <TabAtkins> | just print the number and hold your ruler up to the screen if you want integral log, imo |
| 21:39 | <littledan> | > Private data that isn’t bad |
| 21:39 | <littledan> | devsnek: I would be interested in your thoughts on this |
| 21:39 | <devsnek> | i feel like i've vocalized my thoughts on private fields already :P |
| 21:40 | <littledan> | Oh sorry I guess I forgot what they were, if there was something additive. Do you have a reference? |
| 21:40 | <littledan> | (if not that's fine) |
| 21:41 | <devsnek> | i mean the tl;dr is i think they are a bad thing that doesn't fit the rest of the language |
| 21:42 | <littledan> | Right so I wasn't sure which part you might want to fix with a proposal |
| 21:43 | <devsnek> | something without new syntax and forced throwing |
| 21:51 | <littledan> | throwing? |
| 21:52 | <littledan> | shu: Curious why this comes up from JS and not just Wasm. I thought Math.fround was sort of an asm.js albatross at this point. |
| 21:55 | <devsnek> | littledan: they throw instead of returning undefined |
| 21:56 | <devsnek> | also the emscripten stdlib on the js side has to do various operations, i'd guess that's where the problem is |
| 21:59 | <littledan> | In which case (re private)? |
| 22:04 | <littledan> | I mean, it already does throw in a bunch of cases; that was a design goal. Do you mean how they're default-initialized to undefined as fields? |
| 22:10 | <devsnek> | littledan: no I mean I don't think they should throw if they don't exist on the receiver |
| 22:10 | <littledan> | oh, you meant *without* throwing, OK |
| 22:11 | <devsnek> | now that they are implementation reality one can only hope we can add all the features needed to make them usable |
| 22:19 | <littledan> | so, which features would those be? |
| 22:19 | <devsnek> | probably the simplest would be reifying them |
| 22:20 | <devsnek> | cuz then you could just use the value with `in` and as a computed lookup and add it to existing objects and delete it from things and etc |
| 22:23 | <devsnek> | you could do each one of those as its own bespoke feature but that gets into my point about composing with the language |
| 22:37 | <shu> | littledan: it came up in emscripten generated js fallback code |
| 22:37 | <shu> | so that aligns with your understanding |