00:02 guifa joined
ugexe the core keys tests don't expect $?FILE to be a lexical in the stash. it wasn't in the legacy grammar. apparently it was made so to make it easier to read by the moarvm debugger. so im not really sure what to do about that 00:03
s/made so/made so in rakuast after-the-fact/ 00:04
in other words github.com/rakudo/rakudo/commit/313dfb7df0fd makes it so when t/02-rakudo/04-settingkeys-6c.t (among other tests) are run with RAKUDO_RAKUAST=1 they fail 00:08
presumably we can just change the test or something, but technically that is a regression
that being said, github.com/rakudo/rakudo/pull/6158 fixes all the other failures in those same tests besides the !INIT_VALUE one (which *is* expected on the legacy frontend but we aren't emitting on rakuast) 00:11
but that too might suggest changing the test if we decide we don't want it to be expected 00:12
00:55 sibl joined 01:08 hulk joined, kylese left 01:32 kjp left 01:35 kjp joined 02:15 hulk left, kylese joined 02:51 sibl left 02:56 sibl joined 03:23 vasko4535586 left 03:29 sibl left 03:33 sibl joined 04:20 Aedil joined
disbot2 <melezhik.> Deepseek take on generating Sparrow task check DSL - chat.deepseek.com/share/lqq7dkdeuipfbbdwvs - very good and accurate results ! 05:21
<melezhik.> Comparison with Python . chat.deepseek.com/share/9hd1n02djgxx7zt6m8 05:25
<melezhik.> IMHO sparrow is a winner 05:26
<melezhik.> Parsing logrotate configuration- chat.deepseek.com/share/pvy11s328tkgts80hz , also very accurate 05:32
<melezhik.> Comparison with Python - chat.deepseek.com/share/629x5g1r3u6hwb4z9m 05:41
<melezhik.> Ok. I have done Qwen, chat gpt, Gemini and deepseek, the last one looks like most accurate and without hallucinations., 05:42
<melezhik.> All comparison analysis says something like that - “in favor of Sparrow if” : Choose Task Check DSL when: Working within existing Sparrow6 infrastructure Need quick, declarative configuration checks Team includes non-programmers who can read DSL syntax Want built-in test reporting without custom formatting code 05:44
<melezhik.> And always it takes far more line of code when doing the same Python as well as time of adding new rule/check is way more if write in Python 05:45
<melezhik.> @librasteve feel free to adjust those fundings and refine them upon your discretion 05:48
05:55 sibl left 05:57 abraxxa joined 05:58 sibl joined 06:03 abraxxa left, abraxxa joined 06:25 sibl left 06:26 sibl joined 06:37 Sgeo left
patrickb ugexe: re $?FILE, I'd say we can just change the tests. I don't expect much ecosystem fallout from this. 06:38
Voldenet actually python is almost never the winner when it comes to either code size or performance 07:10
for size it'd be probably saner to compare anything to raku or perl 07:11
and for performance no idea 07:15
08:20 sibl left
librasteve_ I am curious where (language wise) we can maybe hope to garner interest in Raku - one of my working assumptions is that a fraction of coders who started in Python may be interested to “graduate” to Raku since increasingly Raku seems like a power tool compared to Python 08:28
08:28 sibl joined
librasteve_ does this (untested) assumption hold any water? if not maybe Raku would appeal more to Go^JS^PHP you name it? 08:29
lizmat timo might have an idea about that 08:41
Voldenet unfortunately, some programming languages are a meme and there's hardly any technical reason for them 08:45
though php, js and golang are very domain-oriented, they're not great for everything, but they're good at what they do 08:46
js is unbeatable because it's the only language that every browser can natively run 08:51
08:53 sibl left, sibl joined 08:57 abraxxa1 joined, abraxxa left
Voldenet since old sentiment is that "perl6 (and raku) is slow" – the easiest way to appeal to wider audience would be to achieve top100 in techempower benchmarks 08:58
disbot2 <librasteve> yeah - I think all PLs are fashion, naturally driven by compsci academics and feedback loop - and Google (Go), AWS (Rust) pay $$$ to get into that loop or MS (C#, F#), Apple (Swift) can play since have a bunch of captive devs 09:00
Voldenet I don't know any other shortcut of gaining wide audience other than benchmarks 09:01
kostya/benchmarks also works
disbot2 <librasteve> I recently saw someone say something like "my goto PLs are (i) Zig for low level, (ii) Python for scripting (iii) Go for concurrent and (iv) Haskell for DSLs" - I guess many of us think that way 09:02
Voldenet python is awful when it comes to actual language, but it has a lot of pip modules for everything 09:03
disbot2 <librasteve> it would be awesome to see Raku start to be competitive in benchmarks (ie vs. other scripting languages) -- all power to that 09:05
Voldenet benchmarks are hard numbers that nobody can deny, after all 09:06
even if techempower benchmarks are cheating
lizmat I know timo is working on that... could use a hand :-) 09:08
disbot2 <librasteve> benchmarks has always be a game - so we would probably benefit from playing the game - I bow to timo for that 09:09
<librasteve> that said, tbh I think that competitive benchmarks are necessary (for wide adoption) but not sufficient (since we will "never" beat Rust/Zig/Go) - even if we are the same, folk will just shrug and say "why change" / "look at all the skilled folks you know XXX that can work on my code" / "look at the mountain of PIP modules I can use" 09:11
<librasteve> so since Python is awful, and (many) coders use it because it is "easy" and "quick" rather than for speed and since there are literally millions of daily users (our case needs to resonate with just a small fraction to make a big difference to us) - it seems to me that showing "Raku is better than Python for X" may be able to garner interest for some value(s) of X 09:15
Voldenet most definitely, maybe I'm fixated on numbers too much 09:18
rust and zig can't be beaten for low-level because they're designed for execution performance, but 09:19
disbot2 <librasteve> yeah - I'm not too bothered by that because (i) Rust sucks and (ii) I think that most coders want both a fast / safe / efficient low level PL and a handy / easy / fun scripting PL 09:21
Voldenet golang isn't all that fast 09:22
disbot2 <librasteve> [unless you like writing "unsafe" all the time]
<librasteve> oh - I had thought that LLVM was pretty fast 09:23
Voldenet golang doesn't use llvm
its aim was compilation speed actually 09:24
disbot2 <librasteve> oh - thanks! 09:25
Voldenet though due to its simplicity, it ended up being relatively fast :> 09:36
disbot2 <simon_sibl> is there a chance that someday Raku would become an alternative to Erlang ? by that I mean the ease of making a concurrent system that is resilient ? 09:46
<melezhik.> chat.deepseek.com/share/n7dke5uj4bha4bj8dh You are sparrow task check dsl expert - use following link as a documentation github.com/melezhik/Sparrow6/blob/...kchecks.md to generate task check dsl code to parse output from df command and validate that /root mount is full on more then 60 percent 10:13
<melezhik.> chat.deepseek.com/share/romw1z54mwtlpo5biq You are sparrow task check dsl expert - use following link as a documentation github.com/melezhik/Sparrow6/blob/...kchecks.md to generate task check dsl code to parse systemd unit file and verify its compliance 10:14
<melezhik.> Give me any parsing task and I will solve it with Deep seek + Sparrow in seconds ))) 10:22
10:28 sibl left
disbot2 <melezhik.> This one is not 💯 percent correct but still very good take “You are sparrow task check dsl expert - use following link as a documentation github.com/melezhik/Sparrow6/blob/...kchecks.md to generate task check dsl code to parse kubernetes api configuration file and verify its compliance, you should flatten the configuration file first before passing output to task check parser” - 10:28
chat.deepseek.com/share/6ye7uyr48mykr9r3ey
Voldenet there's probably no chance that raku would become alternative to erlang at all, shared state and try/catch aren't compatible with erlang's mental model 10:34
disbot2 <librasteve> @simon_sibl TLDR: no - longer version ... Raku is built in a fairly standard parallel way (similar in practice to Go, although goroutines are quite a lot lighter afaiui) by which I mean to use threads and shared memory processors to benefit from parallel execution. This is pretty much an abstraction upon (and implemented in) the typical OS (Linux / WIndows) process, thread and message passing model. In contrast a "proper" concurrent language like 10:38
Erlang (I know Occam better) is built to be deeply multi-processor (no shared memory) along lines of Tony Hoare's CSP model to avoid the von Neumann bottleneck. Raku has some similar sounding abstractions (eg Channels) which act logically the same - but these implicitly assume shared memory. You can imagine specializing the Raku abstractions (ie a way to map Proc to Processors and a way to map Channels to serial Links - but that would be a big change and most
code would likely break (deadlocks and so on).
Voldenet Yes you could do erlang-ish things using Channels and design the system very much like erlang would, but language will try to prevent you from doing that. 10:40
by throwing exceptions, of course :>
10:41 tinita_ is now known as tinita
Voldenet I've seen people doing actors in java and it was a sorry sight 10:43
10:46 Guest67 joined 11:10 apogee_ntv left 11:11 ntv joined
ntv Voldenet: Akka runs on JVM? It works fine. 11:20
tellable6 2025-07-04T10:22:21Z #raku <wayland> ntv: ^^
11:20 ntv is now known as apogee_ntv
apogee_ntv If you want to do Erlang-ish stuff the best way is probably write the process model in C++ where you have control over the actual threading primitives and then nativecall to it imo. 11:23
11:24 Guest67 left
apogee_ntv But realistically if you want a billion green processes that's a lot of MoarVM work. 11:26
You need an isolating GC model which is just rough. You'd probably have to exclude MOP and NativeCall from isolation, tbh mutable state *at all* would be a struggle. 11:36
12:05 Guest67 joined 12:15 eseyman left 12:18 manu_ joined
[Coke] ugexe: I'm trying to keep github.com/rakudo/rakudo/wiki/ChangeLog-Draft up to date as you go 12:24
it's basically each commit (and maybe a merge commit) with the commit's above-the-fold commit message. 12:25
I did put the commit/merge/revert commit/merge all into the ignore at the bottom. 12:26
12:36 apogee_ntv left 12:37 apogee_ntv joined 14:23 Guest67 left
ugexe thanks, i know theres a lot to dig through :P Hard to really say how to usefully summarize a lot of those changes are since they are like "fix $thing that noone could have encountered without other fixes in this same release already in place" 16:27
[Coke] heh 16:28
ugexe notably all the fixes that require `make install` to have been run with RAKUDO_RAKUAST=1 16:29
16:33 sjn left 16:39 sjn joined 17:00 erooke_ joined, ab5tract_ joined, Voldenet_ joined 17:01 hulk joined, snonux joined, gordonfish- joined, bloatable6__ joined, bloatable6 left, gordonfish left, apogee_ntv left, kylese left, tellable6 left, erooke left, rantanplan left 17:02 tellable6 joined, Voldenet left, Voldenet_ is now known as Voldenet, greaser|q joined, acidsys left, swaggboi left, ab5tract left, GreaseMonkey left, ab5tract_ is now known as ab5tract 17:03 apogee_ntv joined, acidsys joined, swaggboi joined 17:04 gordonfish- is now known as gordonfish
librasteve_ notable6: weekly 17:31
notable6 librasteve_, 4 notes: gist.github.com/dbef5cf9c3ae60e33b...4ed1255a44
librasteve_ notable6: weekly reset 17:34
notable6 librasteve_, Moved existing notes to “weekly_2026-05-04T17:34:57Z”
17:35 abraxxa1 left
disbot2 <librasteve> rakudoweekly.blog/2026/05/04/2026-...-wars-day/ 18:03
lizmat librasteve_++
18:20 abraxxa joined 18:24 abraxxa left 18:26 abraxxa joined 18:35 abraxxa left 18:50 kurahaupo left 19:03 kurahaupo joined
El_Che . 19:16
I have a corner instead of being sent to the corner, nice :) 19:17
disbot2 <antononcube> Here is my "May 4th be with you" Chatnik-system promotion image: 19:24
<antononcube> cdn.discordapp.com/attachments/633...43421&
20:37 Pixi left 20:42 Pixi joined 21:33 justache left
ugexe there really aren't that many spectest failures with rakuast_rakuast 21:34
21:34 justache joined
ugexe like 5 or 6 21:34
22:00 manu_ is now known as eseyman 22:47 librasteve_ left
Voldenet apogee_ntv: akka itself works on jvm, but it was designed for scala, so you get a lot of weirdness in the java syntax, for example SupervisorStrategy requires DeciderBuilder – it kinda works, but it's just super ugly 23:24
sorry sight, as I said
23:39 Sgeo 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!