01:55 kylese left 01:56 kylese joined 02:15 kylese left, kylese joined 02:16 human-blip left
Voldenet maybe MAIN could use something like this for `global options` # proto sub MAIN(:$suspend, :$port) { { * } }; multi MAIN("foo") { say "foo" } 02:34
which would understand proto args before actual MAIN 02:36
03:22 wayland left 03:27 donaldh left
simon_sibl Voldenet: thanks for all the feedback ! So from what I understand, I actually don’t need to use start{} at all ? Won’t that mean my client will be slower and run on a single thread ? (Not saying it’s wrong to be single threaded but I thought having multi thread to handle the peers would help) 03:37
Voldenet multithreading helps a lot if you do cpu-bound work 03:38
simon_sibl And also you say I can put a QUIT phaser at the right place to handle error better
Voldenet and if threads use different regions of memory
I think torrent client is mostly limited by io 03:40
simon_sibl I see, I’ll check how it goes if I don’t use the start block, for the QUIT phaser tho I’m still a bit confused, what would it help with ?
Voldenet QUIT would catch errors from peers crashing 03:41
simon_sibl I see, will help a lot for debugging 😆 or if I make a better ui that could show error 03:42
Voldenet you can integrate it with `start react …` later too, then create a channel and return its supply – it's quite flexible 03:48
04:13 wayland joined
wayland .tell patrickb Once I get the basic Data Oriented Programming stuff done, I'm looking at making something that will replace the standard "sub MAIN". I'd call it CommandObject, and the idea would be that you would follow the exe name with a command and an object (eg. git diff path/to/file, where "diff" is the command, and "path/to/file" is the object). Then various optional options like --option. 04:15
tellable6 wayland, I'll pass your message to patrickb
wayland .tell patrickb: I think you just have to go: sub MAIN($compulsory, :$optional, %_) and the rest of the options get dumped into %_
tellable6 wayland, I'll pass your message to patrickb
04:31 ShimmerFairy left 04:38 swaggboi left 04:40 ShimmerFairy joined 05:10 belluzj joined
patrickb wayland: but that would swallow up args appearing after the single compulsory one that match $optional. 05:21
tellable6 2026-06-27T04:15:42Z #raku <wayland> patrickb: I think you just have to go: sub MAIN($compulsory, :$optional, %_) and the rest of the options get dumped into %_
hey patrickb, you have a message: gist.github.com/ff577ce2dc25e63c34...5f809d7725
05:26 hurufu joined 05:27 belluzj left 05:35 donaldh joined
simon_sibl Voldenet: I am not totally confident with the whole concurrency thingy in Raku, but I removed the start{} for the peers and now it just blocks and doesnt download anything, but I am not sure to see anything blocking the execution for each peer, also instead of .then I changed the for loop to "generate" (if thats the right way to see it) a new whenever for the peer which remove it from the Set once the promise has been 06:11
kept/broken
ah ok thats the difference between react and supply 06:21
06:32 human-blip joined
alright, updated, added the QUIT (still a bit confused with the different layer of QUIT) and without the start{} block, thanks a lot Voldenet ! 06:43
now I am just not happy with the current method !request-work its very ugly but yeah xD 06:45
but it still becomes the highest CPU hungry process of my machine xD 06:47
cdn.discordapp.com/attachments/633...6ddb1&
cdn.discordapp.com/attachments/633...47302& 06:48
probably the way I handle the pieces I and the peers have, maybe using a SetHash with the ID of each piece isnt amazing memory wise.. 06:49
bigger issue is really cant seem to finish downloading a file, the last or lasts pieces dont get downloaded or written even when I add a %file<handle>.flush for some reason 06:56
lizmat: it seems the whenever Promise.in with dynamic interval seems not to work actually xD 07:33
lizmat meh
sorry, not in the right frame of mind to be of any help :-( 07:34
the heat / lack of sleep is just getting to me :-( 07:35
9:30 and already 32 degrees here :-(
simon_sibl no worries xD I just lazily use a start block for this for now 07:39
08:02 abraxxa joined 08:11 belluzj joined
there are no AC nearby you lizmat ? 08:31
lizmat yes, there is, but it's not cutting it :-( 08:32
I mean, where I sit it's 26 now... which is bearable 08:33
but it's that and the combination of not getting enough quality sleep time
and I guess age has also something to do with it 08:34
simon_sibl you mean the AC is not powerful enough ? age must not help when having bad sleep for sure..
lizmat well.. I guess the house wasn't really built for these types of events... the local weather station just declared a "super heat wave" 5 consecutive days of 30+, with 3 of them 35+ 08:50
at night it didn't get below 26.9 here
there's just no cooling off :-(
tadzik :( 09:03
09:13 Sgeo_ left 09:31 belluzj left
melezhik. o/ 09:37
11:41 wayland76 joined 11:42 wayland left 12:24 abraxxa left 13:08 librasteve_ joined
simon_sibl Voldenet or anyone really, there is some kind of bug hiding that makes the download freeze and doesn’t progress but the event loop is still running fine, and I really can’t figure out why or where it happens 13:12
timo have you worked with `gdb` before? 13:15
oh, liz made a cool module that's included in rakudo called "own-up" that lets you send SIGINT to the process and it makes all threads print their current backtrace. all you need to do is put `-Mown-up` in the commandline or `use own-up;` in one of your sources 13:17
or you can `VM.own-up` at any point too
of course that only helps when threads are actually executing at that moment, and not all waiting for tasks to come in on the task queues 13:18
13:34 guifa_ joined
guifa_ okay I think I came up with a very cool little pattern for folks making reusable tokens 13:48
you can override the match method
err made
grammar Foo { method made { ... } } 13:49
normally that'd violate the idea of being able to allow people to give their own actions
but you can grab a user specified action result via ast
grammar Foo { method made { with self.ast { #action overridden# } else { #do default action# } } 13:50
patrickb simon_sibl: Which raku lib? Do you use tls? 13:54
guifa_ oh dangit, it can be simplier, I just forgot that callsame isn't a term but a sub 13:56
callsame() // #defaultAction#
ugexe wayland76: i think both of the rakuast bugs you somewhat recently opened are fixed on HEAD 14:12
timo any reason you wouldn't make a corresponding FooActions? that could also be a role that can be re-used
ugexe i don't know exactly which commits would have resolved them (or even if they truely are fixed), i just feel like they are probably resolved 14:13
ah sorry, only one of the issues i was references was yours
guifa_ timo: actually I'm thinking it more, probably need a different approach 14:21
i dunno why I forgot you can just insert a grammar with <grammarName> lol
I'm wanting to make it such that a token can be reused AND also can provide its own .made method 14:22
timo I would get very surprised if a `.made` doesn't behave just like an accessor for what is installed with `make` 14:39
guifa_ it is, but can be overriden if necessary in typical Raku fashion 14:45
I think I've found a better method for what i'm trying to accomplish, just now having to test it out
guifa_ is mad coding getting ready for his talk lol
timo how about putting a { make blabla } at the end of your token and if the user wants to override it with their own stuff they can use an action method on their own action class; the work will be duplicated, but it's probably less surprising maybe? and the action class can use the value that the token itself put on the $/ as well just by grabbing the `.made` 14:48
guifa_ my current approach is this: Grammar, Actions, and then a token that recreates the grammar. final line of the token is a { token does override-made } line, where override-made is a simple method that processes and caches the result 14:53
that way the action call is only done if requested, and otherwise not called if no one wants it
the token is just a copy of TOP
14:54 belluzj joined
guifa_ right now my trick is figuring out a way to make the Grammar/Actions extensible in an intuitive way such that the base grammar/actions can provide some special handling 14:55
14:55 belluzj left
guifa_ (concretely: for a URI grammar/actions class to enable specialized handling for certain schema) 14:56
timo did you think about parameterized roles yet? those can have tokens/rules/regexes in them too, which can be interesting for some use cases 14:57
guifa_ timo: yeah. so the goal is to figure out in the TOP grammar/actions to do something like the following (untested, purely notional) 14:58
timo oh, but don't let me keep you from your Conference Driven Development sprint if i'm not getting you closer to your goal 14:59
guifa_ token TOP { <schema> [ <?{ schema has registered customGrammar }> <custom> | <default> } 15:00
and then ditto for the actions
100% in the spirit of torturing the implementor to figure out how to do this
and then of course the goal is that i can have a reusable token that than incorporates both that grammar and action per the .made override 15:01
simon_sibl patrickb: not a lib, a BitTorrent client I wrote: github.com/4zv4l/rbt really can’t find the bug after few hours of debugging 15:07
timo simon_sibl, did you try the own-up module yet? 15:11
15:53 guifa_ left 15:59 guifa_ joined 16:15 hurufu left
guifa_ Is there any great way to check if something is a type object versus NQPMu ? 16:21
in an implementation agnostic way
(that is, not relying on NQPMu) 16:22
[Coke] guifa_: 2026.06 is being released today, FYI.
(binary releases probably a few days out)
timo guifa_: try .DEFINITE 16:34
guifa_ timo: say Int.DEFINITE
--> False
timo yup 16:35
and unlike .defined a class can't redefine it for itself
that's why it's all uppercase, it's like WHAT and such
guifa_ yeah -- I guess I should have been more clear
16:35 acidsys left
guifa_ so in grammars, I can access the actions via self.actions 16:35
I'm wanting to check if an actions was set (or not). But quite often, actions are stateless / type objects 16:36
timo ah, you want "was an attribute given any value by the constructor" essentially
guifa_ yeah. It would be easier for Nil, but Rakudo uses NQPMu 16:37
best I can come up with is `$foo ~~ Any && $foo !=:= Nil`
timo Nil is slippery, as it resets containers to their default value if assigned into them 16:41
so handing it around is not a great user or developer experience
this comes down to "what is a valid action class?", and since every regex/rule first checks if the method of the same name exists in the action class (or instance), it's totally fine to put "1" as your actions 16:43
guifa_ yup
I would love if down the road we had a new key word 16:44
simply
actions { ... }
and a method defined as action foo { ... } would autoset $/
16:56 acidsys joined
librasteve guifa++ that’s a cool idea … definitely worth a problem solving issue (even better just implement it ;-)) 17:13
17:16 eseyman left
timo i feel like we've had that suggestion before. it would probably work well together with "naming your token/regex/rule something that is present as a method in almost all classes gives surprising results" as well 17:16
17:18 manu_ joined 17:19 manu_ is now known as eseyman
librasteve I, for one, like the idea that Nil represents the absence of a meaningful result. Sure it is unusual, not just another way to spell null 17:19
guifa_ librasteve: would be interesting if we could have an action inside of a grammar. Occasionally I have some grammars where the reusability of an actions class isn't actually relevant, but being able to place the action right next to the token could improve maintainability
librasteve yeah, I was thinking would be nice to infer that grammar Invoice {…} actions Invoice {…} Invoice.parse: $txt; are coupled too 17:22
but then I think Larry is a genius
17:32 librasteve_ left 17:34 human-blip left 17:36 human-blip joined 17:38 abraxxa joined 17:43 abraxxa left
guifa_ I swear we had a trait for it 17:54
is there a way to prevent a subclass from implementing a given method from its parent? 17:55
timo at the very least you can have your own metaclass that does that ... maybe a COMPOSE phaser can do something? also, putting ^blabla methods in a class is a thing you can do 18:00
18:16 ShimmerFairy left 18:18 ShimmerFairy joined 18:35 guifa_ left 18:36 librasteve_ joined, guifa_ joined 18:41 guifa_ left 18:46 ShimmerFairy left 18:51 ShimmerFairy joined 19:07 MasterDuke joined 19:18 _________ left 19:20 Sgeo joined 19:25 _________ joined 19:36 guifa_ joined
timo gist.github.com/timo/20b8e7f029b21...1be3bdce7f working on new commands for GDB when debugging a moarvm / rakudo process 20:18
20:43 rmv joined 20:44 itaipu left 20:50 itaipu joined 21:06 guifa_ left 21:12 guifa_ joined 21:22 wayland joined 21:24 wayland76 left 21:25 MasterDuke left
aruniecrisps here is my budget migration tool in Raku: gist.github.com/arunvickram/4bb7cc...e0b5fd4128 21:34
@wayland76 here it is, i think you expressed interest in it 21:35
21:52 guifa_ left 21:55 Guest63 joined 21:56 Guest63 left 21:57 guifa_ joined 22:11 rmv left 22:13 librasteve_ left 22:17 rmv joined, rmv left, rmv joined
Voldenet simon_sibl: I would consider adding more debug logging - e.g. `whenever Supply.interval(5) { %peers.log }` or something similar 22:57
it could help you see if it's concurrency problem (log should write things every 5 seconds more or less, if it slows down then it could be cpu speed issue after all)
if it's not, it would simply write some dump to inspect the state 22:58
keep in mind that torrent peers can blacklist you making you unable to download, also ipv6 exists – best thing to test torrent client would be to set up your own tracker in local network I think 23:16
it's not real-life test, but it could be good enough with some VMs with traffic shaping 23:17
23:47 guifa_ left 23:51 guifa_ joined

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!