🦋 Welcome to the MAIN() IRC channel of the Raku Programming Language (raku.org). Log available at irclogs.raku.org/raku/live.html . If you're a beginner, you can also check out the #raku-beginner channel!
Set by lizmat on 6 September 2022.
00:00 reportable6 left, reportable6 joined
Voldenet Using rakumod files out of zef modules is not very fun 00:05
Nemokosch I mean, there is just no directory = package correspondence in Raku, right? 00:08
no matter if we declare something to be "the root folder" or not 00:09
Voldenet Well, since you can define whatever structure you want…
Nemokosch not only can you but from what I know, you have to
Is Foo/Bar.rakumod any closer into being a Foo::Bar module than really-funny-filename.xyz? 00:11
guifa_ inasmuch as `use lib …` would interpret Foo/Bar.rakumod in that lib folder as Foo::Bar 00:14
Nemokosch well, would it? 00:15
because the -I flag didn't, as illustrated by grondilu
Voldenet Foo/Bar would need `use ../Foo` or some similar weirdness 00:18
in fact, unless you define root dir explicitly, no matter what convention you use, you end up with nonsense and ambiguities all over the place
Like should `use Foo` in `Foo/Bar` use `./Foo.rakumod` or `../Foo.rakumod` when invoked from the parent and both files exist? 00:19
Nemokosch what does the "files" key do in META6? 00:20
guifa_ Nemokosh: I just checked it 00:22
`use lib …` treats directories as namespaces 00:23
use lib `foo`; use A; use A::B; use bar::A::X; would load out foo/A.rakumod; foo/A/B.rakumod; foo/bar/A/X.rakumod 00:24
Voldenet What if A did `use B` without explicitly saying `use lib` though?
or B did use X 00:25
it could simply use `current .rakumod directory` as reference, or folder defined in `use lib` above 00:26
00:26 nort left
guifa_ I believe it uses the folder defined in use lib, because that's what you would use even in modules 00:28
Nemokosch use lib sets up a FileSystem compunit repository, no? 00:29
Voldenet you can't use `use lib` in modules though
`'use lib' cannot be pre-compiled and thus cannot be used in a module`
guifa_ My point is why would the naming convention suddenly differ? 00:30
Unless the module itself uses `use lib …`, I would expect Raku to search for any referenced module name from the root directory given in use lib
so if lib/A/B/C/D.rakumod says `use E`, I would expect it to load lib/E.rakumod 00:31
does it in fact load lib/A/B/C/D/E.rakumod ?
Voldenet It does load… both 00:34
if there is lib/A/B/C/D/E.rakumod, it gets loaded first
But then when you have E.rakumod in deep structure, how do you load lib/E.rakumod 00:35
Nemokosch > 'use lib' cannot be pre-compiled and thus cannot be used in a module This sounds stupid. Precompilation is meant to be a feature, not a curse. 00:37
no precompilation exists
guifa_ Well, there would be the bug to me. Only allow the full name — I don't think I've ever seen the use E to refer to the next deeper one 00:38
although perhaps one would need `use GLOBAL::E` 00:39
based on my quick read of S10, but currently reading S11 which deals more directly with modules 00:40
Voldenet imo in case where both E.rakumod and **/E.rakumod can be loaded, compiler should start complaining 00:44
it's okay for it to work in such a way, but ambiguities here shouldn't be allowed
guifa_ (I'm actually now not convinced that S10/S11 say that — I think they're looking for current installed symbols. I think it just doesn't handle it, as the `use lib`statement isn't covered anywhere 00:45
Nemokosch github.com/rakudo/rakudo/blob/e5e5...m.pm6#L318 00:47
this seems interesting for me at least. The construction of the provides section for distributions without any META6
1. replace slashes to :: 2. reduce multiple consecutive colons to a length of 2 3. remove the first part of the hierarchy 4. remove anything starting from a dot (? not terribly descriptive) 00:50
Voldenet So apparently what gets loaded is tied to how DIR-RECURSE works 00:53
I'm not sure how do I feel about that…
before creating `Foo/Foo.rakumod` > {Bar => {rakumod => Foo/Bar.rakumod}, Foo => {rakumod => Foo.rakumod}} 00:57
after creating `Foo/Foo.rakumod` > {Bar => {rakumod => Foo/Bar.rakumod}, Foo => {rakumod => Foo/Foo.rakumod}}
Nemokosch from the directory that contains Foo/ ? 00:58
Voldenet Yes 01:00
I'm under the impression that this all depends in what order you get files
and I'm sort of expecting this to be dictated by filesystem…
so, you can write fusefs to change behaviour of modules loading 01:01
Can you feel the horror yet? :D
Nemokosch Perhaps it would be better if this didn't work at all
Voldenet Some sanity check `if module exists then die` in the line you've pasted probably 01:02
there's none of problems if the hash above has no ambiguities 01:03
01:04 xinming left 01:05 xinming joined
Nemokosch you know, the funny thing is 01:09
I took grondilu's sample
raku -I. -e 'say $*REPO.distribution.dist.meta;'
this yields something sane
{api => *, auth => , files => {}, name => /tmp/grondilu-test, provides => {Foo => grondilu-test/Foo.rakumod, Foo::Bar => grondilu-test/Foo/Bar.rakumod}, resources => (), ver => *}
makes sense, right?
when I edit the code to run to actually use Foo::Bar 01:10
raku -I. -e 'say $*REPO.distribution.dist.meta; use Foo::Bar;'
the whole thing explodes and I get loads of stuff from /tmp
which coincidentally ruins the correspondence between the supposed file and the namespace 01:12
it explodes the same way if I use use lib instead of the I flag 01:14
apparently the "use lib" inside Foo::Bar did something funny that retroactively overwrote the repository 01:17
Voldenet I'm betting that the source of this is that it's not explicitly defined anywhere in tests 01:18
so the behaviour doesn't even have to be the same everywhere
tonyo -I. and -Ilib do different things 01:19
Nemokosch that's good to know 01:20
but it would be even better if they didn't, probably... 01:21
Voldenet otoh, now I'm pretty convinced that `use lib` is evil
Nemokosch by the way, I again think that I found why grondilu's snippet was indeed wrong
Voldenet because it usees Foo::Bar instead of Bar?
Nemokosch and it demonstrates my former point from a couple of days ago, that implicit namespace declarations are evil
> method talk { use Foo; Foo::talk; } 01:22
this is the problem
that Foo is a different Foo
Voldenet in fact, `unit class Foo::Bar` doesn't seem like something that should be in Bar file 01:25
Nemokosch by the way, there is no amount of no precompilation that would keep you from the infamous .precomp folder to show up 01:26
Voldenet it should be `unit class Bar; …` and then users of this class should see it as Foo::Bar _if_ loaded in the context of subdirectory
Nemokosch I don't know, Rakudo is full of that structure
github.com/rakudo/rakudo/blob/main.../Parts.pm6
I think the mere notion of unit class Foo::Bar is damned, though 01:27
it can automagically create a namespace Foo for you
Voldenet but it barely serves any point – C# had that problem for a long time, where random namespaces were in random locations
It was fixed by tools that help you change namespaces when you move the file… 01:28
Nemokosch I mean, this is worse than the Java-ish "package foo.bar.baz" thing
Voldenet So, what was the point of even having namespace if it's supposed to mirror the folder structure
Nemokosch a namespace was born, from thin air 01:29
Voldenet I'm not saying that having one file with multiple namespaces is bad
or that Foo.rakumod can't declare Foo::Bar and 5 other classes
but it should be implicit that when namespace is missing, it gets materialized out of thin air 01:30
Nemokosch and now you are damned to have an implicit name conflict with any other symbol (namespace in particular) called Foo
Voldenet and it should mirror the folder structure
Nemokosch It's not even obvious whether Foo::Bar would be attached to an existing namespace or a new one 01:31
Voldenet m: role A { has $.x }; role B { has $.x; }; class C does A does B { }
camelia ===SORRY!=== Error while compiling <tmp>
Attribute '$!x' conflicts in role '<anon>' composition: declared in both 'B' and 'A'
at <tmp>:1
Voldenet This is what should happen.
Nemokosch a new one - that shadows the old one by usual scoping rules
Voldenet when given `Foo::*`, it should be treated as a namespace, everything under the Foo/ directory must be here 01:33
in cases of conflicts (e.g. Foo.rakumod having `Foo::Bar` and Foo/Bar.rakumod having `Bar`) errors should stop you 01:34
Nemokosch back to the snippet... it should be suspicious that the same error shows up without any use Foo
so yes... use Foo must appear before unit class Foo::Bar 01:35
or else you just use the dummy Foo
I mean... it's very good for my ego that something actually causes problems only days after I bring up how damned this whole concept of implicit namespace declarations is 01:36
but would it not be better if somebody took this caveat seriously at least 10 years ago? 01:37
Voldenet probably 01:38
but I'm betting it was implemented and it worked
Nemokosch yes, it "worked" :painharold:
I think it was Nick who brought up Chesterton's fence a while ago. How ironic. 01:39
Almost every time Raku tries to do something unlike any of the mainstream languages - maybe except for Perl that it was meant to fix, not reproduce - it turns out the other languages had a reason to not do it 01:40
Voldenet dynamic namespaces are difficult, especially if you don't compile whole code 01:44
for instance, nodejs just doesn't do namespaces, you simply load the js file and assign its exports to something
Nemokosch There is no shame in that, I think 01:48
Voldenet but it's ridiculously hard to precompile and cache things
Nemokosch I'm pretty sure the result shouldn't be this system, anyway 01:49
Voldenet by design you're bound to never beat anything compiled
Nemokosch well, you know that much of Node is just plain old C++ 01:50
how do imports work in C#, though? I don't remember.
Voldenet you reference dll files 01:51
Nemokosch how are dll files tied to namespaces?
Voldenet compiler turns Foo.Bar.Baz into `this is the "foo.dll" or "foo.bar.dll"` 01:52
Nemokosch hahaha 01:53
Voldenet so, it's all resolved during linking
Nemokosch so is one namespace turned into one dll?
Voldenet depends on how you declare projects
you can have one project with multiple namespaces 01:54
if there's conflicts (two dlls define the same namespace) you end up with errors
that compiler can't decide
in practice people try to follow the convention of package name following the namespace exported by it 01:55
Nemokosch hm, wait, so are these .NET dll's for kind of static linking as well?
Voldenet assemblies just define referenced assembly names which are resolved by runtime on first use 01:56
Nemokosch usually when I think of dll's, I don't think of the compiler doing anything about the linking
Voldenet technically you could do static linking with them though
Nemokosch okay, I don't say it right - it just happened today... 01:57
so apparently when you build a binary, you can define, at least in this godforsaken CC compiler, where the dynamic libraries should be searched first 01:58
it's so much fun to build Pro*C...
Voldenet existing conventions in dotnet world used to cause pain, since there was 'global assembly cache' that had all the assemblies 01:59
and saying `use assembly entityframework version 6`, since every version of entity framework was described as `6.0.0` could give you version 6 or 6.2 :)
Nemokosch 🤣 02:00
Voldenet in which case your app could suddenly throw missing method exception when compiling il code
(since every direct call needs to be turned into actually loaded address) 02:01
Nemokosch so what's the relevance of packages in C#?
Voldenet Well, dotnet have decided to stop using global assembly cache… ;)
Nemokosch Are packages like distributions in Raku terminology? 02:02
Voldenet Sort of
Now you simply load things from `bin/` which define assembly names, so `bin/EntityFramework.dll` becomes `entity framework version 6`
and when assembly `my business logic version 1.0` loads `entity framework v6`, the runtime tries to fetch loaded one in current assembly load context 02:03
if there's package A and B that both use the same assembly name+version with different metadata being loaded, it's possible to load them into separate assembly load context 02:04
which causes other problem, `common interface dll` suddenly exports `common interface dll from runtime A.IInterface` and `common interface dll runtime B.IInterface` which are not compatible with each other 02:05
even though they are defined by the same file
02:05 linkable6 left, evalable6 left
Voldenet there's ways to fix that by passing common assembly load context, but it's not very elegant at the moment… 02:07
02:07 linkable6 joined
Voldenet however, if you use the same assembly versions everywhere within your application, it doesn't feel bad 02:07
since you don't even have to know assembly load contexts in the first place
Nemokosch yeah it sounds more like a "hard things possible" thing 02:08
02:08 evalable6 joined
Voldenet it's even weirder when you throw in native .dll/.so libs to the mix 02:09
"hard things possible", "easy things easy" really is current state of dll loading in .net though 02:10
it used to be "easy things easy… otherwise simply set registry key HKEY_LOCAL_MACHINE\Software\Microsoft\Fusion\LogPath and pray that nobody uses global assembly cache" 02:12
02:13 jpn joined 02:17 coleman left 02:22 coleman joined 02:24 razetime joined 03:20 lizmat_ joined 03:23 lizmat left 03:33 gcd left 03:36 razetime left 03:51 theesm left, derpydoo left 03:52 derpydoo joined, razetime joined 03:53 theesm joined 03:54 razetime left 04:34 donpdonp- left 05:15 jpn left 06:00 reportable6 left 06:03 reportable6 joined 06:14 tejr left 06:21 grondilu joined
grondilu Hi all 06:21
tellable6 2023-03-24T16:10:43Z #raku <tonyo> grondilu: what's the meta look like in that ?
2023-03-24T23:49:16Z #raku <Voldenet> grondilu: the problem with .rakumod-only modules is that there is no `project root`, so your module could be either Tmp::Foo::Bar or Foo::Bar or Bar depending on the root directory used, which causes ambiguities
grondilu 😕 06:23
ngl it's the first time I read about such thing as a "project root".
06:24 nort joined
Voldenet when you start the file from some directory, missing `META.json` get constructed 06:24
ofc, things would work if you did this ix.io/4rLE 06:25
since Bar in directory Foo becomes Foo::Bar
details github.com/rakudo/rakudo/blob/e5e5...m.pm6#L318
see > raku -I. -e 'say $*REPO.distribution;' 06:27
grondilu Well, my Foo/Bar example was an attempt to replicate an error I got from a project where there was indeed a META6.json 06:29
Voldenet in that case `unit class Foo::Bar;` creates Foo that is empty, so when you `use Foo` later it's, well, empty 06:32
grondilu that is just weird. It seems to me I've been using that kind of modules hierarchy for years and I think it's the first time it creates issues. 06:36
any way, it's ok to name 'Bar' a module located in 'Foo/', instead of naming it 'Foo::Bar', then? The names don't have to reflect the directory structures? 06:41
I think I'd rather write 'use Foo;' just before 'unit class Foo::Bar;', as suggested above 06:46
I mean often 'Bar' is actually a qualifier for 'Foo', so it doesn't make much sense on its own. Like say 'JSON::Tiny' 06:47
06:50 tea3po joined 06:53 teatwo left 07:00 razetime joined 07:05 razetime left 07:10 tejr joined 07:40 Ergo444 joined
Ergo444 hi 07:41
07:46 lizmat_ left, lizmat joined 08:08 lizmat left 08:11 lizmat joined, lizmat left 08:16 abraxxa joined 08:17 Sgeo left 08:19 Ergo444 left, razetime joined 08:22 abraxxa left 08:26 sena_kun joined 08:45 razetime left 08:58 lizmat joined 09:06 sena_kun left 09:08 sena_kun joined 09:19 lizmat left 09:28 derpydoo left 09:31 jpn joined 09:42 QhpAptyj9hj0RQwM joined
grondilu how do I get out of a react block? `last` does not seem to cut it. 09:45
09:48 frost joined, QhpAptyj9hj0RQwM left, jpn left 09:54 frost left 09:56 jpn joined 10:01 jpn left
Nemokosch grondilu: it's called done 10:30
10:36 frost joined 10:37 frost left 10:38 QhpAptyj9hj0RQwM joined 10:39 frost joined 10:44 frost left 10:48 ab5tract joined 10:49 QhpAptyj9hj0RQwM left
tbrowder__ grondilu: i have gotten myself in trouble with building modules on the fly during experimentation. i recommend starting a new module project with App::Mii6 (for extra credit, drive it with Mi: 11:04
Mi6::Helped). that gives you a solid root directory, META6.json, and you're off to a solid project usable throughout our community's structure 11:06
*Mi6::Helper
*App::Mi6
11:09 euandreh joined
Nemokosch not sure if that would have helped with this case 11:09
anyway, grondilu has a rare talent to hit problematic aspects of code composition in Raku 11:10
Not gonna lie, I'm thankful for this one. It has been 2 or 3 days ago when I pointed out that implicitly creating namespaces for something like Foo::Bar::Baz is a can of worms - and here you go. 11:11
It's not even healthy to think about it too much... Somebody managed to invent namespaces with the same problems namespaces were meant to solve. It just makes one mad. 11:15
grondilu tbrowder__: tbh I hate module management. I will look up this App::Mii6 thing, but frankly that sounds like yet another layer of complexity. 11:45
12:00 reportable6 left, derpydoo joined 12:02 reportable6 joined
tbrowder__ well, if you want others to use your work, you have to meet expected standards, yes? and it makes your like easier, too! 12:04
try it, you'll like it :-D 12:05
*life
anyone use app "termius" on iPad? 12:06
12:41 jpn joined 13:01 jpn left 13:29 jpn joined 13:37 jpn left 13:50 jgaz joined 13:52 QhpAptyj9hj0RQwM joined 14:08 jpn joined 14:26 Guest37 joined 14:33 rf joined 14:45 QhpAptyj9hj0RQwM left 14:49 Guest37 left 14:50 QhpAptyj9hj0RQwM joined 15:02 rf left
grondilu ok I'm going to throw a pie in the sky about module management. 15:16
I wish it was simpler. Like waaaaay simpler.
I wish it basically consisted in including the source code, and just that. 15:17
like if you need a Foo module, then you include a "content-based" address of the Foo module. 15:18
like an IPFS CID or something.
and that would literally be all.
I honestly fail to see why we need anything else. 15:20
Nemokosch Precompilation? Being able to distribute your individual work in an isolated manner? 15:22
No, frankly I don't think it's about the complexity in itself. It's about unnecessary voluntaristic approach to many things in this topic. 15:23
grondilu Precompilation? You could put a KID to your compiled module if you really want to. Or your import statement could be in a BEGIN phaser or something.
s/KID/CID/
regarding compilation, after all binary code is just an other programming language. So in a way it's a slang, isn't it? 15:26
Nemokosch This is an interesting take 15:27
grondilu this is something I thought about recently when thinking about chat-GPT's ability to translate code from one language to another. It occurred to me that if it can do that, it could also do compilation since binary code is just an other language. 15:28
Nemokosch Idk if it's effective or even sane to represent binary code encoded as text 15:31
15:31 QhpAptyj9hj0RQwM left
grondilu Nemokosh: I don't think a programming language has to be encoded as text to be considered a programming language. 15:32
Besides, wasm has a textual representation so it's not like it's an absurd concept. 15:33
also good old assembler is pretty close to binary code.
Anton Antonov 🙈🙊🙉 15:35
Can be used to do neural programming. 15:36
Nemokosch I didn't talk about what is considered a programming language, although you are right: binary code probably isn't a programming language by definition 15:37
grondilu I said the contrary, that it is a pl 15:38
Nemokosch You were right in your misinterpretation 😛
There is indeed a reason to object to binary code being a pl 15:39
The other thing is: native binary code is as unportable as it gets 15:40
grondilu well I guess there has to be a difference otherwise "free and open source code" doesn't mean anything.
well... 15:41
there is emulation now, though.
and it's usage is growing.
so one could argue that a vm now can be used to run any machine code. 15:42
in a way what it means for some code to be "portable" may become subjective.
though I admit I may be reaching here. 15:43
Nemokosch emulation is costly 15:44
grondilu still, it's all the rage these days 15:45
Nemokosch I don't have that impression at all tbh
I don't see any convincing level of ARM/x86 interoperability, for example 15:46
Anton Antonov @Nemokosch and @grondilu ChatGPT is "all the rage these days."
grondilu I mean, isn't MoarVM a VM? 15:47
Nemokosch MoarVM is a runtime of its own bytecode, not an emulator
probably you can carry MoarVM bytecode around, as long as you are using the same version on both ends 15:48
15:50 rf joined
grondilu btw what do you guys think about chat-GPT? Without going too deep into the long term implications, on the short term I'm looking forward to it being able to translate from one pl to an other. And, as I was discussing, possibly even compile to binary codes such as wasm. 15:51
It would be quite mind-boggling too if it could de-compile efficiently. Imagine the impact on proprietary software lol 15:52
Also wouldn't it be awesome if it could just create a raku compiler out of the specs? 15:53
Anton Antonov @grondilu "[...] being able to translate from one pl to an other." -- I doubt that will work for non-trivial code, if prior -- and recent -- Microsoft at code generation and completion functionalities are taken into account. 15:54
Granted, I am biased : more or less that is what I am trying to do with my DSLs, so I might claim the competition is underperforming. 15:55
grondilu It can already generate code without having trained specifically to do that. There is no way it won't be able to learn to do it well once it is trained specifically for that with self reinforcment learning. 15:56
IMHO 15:57
Anton Antonov @grondilu I think some of your statements indicate too much belief in the generative machine learning algorithms. 🙂
Right, this is what other AI believers think. But, Google hit sparsity problems just with search n-grams, so I think that sparsity problem is going to show up much stronger in programming code generation / completion. 15:59
grondilu well, I play chess and I was impressed by Deepmind's AlphaZero. If I understand correctly, Deepmind used reinforcement learning with self-play, which is very different from how language models are trained. I think both methods will merge eventually ande the result will blow our minds.
ugexe chatgpt is great at telling you stuff that is wrong but sounds right 16:01
i can tell chatgpt it is wrong about something, it will agree, and then in a later response it will reiterate the wrong thing i correct it on
chatgpt is great for inventing stories, but even asking it to describe what an e.g. sql statement does it gets things terribly wrong too often
and unless you already knew the answer, you aren't going to know its lying or wrong 16:02
Anton Antonov @ugexe Yeah, I used the following command openai-playground does google guava have shortest path algorithm implementation 16:03
16:04 jpn left
And got the very assertive and wrong answer : > Yes, Google Guava includes an implementation of Dijkstra's shortest path algorithm in its graph library. The algorithm can be accessed through the "ShortestPath' class in the 'com.google.common.graph' package. The 'ShortestPath' class provides methods for computing the shortest path between two nodes in a graph, as well as fin ding the distance between those 16:04
nodes.
16:04 jpn joined
grondilu it is not trained to produce a correct output, only a plausible one. But once it is trained to produce correct code, which is possible if the code is verifiable somehow, like with a test suite, then it should perform as well as any human IMHO 16:05
ugexe its not hard to produce plausible but incorrect code 16:06
Anton Antonov @grondilu " I think both methods will merge eventually ande the result will blow our minds." -- And am very pessimistic about that. 🙂
grondilu ugexe: I disagree. It is actally pretty hard, that's why good language models took so much time to show up.
16:06 jgaz left
ugexe i dunno i peer review code like every day and it almost always looks plausible and almost always has something wrong with it 16:07
Anton Antonov Also, Chess and Go are easy for Machine Learning -- rules are clearly defined. There is no ambiguity or incomplete information problems. 16:08
16:09 jpn left
ugexe i actually kind of dread the potential future of having to peer-review jr developers code they are c&ping from an AI helper 16:09
Anton Antonov Since you, grodilu and ugexe, are here discussing this -- can you help me naming a Raku module that combines the ML modules in Raku?
grondilu well, rules for programming languages are pretty well defined too, aren't they? You provide an input, and you expect a precise output. That's why raku can be defined from a test suite, after all.
ugexe naturally i'll get AI to also do the code review for me
Anton Antonov I considered "Acme::AI::Fusion" and "Acme::AI::MushUp" . 16:10
grondilu Antonov: what is Acme?
Anton Antonov " well, rules for programming languages are pretty well defined too, aren't they?" -- maybe, but not the problems solved with programming languages. 16:11
ugexe Anton: in perl such meta packages i think are called Task::*
Anton Antonov "Acme" is the name of the company that provides intgeresting devices in the Road Runner cartoons.
Take a look here: raku.land/?q=acme , for better idea of the Raku culture about the "Acme::*" modules. 16:12
grondilu Antonov: I think you misunderstood me. In my mind, at least in the short term, the problems would still be solved by humans. I'm considering Language translation and/or compilation. That does not require problem solving skills apart from precisely language translation/compilation. 16:14
Nemokosch ugexe told me that a sort had a sample from the song Woodhenge by Jethro Tull 16:15
the problem is, I know that song, I'm a huge Mike Oldfield fan, after all... 16:16
ugexe the only sort i promote is sleep sort
Anton Antonov @ugexe Let me be a more concrete here. We have Raku packages for: utilizing OpenAI, utilizing DeepL, utilizing W|A, generating insults, translating natural language DSLs into programming code, and vice versa. We want to facilitate typical workflows with all that "wealth of AI things." For example, round-trip translations, or different code generations from different languages, or defining chains of fallback 16:17
mechanisms between the AI-contractions.
I am not sure how "meta" those use case scenarios are. I think that any kind of initial take on those is going to be opinionnated. Hence, I am proposing (the more timidly prefixed) "Acme::AI::*". 16:20
ugexe yeah i'm not sure... sounds like a framework, and those generally have branding type names it seems like
Anton Antonov @ugexe Right. Well, I can always design a conversational agent and named it accodingly. (E.g. "DSL::English::AIFusionWorkflows".) 16:23
@grondilu I see. Yes, that kind of incremental text build-up -- via ML algorithms -- was used few decades ago to create poems (in English) with great success. 16:24
ugexe chatgpt suggests the following: PolyglotAI, AI Compose, CodeCraftAI, AI Fallback 16:25
Anton Antonov LOL! Cool! 16:26
Which text did you use as a prompt? 16:27
Also, did you use Raku to access ChatGPT? 16:28
16:29 jgaz joined
Here are my results: 16:33
cdn.discordapp.com/attachments/633...MushUp.png
ugexe I pasted one of your explainations in and told it to name a raku module accordingly 16:36
I just used their UI
lucs Here's something interesting, you may have heard about it: arstechnica.com/information-techno...y-over-ai/ 16:37
Anton Antonov @lucs Thanks! 16:38
@ugexe I used my explanation, but I had to add "How to name such framework?" in order to get name suggestions. (As shown in the screenshot above.) 16:39
Also -- I am shamelessly self-promoting here -- I used the CLI of "WWW::OpenAI". 16:40
16:41 jpn joined
lucs And here's Chomsky's take on such systems: portside.org/2023-03-08/noam-choms...se-chatgpt 16:44
grondilu lucs: that man beats machine article is interesting, though I know about a period during which chess grand masters could still beat or draw against machines with a particular style of play. That time is long gone now. 16:50
lucs That's because chess is easy ;-) 16:54
grondilu well, I love this game but I can easily believe that Go is much more difficult. 16:57
17:01 ab5tract left 17:03 Ergo444 joined
Ergo444 hi 17:03
I use fez
I get this 17:04
D:\m\p6\template-nest-xs>fez upload
=<< ERROR: Failed to write 46 bytes to filehandle: Bad file descriptor
what could be the problem? 17:05
ugexe probably that you're using windows 17:06
does fez work on windows yet?
Ergo444 it worked yesterday
though I updated rakudo today 17:07
How can I upload the module then? 17:12
17:40 Ergo444 left 17:43 nort left 17:45 grondilu left
Nemokosch Is there no verbose mode for it? 17:49
17:54 RakuIRCLogger joined 17:58 RakuIRCLogger left 17:59 RakuIRCLogger joined, jpn left 18:00 reportable6 left 18:03 reportable6 joined, lizmat joined
lizmat PSA: IRC logging and the irclogs.raku.org should be operational again now 18:07
18:16 jpn joined 18:50 teatwo joined 18:54 tea3po left 18:57 rf left 19:08 rf joined 19:39 Sgeo joined 19:40 teatwo left, teatime joined 20:05 euandreh left 20:06 jpn left 20:13 jgaz left 20:19 derpydoo left 20:34 rf_ joined 20:36 QhpAptyj9hj0RQwM joined 20:37 rf left 20:41 QhpAptyj9hj0RQwM left 20:47 QhpAptyj9hj0RQwM joined, QhpAptyj9hj0RQwM left
p6steve yes - works on my iphone, having a few teething issues on ipad ... probably due to very old model 20:55
21:29 sena_kun left 21:52 sena_kun joined 22:19 sena_kun left 22:26 rf_ left 22:28 rf_ joined 22:34 Nemokosch joined, Nemokosch left, Nemokosch joined
Nemokosch m: $_ = 12; say 14 ~~ $_ * 1; 22:34
camelia True
tellable6 2023-03-14T12:19:05Z #raku <Xliff> Nemokosch: I know not. That's why I'm asking. :)
2023-03-15T00:16:36Z #raku <tonyo> nemokosch: python has for else
2023-03-16T09:23:02Z #raku <SmokeMachine> Nemokosch: Yes, I think so! :) and I think Raku already has almost everything we would need in a shell.
2023-03-17T15:03:22Z #raku <Woodi> Nemokosch: only one ? crazy :)
2023-03-19T12:47:18Z #raku <XliffMac> Nemokosch: How so? I use the shorthand +@
2023-03-19T20:11:40Z #raku-dev <SmokeMachine> Nemokosch: but .&{} is not equivalent to .do({}). .do returns the original value, so, 42.do: { $!next = $_ + 1 }; would return 42 and store 43 on $!next…
2023-03-20T16:42:18Z #raku <Voldenet> Nemokosch: well yes, golang is awfully simple and non-abstract, it's nice for doing things
2023-03-22T09:49:56Z #raku <exp> Nemokosch: can you show an example? I'm not aware of anything like that, but boy I haven't heard Coffeescript in a while!
Nemokosch oh gosh 22:35
xD
so yeah... I think that snippet actually DOES make sense. It follows the principles of smartmatching. Both regarding implementation (14 is being bound to the $_ of the RHS), and in its higher-level reading (14 can definitely be taken as a value defined by the x * 1 formula) 22:36
m: $_ = 12; say 14 ~~ $_; 22:37
camelia False
Nemokosch This, on the other hand, is not okay.
Somebody, probably in the spirit of "do what I mean", created a corner case with no theoretical foundation, something you absolutely cannot anticipate. 22:39
As I hope this is simply a benevolent mistake living outside of the language design, I'm going to open a Rakudo issue for this. 22:41
22:41 Nemokosch left
gfldex m: $_ = 12; say 14 ~~ * * *; 22:58
camelia Too few positionals passed; expected 2 arguments but got 1
in block <unit> at <tmp> line 1
gfldex m: say &(* * *).^can('ACCEPTS')[0].&{.file, .line}; 23:04
camelia (SETTING::src/core.c/Mu.pm6 19)
Nemokosch actually, even 14 ~~ ($_) works 23:14
now I'm especially not sure if this isn't just a mere bug... 23:20
23:41 linkable6 left, evalable6 left 23:42 linkable6 joined, evalable6 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!