wayland76 How do I use an anonymous method? 00:27
tellable6 2025-12-31T18:19:29Z #raku <arkiuat> wayland thanks for that tip on git worktrees 5 months ago; I couldn't absorb it then, but I tried them again recently and find them very helpful!
wayland76 .tell arkiuat No worries. Was that the one in this channel (6th June, at 17:40 AEST from wayland76)? Let me follow up my own comment. I'm using cursor and spec-kitty together. I tend to use cursor (Agent mode) for the small stuff, spec-kitty for the medium-sized stuff, and either cursor or ChatGPT for the big stuff. So my process for something larger tends to go A) Spitball with ChatGPT or something, and capture the bits I want into a Markdown spec; 00:37
tellable6 wayland76, I'll pass your message to arkiuat
wayland76 B) Use Cursor to whip the spec into shape a bit, and break it down into features C) Use spec-kitty to turn those features into tasks, then code (it'll include tests). I review each feature before it gets merged. D) Finally, testing and debugging in Cursor, to follow up the problematic items.
00:41 guifa left
wayland76 .tell lizmat the logging bot seems to be down. 00:46
tellable6 wayland76, I'll pass your message to lizmat
[Coke] m: my &x = anon sub { $^a * 2 }; say &x(3) 00:57
camelia 6
[Coke] there's one way
wayland76 [Coke]: That's a sub, not a method. I'm specifically interested in declaring a method without a name. It can be on a class. It's syntactically valid, but I can't find a way to get at it. It doesn't seem to show up in .^methods, for example. 01:00
[Coke] ... whoops. I misread as sub. 01:46
01:48 guifa joined
guifa ack, got disconnected, wayland76 couldn't see from transcripts if you got an answer 02:09
an anonymous method might as well be an anonymous sub that gives you access to self, basically
my &foo = anon method :: ($self: $a) { say "called :: on $self with arg $a" }; 'bar'.&foo(42); foo('xyx', 123) 02:12
evalable6 called :: on bar with arg 42
called :: on xyx with arg 123
02:26 hulk joined, kylese left 02:47 lichtkind_ left 03:15 hulk left, kylese joined 03:37 guifa left
arkiuat wayland76, uh no, it was July 19th in #raku-doc, but tellable6 isn't on that channel. 03:48
tellable6 hey arkiuat, you have a message: gist.github.com/2b430e73a8799135d1...a46721d574
arkiuat .tell wayland yeah sorry I don't know what any of that is. It was 19th July 23:45, on #raku-doc about git worktrees. I posted here because tellable6 wasn't hanging out on #raku-doc 03:54
tellable6 arkiuat, I'll pass your message to wayland76
arkiuat .tell wayland oh wait, I mean 18th July (converted from UT to AEST in my head, forgot to do the date too) 03:57
tellable6 arkiuat, I'll pass your message to wayland76
04:46 Aedil left 05:26 gugod left 06:19 guifa joined 07:28 arkiuat left 07:40 arkiuat joined 07:45 arkiuat left 07:59 gabiruh left, gabiruh joined 08:10 guifa left 08:15 arkiuat joined 08:19 arkiuat left 08:38 gugod joined 08:45 arkiuat joined 08:50 arkiuat left
disbot8 <melezhik.> Happy new year! šŸŽˆšŸŽŠšŸŽ† Have started a new ( Sparrow based ) project , check it out - doubletap.sparrowhub.io/ 08:53
<melezhik.> Happy new year! šŸŽˆšŸŽŠšŸŽ† Have started a new ( Sparrow based ) project , check it out - doubletap.sparrowhub.io/ 09:04
09:09 arkiuat joined 09:13 arkiuat left 09:41 arkiuat joined 09:46 arkiuat left 10:10 arkiuat joined
patrickb what is that about? 10:13
10:15 arkiuat left 10:21 RakuIRCLogger left 10:44 arkiuat joined 10:48 arkiuat left 11:04 arkiuat joined
disbot8 <melezhik.> ā€œDouble TAP is lightweight testing framework where users write black box tests as rules checking output from tested "boxes". Boxes could be anything from http client, web server to messages in syslog. This universal approach allows to test anything with just dropping text rules describing system behavior in black box manner.ā€ 11:06
11:06 itaipu joined
lizmat Everybody: best wishes for 2026, may it be a good one! 11:14
tellable6 2026-01-01T00:46:17Z #raku <wayland76> lizmat the logging bot seems to be down.
lizmat should be good again now 11:17
11:28 arkiuat joined
librasteve_ weekly: rakuforprediction.wordpress.com/20...-examples/ 11:28
notable6 librasteve_, Noted! (weekly)
librasteve_ weekly: rakuforprediction.wordpress.com/20...ata-scienc
notable6 librasteve_, Noted! (weekly)
disbot8 <librasteve> weekly bot is not working on Discord ATM ;-( 11:29
11:33 arkiuat left
disbot8 <librasteve> happy new year! 11:34
11:38 itaipu left 11:45 arkiuat joined 11:47 lichtkind joined 11:53 arkiuat left 11:57 itaipu joined 12:07 arkiuat joined 12:11 arkiuat left 12:13 arkiuat joined 12:15 justache left 12:17 arkiuat left 12:19 justache joined 12:29 arkiuat joined 12:33 arkiuat left 12:45 arkiuat joined 12:49 arkiuat left 13:13 arkiuat joined 13:17 arkiuat left 13:44 annamalai left, annamalai joined 13:47 arkiuat joined 13:53 arkiuat left
lizmat new, updated! dev.to/lizmat/raku-2025-review-4m0l :-) 14:03
14:14 arkiuat joined 14:18 arkiuat left 14:29 arkiuat joined
disbot8 <antononcube> Happy New Year! 14:53
16:22 Aedil joined 16:33 abraxxa joined 16:38 abraxxa left 16:39 abraxxa joined 16:43 arkiuat left 16:45 annamalai left 16:57 kaskal left, kaskal joined
korvo Has anybody figured out how to use a Raku grammar to constrain an LLM's output? llama.cpp and friends have a builtin BNF metalanguage for CFGs but I want something PEG-shaped. 16:57
17:22 annamalai joined 17:38 arkiuat joined
disbot8 <aruniecrisps> Happy new year everyone! 17:51
<aruniecrisps> @Voldenet @wayland76 I'm wondering if it's not worth it to make a community slang module that allows you to define new trait names like provides 17:52
17:58 wayland76 left 17:59 wayland joined 18:13 abraxxa left 18:15 arkiuat left 18:16 abraxxa joined
disbot8 <antononcube> @korvo Can be done via llm-synthesize and/or llm-function and/or "LLM::Graph". But those are more generic approaches than just specifying the grammar directly as filter for a particular LLM. (Like llamafile models.) 18:26
<antononcube> Using grammars or regexes to filter or validate LLM results is mentioned in "Day 6 ā€“ Robust code generation combining grammars and LLMs", raku-advent.blog/2025/12/06/day-6-...-and-llms/ . 18:30
18:31 arkiuat joined
disbot8 <antononcube> The first LLM graph example, "Parallel race", can be refitted to use a Raku grammar by the "judge" node. 18:32
korvo antononcube: Fun! FWIW specifying the grammar directly has the useful property that we *provably* generate valid texts, and that's what I use it for in my own harness. 18:36
18:37 arkiuat left
korvo It's been nearly 3yrs since I pointed this out to simonw. They didn't understand, but a peer introduced github.com/1rgs/jsonformer and local-LLM generation has had perfect JSON ever since. 18:37
I *could* find some way to negotiate between the typical local harness, written in Python 3, and Raku. It might be more useful to have Raku modules which natively call the LLM's inference routines instead. 18:38
antononcube: To give some idea of what local harnesses do, here is some distilled Python 3 which considers all of the possible tokens for an LLM and masks off any token which wouldn't be acceptable in a sanitized chat window. bpa.st/VZZA 18:45
So if there's some way to neatly invert a grammar, to feed it incrementally and ask it for a set of valid next characters, then it would be possible in principle to use Raku for these sorts of token-restricting harnesses. 18:46
18:47 arkiuat joined
disbot8 <antononcube> The package "WWW::LLaMA" can be made to use grammar spec. Right now it does not, but you can see here that it can pass a fair amount of arguments to LLaMA LLMs: github.com/antononcube/Raku-WWW-LL...ns.rakumod 18:51
18:51 arkiuat left
disbot8 <antononcube> At the time I wrote "WWW::LLaMA" I did not consider passing grammar filters "universal" enough to include it in the signature. 18:53
<antononcube> I assume "WWW::LLaMA" has to be revisited with the newest LLaMA models signatures. 18:54
<antononcube> TBH, currently my highest LLM-related development priority is making an Ollama client. 18:55
<antononcube> As for "generating JSON for sure" valid JSON code is produced fairly robustly by adding llm-prompt('JSON') to a llm-synthesize or llm-function specs. 18:59
<antononcube> @korvo "It might be more useful to have Raku modules which natively call the LLM's inference routines instead." -- I aggree. To large extend that is why I think it is better to have separate, dedicated LLM-service utilization packages ("WWW::OpenAI", "WWW::LLaMA", etc.) that a generic one. 19:02
19:03 arkiuat joined
korvo Yeah. BTW Ollama and llama.cpp can run various non-LLaMA models; there's been a bit of feature creep upstream. 19:04
So that e.g. I'm working with RWKV, using rwkv.cpp, and I don't know what a Raku module might be called. Is "WWW::RWKV" too opaque? Does it matter that it's not a Web interface? 19:05
disbot8 <antononcube> @korvo "WWW::RWKV" seems fine to me. It is a family of models, from what I see. 19:12
<antononcube> Meaning "WWW::RNN-LLM", or something like that, might be more informative, but it is not concrete enough. 19:13
korvo There's no WWW in it, I mean. My current harness opens librwkv.so and does FFI. 19:14
disbot8 <antononcube> Then "RWKV::Client" or "RWKV" might be good. 19:15
<antononcube> I have similar dilemma naming modules that access H2O.ai or OR-Tools. 19:16
<antononcube> Here I use "H20::Client" : github.com/antononcube/Raku-H2O-Client 19:17
korvo Some languages have namespaces for FFI bindings. I gather that that's not something the Raku community does, though. 19:18
19:22 Sgeo joined
disbot8 <antononcube> I think the closest thing is ā€œ::Nativeā€ to the package name. Or just use the name of the library or service. Like, ā€œRWKVā€ or ā€œH2Oā€. 19:23
19:26 xinming joined
korvo It's too bad that Grammars can't have inverted control or incremental parsing. It would be useful for many tasks, not just for LLMs. 19:28
disbot8 <antononcube> @korvo What do you mean by ā€œgrammar inversionā€? Generation of sentences using a grammar? 19:54
korvo Yes, precisely. Or, more generally, having a suspended parse with a set of acceptable characters.
disbot8 <antononcube> Yes, that is very useful and use it often. I would say it is even a measure of maturity of the corresponding grammar systems. 19:56
<antononcube> These three Raku packages provide random sentence generation: - Grammar::TokenProcessing - EBNF::Grammar - FunctionalParsers 19:58
<antononcube> The ā€œsuspended parseā€ is a good idea! (If understand it correctlyā€¦) 19:59
korvo It's just how some parsers naturally work. Instead of taking an entire string, they take individual characters and return a new parser after each character. 20:04
And an efficient encoding leads to each parser having a set of expected characters, in the sense that you could try feeding a parser every possible character and see which parsers generate errors. 20:05
disbot8 <antononcube> I see. And yes, this how functional parsers can be also seen. 20:08
20:08 jmcgnh left
disbot8 <antononcube> I was thinking that "suspended parse" can be (relatively) easily implemented using functional parsers. 20:09
<melezhik.> I have brought some examples of TTAP protocol here -doubletap.sparrowhub.io/examples , including Raku inlines 20:10
korvo FunctionalParsers definitely has the sort of API that I would want. 20:11
disbot8 <antononcube> I think what you are suggesting/wanting can be done by modifying the sequential parser sequence. 20:23
<antononcube> An alternative, is replace each of the initial k-parts of a parser expression with the parser success or satisfy({True}). 20:25
21:07 arkiuat left 21:12 arkiuat joined 21:17 arkiuat left 21:30 arkiuat joined
wayland .tell arkiuat In that case, ignore what I said about AIs, but, you're welcome :). 21:33
tellable6 wayland, I'll pass your message to arkiuat
21:34 arkiuat left
wayland .tell aruniecrisps It might be, in that I would've used it. 21:36
tellable6 wayland, I'll pass your message to aruniecrisps
21:50 arkiuat joined 21:54 Aedil left 21:56 arkiuat left
wayland .tell guifa Maybe an example would be better. gist.github.com/wayland/8a52577055...6f4964c3c5 . It runs, but the anonymous method doesn't show up. It's not so much that I'm dead keen on anonymous methods, it's that it's possible to declare them, so I thought that, if there were a way to use them, I might have a use for it. 21:57
tellable6 wayland, I'll pass your message to guifa
wayland [Coke]: ^^
tellable6 2026-01-01T03:54:24Z #raku <arkiuat> wayland yeah sorry I don't know what any of that is. It was 19th July 23:45, on #raku-doc about git worktrees. I posted here because tellable6 wasn't hanging out on #raku-doc
2026-01-01T03:57:19Z #raku <arkiuat> wayland oh wait, I mean 18th July (converted from UT to AEST in my head, forgot to do the date too)
22:05 abraxxa left 22:17 arkiuat joined 22:26 arkiuat left 22:32 arkiuat joined 22:50 arkiuat left 22:56 itaipu left
lizmat m: class A { }; BEGIN A.^add_method("foo", my method () { "foo" }); dd A.foo # use of anonymous method 23:24
camelia "foo"
wayland lizmat: So the declaration that I had doesn't work then? 23:27
lizmat yeah, looks like anonymous methods are not getting added to the method table 23:28
wayland lizmat: OK, good to know. Does that count as a bug? 23:29
lizmat not sure: I guess it's somewhere between a DIHWIDT and maybe a warning / error at compile time
wayland Hmm. OK. I'm not sure either :p 23:31

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!