🦋 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.
cfa like, someone else my want to use MyMod 'widget' isntead? 00:00
00:00 reportable6 left
cfa (to alias &foo as &widget) 00:00
lucs Exactly. 00:01
cfa urgh, typos, sorry; someone else may want to use MyMod 'widget' instead?
lucs Yep.
00:02 reportable6 joined
ugexe so just make a sub EXPORT that exports everything regardless of whatever is on the right side of `use Foo :bar` 00:02
lucs You know, ultimately, rf's solution could be fine, but I'm really curious about how to do this inside the module itself.
cfa i think that's almost the example here: docs.raku.org/language/modules#EXPORT
look for $short_name => MyModule::Class
lucs ugexe: I tried a bunch of stuff with EXPORT, but I wasn't able to make it work.
cfa: Yeah, I saw all that. 00:03
cfa you want to just map whatever's passed in to the sub
lucs Looks easy, right? Welp, I wasn't able to come up with the right expression. 00:04
In that $short_name example, I think I'd want something like ｢Map.new: do "&$short_name" => ??? if $short_name｣, but is it indeed ｢&$short_name｣ and what goes at ｢???｣? 00:18
cfa with the caveat that i have no idea what i'm doing: 00:22
gist.github.com/cfa/13c5088670b6a1...b237ebca9f
00:22 derpydoo left
cfa (i used `our` in this example so that i could easily reference A's &foo from within EXPORT) 00:23
i'm sure someone who's used EXPORT before can provide better guidance; i was just riffing on that example 00:24
lucs I didn't think of 'our'.
But this, copied from my failing experiments, fails: 00:25
Map.new: do "&$new-subname" => &MyMod::subname if $new-subname;
(same as yours, eh)
cfa yeah but subname will just be lexically scoped by default
i.e. it's effectively `my sub subname` 00:26
so EXPORT can't see it
lucs Won't the EXPORT take care of that?
Maybe I'm not following...
cfa EXPORT isn't in the same scope
lucs Right.
cfa from the docs: it looks like EXPORT is part of the compunit rather than the package/module, so we can't place it inside the module definition (and hence the module's scope). 00:31
lucs Correct. 00:33
Um, I'm going to eat now, and resimplify my gist again.
(there is also a potential "multi" problem) 00:34
TTYL
cfa cheers
00:36 Manifest0 left
cfa hmm 00:49
regarding this,
m: module A { sub foo is export { ... } }; module B { sub foo is export { ... } }
camelia ===SORRY!=== Error while compiling <tmp>
A symbol '&foo' has already been exported
at <tmp>:1
cfa i might have answered my own question, though the behaviour still seems surprising
i.e. if the exporter relates to the compunit rather than the module itself, then this makes sense 00:50
(as does this differing from the .rakumod case)
anyway, & 00:51
rf I successfully made cinnamon buns 00:58
01:36 evalable6 left, linkable6 left 01:37 linkable6 joined 01:39 evalable6 joined 01:50 derpydoo joined
Anton Antonov @rf Using Raku? 02:09
rf I wish. No, sadly just in real life. 02:10
lucs Is this legit?: $condition ?? (use A) !! (use B)
'use' is compile-time, iiuc, so I'm wondering. 02:11
Anton Antonov @rf If you proclaimed that earlier I might have used cinnamon buns making example in my Gherkin grammar post...
el compile time conditions hmm 02:12
rf Lol. That would have been excellent
lucs Well, the $condition would be evaluated at runtime, eh. 02:13
rf You're much better off using require vs use in that scenario. 02:15
lucs Oh, maybe 'require'.
rf ^
lucs Yeah, just thought of that :)
el @if $?VERSION ~~ v2.d+ foo @else bar @endif 02:19
Anton Antonov @rf The "Highlander" movie example is the same as the one in cucumber.io . Food preparation would be also good.
02:20 cfa left 02:22 frost joined
Nemokosch the BEGIN phaser runs at compile time, though 02:32
so who knows...
02:46 razetime joined
lucs Okay, can't make it simpler. Hope I'm clear in what I'm trying to do (previously explained): gist.github.com/lucs/5335c195d3642...48606c21d8 03:16
03:21 coleman left, coleman joined 03:35 razetime left 03:58 razetime joined 03:59 codesections joined 04:17 rf left 04:31 codesections left, codesections joined 04:37 razetime left 04:47 razetime joined 04:52 codesections left, codesections joined 05:21 codesections left 05:23 codesections joined 06:00 reportable6 left 06:03 reportable6 joined
Nahita @lucs only slight change is needed. problem is, A::orig is not really accessing the &orig residing inside the class A because subroutines are "my"-scoped, i.e., lexical by default. So to attach that symbol to the class, you can "our"-declare it instead. In multis, it wants an our-declared proto, so our proto sub orig(|) is export {*} multi orig (Int $n) { $n * 2 } multi orig (Str $s) { $s ~ $s } inside the 06:10
class A should do it
06:10 bbrtj left 06:56 pingu joined 06:57 razetime left 07:02 jpn joined 07:04 razetime joined 07:07 jpn left 07:19 frost left 07:29 razetime left 07:30 razetime joined 07:41 abraxxa-home joined 07:44 razetime left 07:45 sena_kun joined 07:53 razetime joined
el will raku have implicit interfaces (static duck typing)? 07:55
07:56 p6steve left 08:08 frost joined 08:24 frost left
moritz I don't think so, that doesn't fit with the dynamic vibe of raku 08:28
08:30 jpn joined
el moritz: then why have gradual typing in the first place 🤔 08:35
08:35 jpn left
moritz so that you don't have to specify types when you don't want to care about them, but you *can* specify types if you want, and get benefit from them 08:39
08:40 Sgeo left 08:42 grondilu joined
grondilu Hi all (today I'm trying Emacs's IRC client) 08:42
tellable6 2023-02-18T21:26:37Z #raku <tbrowder> grondilu the raku/perl6 emacs mode does have the capability to show an index of raku classes, subs, and variables
moritz waves to grondilu 08:43
el m:perl role R { sub a() { ... } } class C { sub a() { say ‘hello’ } } say C ~~ R;
Raku eval False
el roles can’t be checked for implicit satisfaction 08:45
grondilu that seems sensible to me
el i want structural typing 08:46
roles is not a viable way to do that 08:47
grondilu subsets then? 08:49
08:49 freeside left
grondilu thinks raku's subsets are pretty neat 08:49
08:50 dakkar joined
moritz subsets are neat, but they don't provide any benefits at compile time 08:55
el how can i do this with subsets then protocol Pet { sub cry(); sub pet(); sub feed(Food $); } 08:57
moritz subset Pet of Any where { .^has_method('cry') and .^has_method('pet') } 08:58
not ideal 09:00
el it doesn’t have any constraints on the signature
09:00 jpn joined
grondilu: how would you do that then 09:01
09:05 jpn left
grondilu Apart from something like what moritz showed above, I don't know. 09:05
el see 09:06
09:08 Manifest0 joined
moritz there's not really structurual typing in raku, it's mostly nominal 09:14
Nemokosch Which is a bit strange, given that you can create a class that respects indexing, without making it Positional 09:16
el raku allows dynamic duck typing but not static duck typing smh 09:24
El_Che static duck typing? 09:25
el structural typing
El_Che millions of static language programmers cried in error! A unbalance in the force!
09:26 jpn joined
Nemokosch by the way, static duck typing is present in traditional C++ templatess 09:41
I remember the uni teacher even said "it's the only reasonable form of duck typing"
09:43 ab5tract joined 09:55 pingu left 09:59 ab5tract left 10:05 freeside joined
el c++ - the language that tries to do everythin 10:11
10:30 sena_kun left 10:38 jpn left 10:53 freeside left 11:06 sena_kun joined 11:14 grondilu left 11:26 razetime left 11:39 jpn joined, TieUpYourCamel joined 11:41 razetime joined 11:57 razetime left 12:00 reportable6 left 12:02 reportable6 joined
Geth ecosystem/main: 774f4da9cf | (Elizabeth Mattijsen)++ | META.list
Remove Intl::Format::Number, it now lives in zef ecosystem
 12:03
12:04 codesections left 12:18 Xliff joined
Geth doc/finanalyst-patch-1: f03b281df7 | (Richard Hainsworth)++ (committed using GitHub Web editor) | doc/Language/101-basics.pod6
remove TM char for consistency

This addresses issue #119 in doc-website, was in Raku/doc Either all references to Raku should have tm or none. My preference is for none as it is obvious		 12:24
doc: finanalyst++ created pull request #4205:
remove TM char for consistency
[Coke] .ask codesections can we get feedback on github.com/Raku/doc-website/issues/119 12:32
tellable6 [Coke], I'll pass your message to codesections
[Coke] what do folks think of something like "use rakudo *>2022.07;" 12:37
(Trying to think of a syntax in the docs to say this was introduced in some compiler version, and then wondering if we could use the same syntax in the actual compiler)
12:39 Xliff left
lizmat that syntax could actually be made into a module 12:40
[Coke] right!
lizmat and possibly added to the core
[Coke] then I can make it *your* problem, and just document it. :)
lizmat but better in ecosystem, as then older versions of Rakudo would also be able to use it
[Coke] as all things should be? ;)
12:41 gcd joined
lizmat I just wonder about what the name of the distribution should be 12:41
"rakudo" feels.. ambiguous ?
Rakudo::VersionMatcher ? 12:42
[Coke] something from <use rakudo compiler version requirement>
That seems reasonable.
then we can document "use this module" somewhere in the docs, and add these lines wherever we have a known version requirement. (We already have the version specification one.) 12:43
12:46 derpydoo left
lizmat actually, make it a real version object, so "use rakudo * > v2022.07" 12:46
to allow for better version semantics
this would also allow: "use rakudo v2022.07+" 12:47
[Coke] Yes, please.
lizmat I'll whip up the module after finishing today's Weekly
[Coke] ... Crap. Ok, I'll get this incorporated into the docs for the next milestone (end of March). 12:48
(sooner if possible)
github.com/Raku/doc/issues/302 was the original ticket where we were discussing what this might look like. 12:50
Nemokosch what would that module do exactly? 12:53
Iirc there is a module that dies if you don't fulfill some version constraint, courtersy of Zoffix
12:54 hythm joined
[Coke] that's it, die if your compiler version didn't match. 12:56
Sure, if it's already out there, maybe we can save lizmat a few minutes here.
Nemokosch github.com/raku-community-modules/RakudoPrereq 12:57
it doesn't seem like a lot of work; if you decide on a different interface, this could probably even be archived, with some pointers to the new thing 12:58
[Coke] looks like the syntax there is a little different (and only lets you set a minimum version, not a where clause), but seems good enough.
added that to the ticket, thanks. 12:59
lizmat fwiw, I think if you want to match the documentation with the use statement, a new module may make sense after all 13:05
hythm .tell Nemokosch: wrt to grammar question I posted a few days ago, looks like I needed to do two things to solve the issue, first needed to use `|` semantics (as you suggested), instead of `proto` which seems uses `||` semantics. second needed to use `regex` instead of `token`, looks like regex backtracks to find best match. updated the gist with
the changes gist.github.com/hythm7/128bc02c190...f9b827f892
tellable6 hythm, I'll pass your message to Nemokosch
lucs Aha! Yes, that appears to work fine. Thanks for the code and for the explanation! 13:06
13:14 jpn left
[Coke] lizmat: I agree. we can always point to the prior art and include the info about why a new one in the readme. 13:15
Nemokosch oh yes, regex backtracks, I did know that part but I didn't know that would have the right consequence here 13:17
moritz I generally recommend to avoid backtracking over regex boundaries in a grammar, things can get very confusing very quickly 13:33
hythm If no backtracking over regex, what would be the alternative 13:36
13:42 jpn_ joined
moritz I have no idea what you're parsing, but generally you can chose your tokens/regexes in a way that once you have parsed something as a token, you commit to that 13:42
(at least when parsing computer code/languages; might be different for natural language) 13:43
Nemokosch well a lot of the issue boils down to the grammar definition, that's for sure 13:47
hythm I', trying to parse a command line options by grammar. lets say I have an option that is both can be specified as boolean or as having value. lets say an option named "deps", when grammar sees "deps", actions method will make `deps => True`. when grammar sees "deps test", action method will make `deps => 'test'. so far so good, but if there is 13:49
exisiting another option named "test", therre will be an issu
e how to parse it, should be paresd as two options "deps" and "test", ot one option "deps => test"
moritz and if the input string is "--deps --verbose", then --verbose will be parsed as a separate option? 13:51
token option-with-optional-argument { '--deps' \s <argument-that-does-not-look-like-option>? } 13:52
token argument-that-does-not-look-like-option { <!before '--'> \S+ } 13:53
hythm Thanks will try these and see how it goes. 13:54
re: "and if the input string is "--deps --verbose", then --verbose will be parsed as a separate option?"   yes.
moritz token option-with-optional-argument { '--deps' [ \s+ <argument-that-does-not-look-like-option> ]? } # likely better, the separating space should also be optional 13:56
hythm regarding the optional space, I'm thinking to make use of it in an edge case when the user really want "deps test" to be parsed as `deps => "test"`, instead of parsed as two options"deps" and "test". so the user can remove the optional space "depstest" 14:02
anyway I'll try the above and see how it goes, thanks 14:03
14:05 jpn_ left 14:10 razetime joined
hythm if I'm not prefixing options with `--` (meaning "verbose" instead of "--verbose", how token argument-that-does-not-look-like-option will look like 14:10
moritz well, you have to have *something* to distinguish options from their arguments 14:14
even if it's a hard-coded list of possible option names
if you don't, neither the grammar nor a human has a good chance to parse it correctly 14:15
el oh no 😢 ===> Searching for: fez ===> Updating fez mirror: 360.zef.pm/ !!!> Failed to update fez mirror: 360.zef.pm/ ===> Updating rea mirror: raw.githubusercontent.com/Raku/REA.../META.json !!!> Failed to update rea mirror: raw.githubusercontent.com/Raku/REA.../META.json No candidates found matching identity: fez 14:46
14:48 jpn joined 14:53 rf joined
rf Good morning folks. 14:55
Anton Antonov @rf Did you eat all buns yet? 15:04
rf No, but i'm ashamed to say I ate 6 out of the 14 I made haha 15:05
lucs Solved: Renaming an exported sub: gist.github.com/lucs/cc60e0b0f6a34...bf704da061
Anton Antonov @rf Well, it is a good idea to eat you own food -- helps aligning the food preparation process. 15:06
@rf So, what is the verdict? You did good?
el can someone add my module to p6c github.com/dangduomg/Super
rf Yeah I like them, by far my best attempt yet!
@el gatito you should upload to zef 15:07
Anton Antonov @rf Ok, not to be too obnoxious here, but how many attempts did you do? 15:08
el i can't install fez
rf Lol. I've been trying to make them for about 2 months, maybe 3 tries total I am not a very good baker :D
Why not @el gatito? 15:09
Anton Antonov @rf Cool. (Persistence is the key.)
Nemokosch it would be better to fix that installation because p6c is not usable in the long run
rf ^^
Nemokosch is there a reason you put the subs inside a class?
el rf: C:\Users\jack9>zef install fez ===> Searching for: fez ===> Updating fez mirror: 360.zef.pm/ !!!> Failed to update fez mirror: 360.zef.pm/ ===> Updating rea mirror: raw.githubusercontent.com/Raku/REA.../META.json !!!> Failed to update rea mirror: raw.githubusercontent.com/Raku/REA.../META.json No candidates found matching identity: fez
lizmat and yet another Rakudo Weekly News hits the Net: rakudoweekly.blog/2023/02/20/2023-...herkining/ 15:10
rf lizmat++
lucs Nemokosch: I'm not quite sure. The idea is to make them available when the user does "use A". 15:11
Nemokosch to be fair...
===> Searching for: fez ===> Updating fez mirror: 360.zef.pm/ ===> Updated fez mirror: 360.zef.pm/ The following distributions will be upgraded: fez:ver<40>:auth<zef:tony-o>:api<0> ===> Updating: fez:ver<40>:auth<zef:tony-o>:api<0> Segmentation fault
rf Lol.
Nemokosch An unexpected old friend... 15:12
rf el gatito: What version of rakudo?
Nemokosch (installed for the second time though)
el C:\Users\jack9>raku -v Welcome to Rakudo™ v2022.12. Implementing the Raku® Programming Language v6.d. Built on MoarVM version 2022.12. C:\Users\jack9>zef --version v0.14.5 15:13
Nemokosch looks like Windows
el my terminal doesn't support unicode lmao
Nemokosch @lucs frankly I keep mixing up what name signifies what but I'm assuming it didn't work without
a class is a package probably 15:14
rf @el gatito maybe clone and build from master
el tenor.com/view/crying-emoji-dies-gif-21956120 15:15
Anton Antonov @lizmat The link to "Intl::Format::Number" is not working; it should be raku.land/zef:guifa/Intl::Format::Number . (And thank you very much for posting!)
rf git clone github.com/tony-o/raku-fez.git && cd raku-fez && zef install . --force-install
el gatito ^
Adjust that to be windows compatible lol. Sorry I am not sure how windows works at all for command line 15:16
I think powershell is posix compliant so maybe use that
Nemokosch zef can still install from a github link iirc 15:17
lizmat Anton Antonov thanks for the headsup, fixed: the link was to the version I just removed from the ecosystem :0)
Nemokosch zef install github.com/tony-o/raku-fez.git
rf Oh yeah forgot about that. zef install --force-install -v github.com/tony-o/raku-fez.git
I find I need force-install if the build has failed sometimes
Nemokosch hmmm 15:18
not sure if that's intended or not
el also why raku has 3 ecosystems 😭
rf Don't ask why.
Nemokosch historical reasons - nowadays, only "the zef ecosystem" should be relevant
rf Yeah, only use zef.
Nemokosch the other two were experimental attempts, so to speak
el Unable to find a suitable handler for web (tried Fez::Util::Curl, Fez::Util::Wget) 15:21
failed 15:22
lucs Nemokosch: Interestingly, the only apparent difference when having the proto and the multis outside the class (no class necessary then) and changing the 'Map.new' line from ｢… => &A::orig …｣ to ｢… => &orig …｣ is that when doing ｢use A 'mine';｣, the 'orig' sub remains visible instead of being undeclared (I have no idea why -- I'm semi-cargo-culting all this). 15:23
Nemokosch that's probably because it is our-scoped 15:25
what if you take your former code and check if you can see A::orig?
(iirc classes are our-scoped themselves) 15:26
el full text: gist.github.com/dangduomg/7c4f8f9b...9f5e8c69ca
in conclusion: can't install fez on windows 😭
lucs Nemokosch: Not sure which former code you mean -- I've done so many variations, eh. What you're thinking about is not mentioned in my latest gist? 15:27
Nemokosch well you are doing a great service by at least figuring that out and illustrating the demand
lizmat [Coke]: considering Rakudo::Version as the name of the module
allowing for "use rakudo v2022.01+" 15:28
Nemokosch the one you said about that "it worked"
rf el gatito: You need curl installed 15:29
el rn only way is to download the module yourself and zef install it
rf curl.se/windows/
[Coke] lizmat: sounds good to me.
lucs Nemokosch: Sorry, I'm really not sure which one you mean -- the code would need to be spelled out. 15:30
Nemokosch the code you made the "difference" to 15:33
and the "difference" was that orig became visible
now I'm hypothesizing that A::origin became invisible, in return - something that we earlier haven't paid attention 15:34
15:36 cfa joined 15:38 jpn_ joined
el rf: ok, where do i place the directory for curl for zef to recognize 15:41
15:41 jpn left
lucs Nemokosch: When the A class wraps the "our proto" and the multis and the 'Map.new' has ｢"&$new-name" => &A::orig｣, then having ｢use A 'mine'｣ makes 'orig() undeclared, but 'A::orig()' is visible. 15:43
Nemokosch then gotcha - there is the one extra sub 15:46
lucs The 'our' at work, eh. But for my use case, this is okay, since it's the plain 'orig()' that I want to hide. If the user does 'A::orig()', well they know exactly what they're getting. 15:48
el rf: 15:51
oh just add curl to PATH 15:53
found that out by reading the code lol
15:54 Sgeo joined
cfa lucs: if you don't want the our, how about gist.github.com/cfa/4efb1fa5eb2f98...11ccc6abaf ? 15:55
lucs cfa: Not that I don't want it per se, but that's an interesting idea, I'll play with it. 15:56
15:58 freeside joined
lucs freeside: Hallo! 15:58
rf el gatito: Did it end up working? 15:59
16:01 jpn joined 16:02 freeside left 16:03 jpn_ left
el i also have to install tar 16:04
it worked now
poggers
rf Nice. 16:05
16:13 freeside joined
el what C:\Users\jack9\Documents\raku\Super>fez upload =<< ERROR: Failed to write 67 bytes to filehandle: Bad file descriptor 16:14
16:18 freeside left
bruh 16:19
ugexe Modern versions of windows 10 and 11 come with both curl and tar
shouldn’t need to update PATH when using them either 16:20
cfa as part of wsl, or just normal binaries?
ugexe normal binaries
cfa (i haven't used windows since, hmm, 7?)
and then, only very briefly
ugexe Windows 7 is no longer supported so I suspect there won’t be much more catering to it 16:21
cfa oh, i just meant that my knowledge of the platform is fairly dated 16:22
the last windows i used "properly" was probably 2000
that's *cough cough* years old at this point 16:23
cfa sighs
ugexe pureinfotech.com/tar-curl-windows-10/
rf el gatito are you on windows 7? 16:25
ugexe Depending on a specific version of rakudo (or even rakudo in general) should be discouraged in distributed modules 16:26
el i use older versions of windows 10 (because my computer is 10 years old and screams at higher versions)
rf Maybe its time to use linux :^)
ugexe otherwise in the future we end up with ecosystems for each raku compiler
el rf: need spare flash drives for that lmao 16:27
ugexe I would suggest updating windows 10 to something more modern. Not necessarily 11, but at least the latest patches / service packs 16:28
el does anyone know this error
rf I can almost gurantee that is something to do with windows version
16:29 pingu joined
ugexe Windows 10 build 17063 and later 16:30
el in case you are curious i use build 10240 lmao the og windows 10
16:31 ab5tract joined
rf Virtual machine? 16:43
If you aren't willing to update windows, or install linux VM is probably your only option
[Coke] ugexe++ 16:47
el windows 10 version 1803 should be good enough imo
it contains build 17063 16:50
cygwin? 16:51
rf WSL
[Coke] I have a build version higher than that here (still on windows 10), been mostly ok.
(using neither cygwin nor wsl)
16:52 abraxxa-home left 16:56 freeside joined
lizmat [Coke]: raku.land/zef:lizmat/Rakudo::Version 16:59
17:04 freeside left
Nemokosch yes, one shouldn't depend on Rakudo versions, yet it's probably an invisible bomb for a long time 17:06
with modules that downright use NQP
[Coke] lizmat: thank you. I may provide a PR that addresses ugexe's concern about balkanizing the ecosystem.
17:13 QhpAptyj9hj0RQwM joined 17:26 freeside joined 17:28 perlbot left 17:29 simcop2387 left 17:30 freeside left 17:31 simcop2387 joined 17:33 perlbot joined
[Coke] (just a note in the README) 17:35
17:36 razetime left 17:39 dakkar left 17:40 jpn_ joined 17:42 jpn left 17:47 jpn_ left 17:48 freeside joined 17:52 freeside left
ugexe We don’t try to make it easier to e.g. use nqp 17:54
Nemokosch How to put it. The guidance to not use nqp just to gain performance, is indeed out there. But then there is also JSON::Fast which must be among the most used Raku dists out there - all NQP. And truth be told, I'm also relying a lot on stuff like HTML::Entity::Fast, simply because the performance benefit couldn't be overlooked... 17:59
18:00 reportable6 left
ugexe truth be told it shouldnt be written in nqp 18:00
or rather it should be written in nqp and in the core 18:01
(which it is, so rather it should be a public feature)
18:03 reportable6 joined
ugexe if we want to bless a single compiler into existence and base our future outlook on a single compiler then i'll change my stance though 18:03
Nemokosch Fair enough. It seems that these days, it's worth considering. Currently, it doesn't seem legitimate to create another Raku compiler - neither technically, nor specification-wise (it's not worked out well to actually make that possible), nor ecosystem-wise (a lot of stuff would just downright break) 18:05
Also, the development of the language itself is driven by what can be achieved with Rakudo 18:06
ugexe technically there actually is a public interface to JSON parsing through CUR, you just have to use it in a way it wasn't intended
cfa specification-wise, isn't it the case that any compiler that passes the test suite is spec-compliant? 18:07
ugexe there was a time that zef ran on parrot, moarvm, and jvm (and maybe even niecza)
so the dream technically was alive at one point
cfa i 18:08
(also, i'm not sure what "doesn't seem legitimate" means)
infeasible, perhaps?
Nemokosch cfa: what I mean is that a lot of what we usually just consider "the language", is actually not specified - I guess because that would put a lot of burden to a possible other implementation, pressure onto it to be like Rakudo 18:09
a massive example is the metamodel
cfa yeah, good point
i remember controversy at documenting some of the *HOW stuff on the doc site
lucs Which Discord channel would be appropriate to suggest an IRC-bridge improvement? 18:10
Nemokosch #meta probably
lucs Okay, I'll go there.
Nemokosch back to metamodel business - ... but to give up on actually using the metamodel to any extent, seems like a huge cost
18:11 sena_kun left
a lot of things that you could do easily, "by writing Raku code", would suddenly become unavailable 18:11
another itch that I keep getting back to is the versioning. Right now, there is no precedent of handling bigger changes - although it seems at least mainstream to think that some breaking changes would be beneficial at the right time and in the right form 18:12
cfa i mean, there's a difference between leveraging implementation-specific details and calling an established api 18:13
18:13 freeside joined
Nemokosch then we can say that there is no "established api" at the moment, and I don't know what cost it would take and who would be willing to work it out 18:13
ugexe using features that were just implemented are usually a bad idea anyway. they need to be baked into a few releases so e.g. apt-get users using a rakudo thats only like a year old don't have a terrible experience 18:14
in a distribution other people will use at least 18:15
for programs that aren't distributed knock yourself out
Nemokosch That's also a point 18:16
At the same time, it's good to feel that there is continuous development in the most literal sense of the word, and perhaps find ways to show it, too. 18:17
18:18 freeside left
ugexe i'm trying to think of a feature that a program like zef would be better off using immediately 18:18
cfa hmm, how do distributions like debian treat rakudo versioning?
and are dated bugfix releases still made?
like would there be a 2022.12.3?
ugexe i imagine i'd probably just implement that feature in zef until that feature is in a few rakudo releases
Nemokosch I guess that's again the cost of "CoreHackers" pretending to be users, like myself. For me, a 2 years old Rakudo compiler is unusable, not even because of missing features but mainly because of all the bugs that I'm aware of and would step on.
ugexe the p6c ecosystem gets generated by a pretty old rakudo 18:19
Nemokosch When I need to get something used by others for company, I'd tell them to get rakubrew by running like 2 commands in the command line, and come straight to the "bleeding edge" 18:20
ugexe for a regular user i dont think old rakudos are that much of a show stopper other than when they try to use modules that don't work on old versions
cfa or when they run into a bug that was that was fixed in a later release
18:20 jpn joined
ugexe i mean bugs still get introduced in new versions as well 18:21
cfa sure -- but that might be a 'show stopper' scenario
ugexe but what i mean is i almost never hit rakudo bugs when i'm writing programs
Nemokosch yes, fair enough - and this is where the bias kicks in. If you know the old bugs, you are probably much more sensitive to that
ugexe only when im writing modules or doing core development
cfa i've run into quite a few
Nemokosch because you might remember them being fixed
or how much you were bothered by them 18:22
Rog Yes, I would definitely like some consistency and some way to handle moderation across the “divide” so to speak
cfa i suppose the sensible response hereis a good test suite, plus rakubrew to test against a reasonable set of older versions
Nemokosch For Roast, I think it's just not what it claims to be. It's a semi-decent test suite for Rakudo, as things stand 18:23
cfa why semi-decent, and what's missing?
Nemokosch Remember the bug with lists of custom-key hashes? 18:24
cfa you mean the hash bug i reported?
Nemokosch yep
cfa yeah
Nemokosch That's just one handy example but bugs like that happen quite often. That was a regression that nobody noticed for seven years 18:25
I added like two tests for it but one can quickly think of cases that still aren't covered
cfa sounds like a great initiative 18:26
Nemokosch well I'm planning to return to the whole "storing (and storing into) different Maps" topic, just didn't want to jump into it right now 18:27
rf The only bug in Rakudo I ever hit is the `our %map is Map;` bug
cfa i meant spending time adding tests if you can think of cases that aren't covered already
perhaps not as glamorous as other core development but surely very important
Nemokosch so yes - I'd say that's the easier part. To make Roast a decent Rakudo test suite 😛 18:28
cfa oh huh, github.com/rakudo/rakudo/issues/2753 is still causing a panic 18:29
18:29 QhpAptyj9hj0RQwM left
cfa Nemokosch: saying it's not 'decent' strikes me as dismissive but sure, being able to catch regressions early is always good 18:30
tellable6 cfa, I'll pass your message to Nemokosch
Nemokosch What word would you fancy to express the idea that it's not as good as it should be?
cfa to make a more comprehensive suite, i guess? 18:31
Nemokosch Okay, so - to make Roast an actually comprehensive Rakudo test suite seems to be the easier part.
Rog lizmat - Discord owner here: as far as how I want to handle moderation concerns in the future, I have two goals 18:32
Nemokosch To make it a generic language specification that provides one a usable interface and followable, consistent features tied to language versions, regardless of the internals of a compiler - that would be much harder.
Rog 1. Shutting down the bridge entirely should no longer be (or need to be) our first course of action, and it sounds like we are moving toward that which is good 18:33
2. We need some channel of communication between moderators on either side and some standards on how to handle situations where there is inappropriate behavior spotted by someone who has no power on the “other side” 18:34
lizmat agree on 1 and 2 18:35
Rog On our end we essentially have a team of two, one of whom is me, and I have quite a bit going on atm
Zephyr is doing an excellent job thankfully but maybe we need one more person in my time zone for coverage. idk, will figure that out later 18:36
As for this situation in particular, I have only had time to glance through the relevant discussion, but my impression is that everybody has the same ultimate goal of making Raku a success, but the desire to be argumentative if creating division where there need not be any 18:38
18:38 freeside joined
s/if/is/ 18:38
lizmat Rog indeed
indeed : some of the users describe interactions as ranging from "everything is a fight" to "toxic" :-( 18:39
Rog My expectation here is that we drop the negative tone and always assume the best of other people rather than jumping to snark or attacks 18:40
lizmat whereas there's never a mention of malicious intent to harm Raku
Nemokosch by the way: Daniel Sockwell texted me yesterday and hopefully you received my mail, too
seemed like the intention was to have a discussion in the upcoming days
lizmat Rog I hope so 18:41
Nemokosch indeed
Rog I will try to stay more active over the coming weeks and get a bead on how things feel 18:45
Nemokosch Just sticking to the tighter events of the last couple of days, that consists of multiple topics, ranging from my personal contribution to the relations within the active part of the community and better targeting the right messages to the right audience overall 18:46
18:47 freeside left
lizmat Rog thanks! 18:48
18:49 freeside joined 18:50 swaggboi joined
cfa uh, sorry if i prompted that 18:51
18:53 freeside left
Nemokosch I don't think so; it's just these things kinda look different on Discord. 18:59
I don't think replies are attached in any shape or form
19:02 jpn left, sena_kun joined 19:09 pingu left 19:11 derpydoo joined 19:25 jpn joined 19:34 freeside joined 19:39 freeside left
[Coke] on windows, I have $dir = "C:\...", and can do `dd dir $dir.IO` and get a subfolder. if I add that subfolder to the path, I can get True on .IO.e, .IO.d, but if I dd dir $dir.IO, I get an error reading the dirhandle. 20:29
gfldex [Coke]: is .resolve or .absolute changing the behaviour? 20:31
20:32 jpn left
[Coke] gfldex: just tested: nope. 20:34
20:34 freeside joined
[Coke] before invoking dir and failing, the path name is 258 chars. 20:35
gfldex [Coke]: there used to be filemon from sysinternals that would should the error as reported by the win32 API. Sadly, I can't remember to what they renamed that tool. 20:36
the limit is 260 characters. Unclear if that is really characters or bytes. 20:37
[Coke] I'm guessing this is related to the long filename issue (but the file in that directory *exists*, and I can cd there in git bash and look at it.)
gfldex ohh, the volume is not included
so you are beyond the limit
see: learn.microsoft.com/en-us/windows/...limitation
20:39 freeside left
[Coke] ok, so if I update the registry this might work. 20:40
gfldex Frankly, if you got the problem, you are very likely heading for another cliff.
Have you considered to use a proper OS? :-> 20:41
20:45 jpn joined
gfldex [Coke]: I believe the proper way of dealing with this is to avoid absolut paths. 20:46
ugexe Thing is, libuv uses UNC paths which isn’t beholden to that rule 20:48
so I dunno why it this length issue occurs in the first place
[Coke] my path explicitly has a C:\ in it, if that matters.
(well, this test file. the original was relative, I think) 20:49
I updated the registry, rebooted, the error is still occuring.
the page you linked says we also have to update the "application manifest", but I don't think we have one of those for rakudo. :|\ 20:50
ugexe github.com/libuv/libuv/issues/2331 For example 20:51
I wonder if we use a non libuv file op somewhere
[Coke] Looks like MVM_dir_open is not using libuv on windows? 20:56
ugexe Nice, can you link it? I’m on my phone 20:57
[Coke] github.com/MoarVM/MoarVM/blob/mast...ops.c#L217
in fact, it looks like non windows is just calling "opendir", not anything in libuv. 20:58
looks like nine wrote most of that in 2019. 20:59
ugexe yeah I think replacing that with whatever the libuv variant is might solve the path length issue 21:04
21:05 jpn left 21:06 cfa left 21:08 swaggboi left
ugexe hmm maybe thats not it either though. `raku -e "my $path = q|a| x 400; say $path.IO.open(:create)"` fails with no such file 21:13
if i s/400/40/ it works
github.com/MoarVM/MoarVM/blob/34ba...rops.c#L71 this function looks like it might need to remove the MAX_PATH check and just always run the logic 21:19
i wish i had an easy way to get a stack trace on windows... or how even :P 21:20
knew how^
[Coke]: do you get the same error with that one liner i posted? and have you done the supposed enabling of long paths in the windows registry or whatever? 21:27
my vmware trial ends in 3 days :( 21:28
[Coke] ugexe: your oneliner: No such file or directory 21:36
(works with 40)
I did update the registry and rebooted
I can probably setup an azure VM and get folks access, but I don't recall what version of windows you can get with that. 21:37
MAX_PATH is also checked in mkdir 21:39
ugexe same with MVM_dir_cwd 21:43
[Coke] I'll try a build that just removes that check and see if it explodes. 21:44
(and put the registry back and see how it explodes there.)
ugexe thing is the one-liner i posted shouldnt use mkdir 21:45
it just creates a file in the current directory
so while reading a directory might be the cause, creating one would seem unlikely (at least in the case of the one liner)
the good news is updating moarvm to use new libuv uv_ ops is pretty easy and one can basically just look at a PR for any of the existing ops 21:46
usually just writing a small wrapper around the function
they also don't need a windows machine to do it 21:47
[Coke] OK. that's probably better than trying to half-rip out MAX_PATH. 21:53
ugexe: added github.com/MoarVM/MoarVM/issues/1743 22:02
22:08 evalable6 left, linkable6 left, evalable6 joined 22:11 linkable6 joined 22:14 ab5tract left 22:29 QhpAptyj9hj0RQwM joined 22:30 NemokoschKiwi joined 22:34 QhpAptyj9hj0RQwM left
ugexe so my golf is a bit misguided in a way 22:43
22:44 sena_kun left
ugexe `raku -e 'my $path = q|a| x 255; say $path.IO.mkdir'` # works, even though im inside a directory thats like 30 characters long already 22:44
`raku -e 'my $path = q|a| x 256; say $path.IO.mkdir'` # mkdir error 123
i mention this because even with long path names enabled, each directory part / file name is limited to 255
thing is while the precomp errors on windows suggest path length issues, they aren't creating single path parts that are that long 22:45
the above examples i ran with long path support enabeld in windows registry 22:46
something like a rakudo-gdb-m dump would be pretty valuable 22:48
it seems like when i try to create a file it takes the entire path length into consideration, not each path part individually 22:56
22:59 NemokoschKiwi left 23:07 swaggboi joined 23:35 freeside joined 23:39 freeside left 23:40 jgaz 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!