| [Coke] | ugexe: any concern about order of processing those PRs? | 00:01 | |
| ugexe | i don't think it should matter but it probably makes sense to do them in the order they were submitted. there are probably a couple of commit messages that reference fixes from other PRs | 00:04 | |
| [Coke] | +1 | 00:06 | |
| just opened a bunch of depreated warnings on modules... and then realized many of them point back to JSON::Fast | 00:45 | ||
| and then there's several coming from various Cro::* | 00:47 | ||
|
01:16
kylese left
01:18
kylese joined
|
|||
| SmokeMachine | apogee_ntv: my first try on writing Selkie::UI: usercontent.irccloud-cdn.com/file/....17.46.gif | 01:20 | |
| apogee_ntv | SmokeMachine: Nice, it's very readable. | 01:28 | |
| SmokeMachine | apogee_ntv: thanks. Yes, I think that’s starting well | 01:38 | |
|
01:58
annamalai left
02:15
kylese left,
kylese joined
02:56
johnjay joined
03:29
lichtkind joined
03:32
lichtkind_ left
03:57
Aedil joined
05:00
topnep left
05:01
topnep joined
05:36
annamalai joined
07:01
Aedil left
07:08
kurahaupo joined
07:37
soverysour joined
07:42
soverysour left
08:13
masayoshi left
08:38
annamalai left
08:42
annamalai joined
09:47
[Coke]_ joined,
[Coke] left
10:11
Sgeo left
11:17
topnep left
11:22
topnep joined
11:34
soverysour joined
11:42
soverysour left
13:27
annamalai left
13:51
soverysour joined,
soverysour left,
soverysour joined
13:56
soverysour left
13:57
Aedil joined
14:34
tinita_ is now known as tinita
|
|||
| SmokeMachine | apogee_ntv: a new change on that. It's kinda handling state values using the Selkie::Store internally and automatically creating subscriptions to path to update the what should be changed (label on this case): usercontent.irccloud-cdn.com/file/....01.12.gif | 15:06 | |
|
15:11
[Coke]_ is now known as [Coke]
|
|||
| Voldenet | but there's no need to type in "val", is there? It can be automatically assigned (a'la react style) | 15:26 | |
| with val as optional description | 15:27 | ||
| and maybe related component instance | |||
| `my UInt $val := use-state 0` | 15:32 | ||
| SmokeMachine | I use the “val” there to use as the path on the store | 15:36 | |
| I think I should use the name of the var as the path… | 15:37 | ||
| Voldenet | you could autogenerate it | ||
| SmokeMachine | the problem on autogenerating it is if you need to update/use it from another place | 15:38 | |
| Voldenet | IMO state is not supposed to be passed around like that | ||
| it's like having global variables | |||
| Yes, global variables are useful so it should be possible to assign them | 15:39 | ||
| but you kinda don't want them to be the default | |||
| something like `sub new-id { <$> ~ ++$ };` maybe | |||
| it'd be the best to pass the proxy directly to the component | 15:44 | ||
| kinda similar style to what signaljs | 15:45 | ||
| or angular do | |||
| erm, solidjs* | |||
| it's very cheap to propagade if the proxy would also have subscribe | |||
| propagate, even | |||
| since then store could be even fully bypassed - store can contain all proxies still for replayability ofc | 15:46 | ||
| or ui transitions | |||
| something like `$fetching = True; $data = ""; start-transition { $address = await fetch-text "/api/address"; $data = await fetch-text $address; $fetching = False }`, that runs in the bg and only commits values to the store upon completion` | 15:49 | ||
| but that'd obviously complicate proxy, because then it can't simply assign values, but needs to be aware of pending transition somehow, with $*transition maybe | 15:50 | ||
| SmokeMachine | I changed `get-state-var($name, $default?)` to `get-state($default?, :$name = UUID.new.Str)` but about the name, I'm still not sure... you will not call it every time you need state, it returns a Proxy that automatically makes the "widget"'s methods subscribe any change on that path, so, when I do `Button.label({ $var ?? "BLE $var" !! "BLA $var" })` it automatically makes the `.label(...)` be called every tine `$var` changes... | 15:51 | |
| (not only when `$var` changes, but mostly because when $var changes, it updates the store and when the store changes, it calls the `label`...) | 15:53 | ||
| Voldenet | I don't mean the dev's perspective, I mean that Selkie::Store itself uses %!db for storage | 15:54 | |
| so it's kinda maintaining large hash that's bound to be slow for large trees | 15:56 | ||
| However, it could store proxies! | |||
| SmokeMachine | I'm currently not using `register-handler`, only the path store, I'm not really sure how both are related... | 15:58 | |
| apogee_ntv | register-handler picks up when component emits event, writes to the store. | 15:59 | |
| SmokeMachine | This is how I'm handling state: www.irccloud.com/pastebin/Oc8J9RK4/ | ||
| apogee_ntv | It means components are completely isolated from writing state, single responsibility. | ||
| Voldenet | `$store.get-in: $name` is the lookup | 16:00 | |
| apogee_ntv | Yeah canonically you would have the component emit an event, then 0..n handlers can handle the event to update state. | ||
| github.com/m-doughty/Selkie/blob/m...unter.raku Like this | 16:01 | ||
| Otherwise you can get a race condition where 2 components try to write state at the same time. Handlers emit effects and effects are applied sequentially. | |||
| Voldenet | m: my %db = :42foo; sub get-in($id) { %db{$id} }; say get-in <foo>; sub get-proxy($id) is rw { %db{$id} }; my $foo-var := get-proxy <foo>; $foo-var++; say get-in <foo>; | 16:03 | |
| evalable6 | 42 43 |
||
| Voldenet | this kinda works, value is still in the %db - Proxy can throw in a global state lock too | 16:04 | |
| I'm mentioning it because anything string-related is expensive | |||
| data races it the problem that js kinda solves by default, since it can't do this… :D | 16:07 | ||
| but you can cheat and emulate js event loop | |||
| SmokeMachine | apogee_ntv: so, instead of using assoc-in, should I be editing events and auto subscring it to update the value? | ||
| apogee_ntv | Yeah, emit event from component, have a handler that's subscribed to the event and then emit an effect from the handler. | 16:09 | |
| or multiple effects if you need. | |||
|
16:13
annamalai joined
|
|||
| Voldenet | uh so | 16:15 | |
| m: my %db = :42foo; my $ch = Channel.new; my $event-loop = start react whenever $ch { $_() }; END { $ch.close; await $event-loop; }; sub get-proxy is rw { my $var := %db{$^id}; Proxy.new(FETCH => {…}, STORE => sub ($, $value) { $ch.send({ $var = $value }) }) }; my $p := get-proxy <foo>; $p = 1; sleep .5; say %db | |||
| evalable6 | {foo => 1} | ||
| Voldenet | event-loop would have to also get read queries and resolve them through promises | ||
| so i'm not sure if that's a good idea after all, it sounds heavy too | 16:16 | ||
| SmokeMachine | apogee_ntv: is there a way to to make Selkie stop swallowing the error messages? | 16:20 | |
| apogee_ntv: Is it looking better now? www.irccloud.com/pastebin/sBfShiWv/ | 16:29 | ||
| apogee_ntv | SmokeMachine: You can set an error log file or handle. It swallows them by default to avoid printing them all over the screen. | 16:45 | |
| Yeah that looks right. You might want to let people specify the event name for easier debugging. | 16:46 | ||
| disbot2 | <jany9781> Hello all, what is the latest status on running Raku in the browser? I saw that there's a JS backend, and I tried to build Rakudo locally with --backends=moar,js. But that eventually fails with ``` /home/.../nqp/nqp-m --module-path=/home/.../nqp/gen/js/stage1 --target=mbc --output=/home/.../nqp/gen/js/stage1/HLL/Backend.moarvm src/vm/js/HLL/Backend.nqp No registered operation handler for 'setmethcache' at | 16:52 | |
| NQP::src/vm/moar/QAST/QASTOperationsMAST.nqp:96 (/home/.../nqp/QAST.moarvm:compile_op) from NQP::src/vm/moar/QAST/QASTCompilerMAST.nqp:1503 (/home/.../nqp/QAST.moarvm:compile_node) from NQP::src/vm/moar/QAST/QASTCompilerMAST.nqp:660 (/home/.../nqp/QAST.moarvm:as_mast) etc. | |||
|
16:56
belluzj joined,
belluzj1 joined
16:57
belluzj left,
belluzj1 is now known as belluzj
|
|||
| [Coke] | JS backend is bitrotted at this point, apologies. | 17:33 | |
| If there isn't a ticket with that particular error, please feel free to open one - probably on the raku/nqp repo. | 17:34 | ||
| [Coke] tries a JS build | 17:35 | ||
| I can open the ticket, presumably I'll have this error shortly | 17:38 | ||
| oh, wow, I'm several versions behind on node, one moment... | 17:39 | ||
| El_Che | hi | 17:46 | |
| tellable6 | 2026-04-25T19:40:31Z #raku-dev <[Coke]> el_che binary releases for 2026.04 can start | ||
| El_Che | ah nice | ||
| will start them on my side as well | 17:47 | ||
| [Coke] | I'm not even getting past the "install npm deps" stage. | 17:56 | |
| El_Che: this release was surprisingly smooth, so I'm sure I screwed something up. :) | |||
| El_Che | bad timing. New fedora in a few days | 17:58 | |
| goodtiming for ubuntu, though | |||
| [Coke] | if I run 'npm install' in the nqp dir, I get a bunch of npm warn deprecated errors, about source-map-url, inflight, urix, resolve-url... | 18:00 | |
| source-map-url doesn't appear in the repo. | |||
| just trying to make sure this is an issue with raku/nqp and not a very old set of installed modules from the the last time I used node | 18:05 | ||
| El_Che | I get some flapping tests | 18:08 | |
| will be ok if I rebuild, probably | |||
| [Coke] | jany9781 - I can't get past 'npm install' here on the mac. | ||
| disbot2 | <jany9781> It needs node version 12 and python 3.10; I used mise locally to get the right ones: mise use node@12 and mise use python@3.10 | 18:09 | |
|
18:09
Sussy joined,
Sussy left
|
|||
| El_Che | t/09-moar/Line_Break__LineBreak.t, t/12-rakuast/xx-fixed-in-rakuast.rakutest | 18:09 | |
|
18:10
Sussy_Touriat joined
|
|||
| [Coke] | oh, whoops, I just upgraded to latest node | 18:10 | |
| I'm at 2x that version number. :) | |||
| Sussy_Touriat | Hey how is everyone | ||
| The Sussy tourist is here | |||
|
18:12
Sgeo joined
|
|||
| disbot2 | <jany9781> Thanks for investigating in any case 🙂 I also had a thought that maybe MoarVM could be compiled to WASM with emscripten (instead of using the JS backend), as an alternative way of getting to the same result of Raku in the browser? But I don't know enough about the build process to have an idea whether that would work at all. I just started trying locally to use emscripten to build MoarVM, but I'm just playing for now, I've never used that tool, | 18:13 | |
| and never contributed to MoarVM or Rakudo either. | |||
| timo | the biggest difficulty you will face is that libuv doesn't compile to wasm | 18:16 | |
|
18:21
Sussy_Touriat left
|
|||
| Voldenet | there's hardly any point of making libuv work on the browser, so it's not a surprise that it doesn't work | 18:36 | |
| in making, even | |||
|
19:08
coleman left
19:09
coleman joined
19:24
abraxxa joined
19:56
belluzj left
20:08
Aedil left
|
|||
| El_Che | [Coke]: I could build everything after 3 runs. Now waiting on ClousSmith to add Fedora 44 to the repos | 20:25 | |
| disbot2 | <librasteve> Jany: hi! just caught your question - with something like HTMX you can run raku (eg in a sandbox) on the server side and stream back to browser - perhaps that is not useful for you, but I thought I would mention as an option... | 20:30 | |
|
20:36
havochunter left
|
|||
| timo | Santa Claus Smith? | 21:03 | |
| El_Che | hehe | 21:08 | |
| timo: I saw your ticket. Is there a speed price to be paid when moarvm is build with --debug=3? If so, maybe a new debug package would be in order for when debug is needed | 21:09 | ||
| timo | normally, distros put debug data in separate files, but i've never done it myself, it's always been something the distro tooling does | 21:11 | |
| otherwise, the debug data is in a section of the file that isn't mapped when you normally load a library | |||
| the debug data can go in a separate package too | 21:12 | ||
| and of course it could also go on a debuginfo server | 21:13 | ||
| El_Che | I would need to check that. Thing is that with the new arm packages, it may be a steep price. There is no JIT there | ||
| timo | why would the price be related to the jit? | ||
| El_Che | I mean that raku on arm is not fast | ||
| of debgug build make it run slightly slower, it would not be nice | 21:14 | ||
| *if | |||
| timo | the debug info parts of the file aren't loaded unless you have a debugger do it for you or some other kind of tool | ||
| i don't think there's a performance difference with -g3 or without | 21:15 | ||
| El_Che | I see | ||
| so the drawback is size then? | 21:16 | ||
| timo | yes, though if you find the right invocation it'd go into a separate .rpm | ||
|
21:18
abraxxa left
|
|||
| timo | more importantly maybe, you can grab the debug info stuff to help figure out a core dump, or to symbolize stack traces where all you have are addresses | 21:23 | |
| lizmat | El_Che: re I mean that raku on arm is not fast: Rakudo is faster on my MBP (on Apple Silicon) than anything Intel based I've ever used | 21:30 | |
| timo | the apple silicon macbooks are also more powerful than a lot of devices you might buy with an arm on it | 21:43 | |
| like, compare a raspberry pi 5 with a macbook pro. the price difference is huge, and the difference in computing power is also really large | |||
| lizmat | true... but therefore the general remark "I mean that raku on arm is not fast" is at least incomplete :-) | 21:44 | |
| timo | or even a high-end flagship android smartphone is not going to get anywhere close to apple silicon laptops | ||
| i have honestly not attempted to use an arm based server or desktop device so far | 21:48 | ||
| so no clue how they compare | |||
|
21:50
topnep left,
topnep joined
|
|||
| El_Che | I have arm VM's | 22:26 | |
| they are not as fast as a macbook | 22:27 | ||
| arm, think of containers on alpine on cheap VMs | |||
| (I have no idea whatsoever what's the speedup is of x64 JIT) | 22:28 | ||
| SmokeMachine | apogee_ntv: my second implementation (with some changes): usercontent.irccloud-cdn.com/file/....40.25.gif | 23:41 | |
| github.com/FCO/Selkie-UI/blob/main/text.raku | 23:44 | ||
| it's not much, but what I currently have is here: github.com/FCO/Selkie-UI/tree/main | |||
