🦋 Welcome to the IRC channel of the core developers of the Raku Programming Language (raku.org #rakulang). This channel is logged for the purpose of history keeping about its development | evalbot usage: 'm: say 3;' or /msg camelia m: ... | Logs available at irclogs.raku.org/raku-dev/live.html | For MoarVM see #moarvm
Set by lizmat on 8 June 2022.
00:00 reportable6 left, reportable6 joined 00:05 kjp left 00:37 kjp joined 04:00 notable6 left, unicodable6 left, squashable6 left, tellable6 left, coverable6 left, linkable6 left, sourceable6 left, evalable6 left, bisectable6 left, reportable6 left, nativecallable6 left, shareable6 left, releasable6 left, benchable6 left, quotable6 left, greppable6 left, statisfiable6 left, bloatable6 left, committable6 left, statisfiable6 joined 04:01 bisectable6 joined, tellable6 joined, shareable6 joined, sourceable6 joined, evalable6 joined, committable6 joined, notable6 joined 04:02 coverable6 joined, reportable6 joined, greppable6 joined, nativecallable6 joined, bloatable6 joined, quotable6 joined, squashable6 joined, unicodable6 joined 04:03 releasable6 joined, linkable6 joined, benchable6 joined 05:41 Xliff left 06:00 reportable6 left 06:01 reportable6 joined
bartolin btw, it looks like this would unbreak the JVM build: gist.github.com/usev6/020176b7f817...01e7f4b917 06:21
It seems to have worked fine in the past and I didn't see yet when/why the old code started to fail. 06:22
anyway, off to $dayjob now &
07:25 squashable6 left 07:26 squashable6 joined 07:28 ab5tract joined 08:29 sena_kun joined 09:49 squashable6 left, bisectable6 left, bloatable6 left, evalable6 left, statisfiable6 left, releasable6 left, benchable6 left, greppable6 left, reportable6 left, nativecallable6 left, linkable6 left, quotable6 left, sourceable6 left, coverable6 left, shareable6 left, unicodable6 left, tellable6 left, committable6 left, notable6 left 09:50 unicodable6 joined, squashable6 joined, sourceable6 joined, statisfiable6 joined, linkable6 joined, quotable6 joined, shareable6 joined, notable6 joined 09:51 tellable6 joined, coverable6 joined, evalable6 joined, releasable6 joined, bisectable6 joined, bloatable6 joined 09:52 greppable6 joined, nativecallable6 joined, committable6 joined, reportable6 joined, benchable6 joined 09:57 Altai-man joined 09:59 sena_kun left
lizmat m: dd "AA".parse-base(16) 09:59
camelia 170
lizmat m: say "A".uniname # TIL you can also use full-width letters for hex numbers ?
ab5tract I'm starting to understand why qualified method calls are NYI 10:28
talk about torturing the implementor... 10:30
lizmat hehe :-) ++ab5tract 10:33
ab5tract I've got a situation where I need a type object in order to fill in a QAST::WVal, but if I use implicit-lookups and .resolution.compile-time-value, it will fail if the type doesn't exist. 10:42
I also don't have a resolver object in scope for this method 10:43
Geth rakudo/main: d3b53807ea | (Elizabeth Mattijsen)++ | src/Raku/Grammar.nqp
RakuAST: moved some elements of HLL::Grammar to Raku::Grammar

As a start to completely lose NQP's HLL::Grammar as a base class, which is needed to be able to use OperatorProperties properly during parsing and be able to get rid of the operator info hashes.
Also added some documentation markers for better readability
lizmat ab5tract: I think in some cases his is fixed by adding an attribute, and putting the type object in it at resolve timer 11:06
ab5tract ack, thanks 11:08
lizmat: is there any way to "create" a type object (just the object, no installation or anything) without resolution? 11:37
lizmat does it exist in the setting?
ah, no, you'd need a resolver object for that as well 11:38
ab5tract it's a user-defined (or not) type 11:51
lizmat I don't know of a way without a resolver 11:53
ab5tract m: role A { method m() { self.x } }; class B does A { has $.x = 42; method m() { self.D::m() } }
camelia ( no output )
ab5tract m: role A { method m() { self.x } }; class B does A { has $.x = 42; method m() { self.D::m() } }; B.new.m 11:54
camelia Cannot dispatch to method m on Failure because it is not inherited or done by B
in method m at <tmp> line 1
in block <unit> at <tmp> line 1
ab5tract ^^ the above doesn't check whether the type is defined or not
but that actually seems to make sense according to the base compiler (it doesn't complain about not finding D until run-time) 11:55
12:00 reportable6 left, reportable6 joined
lizmat yeah, method call lookup is runtime, whether fully qualified or not 12:00
m: class A { }; A.B::foo() 12:01
camelia Cannot dispatch to method foo on Failure because it is not inherited or done by A
in block <unit> at <tmp> line 1
lizmat that's actually an LTA error
looking at the QAST in the legacy grammar 12:02
it appears to be doing a nqp::who($?PACKAGE).AT-KEY("B").foo 12:04
Geth rakudo/ugexe/bartolin-fix-jvm-maybe: 3c3c40cca4 | (Nick Logan)++ | src/Perl6/Actions.nqp
Fix JVM build
rakudo: ugexe++ created pull request #5365:
Fix JVM build
rakudo/ugexe/bartolin-fix-jvm-maybe: 1227b40723 | (Nick Logan)++ | src/Perl6/Actions.nqp
Fix JVM build
ab5tract well, my half-baked fix does bring +3 spectests into the fold 12:11
lizmat so why is it half-baked then? 12:16
ab5tract it behaves differently from base in that it dies when the qualified package is not found 12:22
I have a chicken-egg situation with constructing the call to dispatch:<::> -- it wants a type object, but if the qualified package is not resolvable... 12:23
let me push, maybe that will make things clearer
Geth rakudo: ab5tract++ created pull request #5366:
RakuAST: Fix t/spec/S14-roles/basic.t
ab5tract lizmat ^^ 12:28
lizmat ab5tract some nits and questions 12:30
otherwise looks good :-)
ab5tract nits and questions should now be addressed 13:10
&& shortcircuits in NQP and it's RakuAST variant, right? 13:17
lizmat yes, && shortcircuits everywhere 13:22
afaik :-)
Geth rakudo/main: fc9debac98 | (Elizabeth Mattijsen)++ | 2 files
RakuAST: Streamline integer handling a bit

  - make these tokens better readable
  - don't capture what doesn't need capturing
  - copy "integer" action from HLL::Actions to reduce dependency
rakudo/main: a369b57fbe | (Elizabeth Mattijsen)++ | 2 files
Make X::Worry::P5::LeadingZero exception more readable

And adapt test that was a little too keen on specific format
rakudo/main: 1227b40723 | (Nick Logan)++ | src/Perl6/Actions.nqp
Fix JVM build
rakudo/main: c12f478a1c | (Nick Logan)++ (committed using GitHub Web editor) | src/Perl6/Actions.nqp
Merge pull request #5365 from rakudo/ugexe/bartolin-fix-jvm-maybe

Fix JVM build
ugexe indeed, that fixed the JVM ci tests
hmm actually there was a linux and mac test failure that look like they are unrelated, but maybe they werent 14:18
Lin_MVM_reloc passed but Lin_MVM failed a test, and Mac_MVM_reloc failed a test but Mac_MVM passed 14:19
14:22 evalable6 left, linkable6 left, evalable6 joined, linkable6 joined 15:22 committable6 left, linkable6 left, reportable6 left, statisfiable6 left, unicodable6 left, nativecallable6 left, bisectable6 left, tellable6 left, benchable6 left, releasable6 left, coverable6 left, evalable6 left, shareable6 left, sourceable6 left, greppable6 left, quotable6 left, squashable6 left, notable6 left, bloatable6 left 15:23 greppable6 joined, releasable6 joined, notable6 joined, sourceable6 joined, bisectable6 joined 15:24 linkable6 joined, unicodable6 joined, benchable6 joined, reportable6 joined, shareable6 joined, nativecallable6 joined, statisfiable6 joined, committable6 joined, bloatable6 joined, coverable6 joined 15:25 squashable6 joined, quotable6 joined, tellable6 joined, evalable6 joined
ab5tract hmm, so the regressions in my branch are both related to assigning Inf/Nan to Int 15:45
Geth rakudo/main: c4674f7000 | (Elizabeth Mattijsen)++ | src/Raku/ast/operator-properties.rakumod
Add OperatorProprties support for allowable openers/closers

For custom (post)circumfix operators. Adds the following class methods:
  - bracket-openers returns all ok openers as a single string
  - bracket-closers returns all ok closers as a single string
  - bracket-closer-for-opener return closer for given opener, if any
  - bracket-opener-for-closer return opener for given closer, if any
lizmat ab5tract: that feels.... odd?
ab5tract It goes from a sorry t(with slightly misleading instructions that suggest coercing Inf.Int or Int(Inf)) to a runtime assignment failure (Type check failed in assignment to $f; expected Int but got Num (Inf))
yeah, I mean I fiddled with the logic that controls dispatching to that sorry, but I can't see where the current code would let NaN and Inf slip through 15:51
"not ok 22 - trying to assign NaN to Int gives a helpful error"
technically the new behavior still fulfills the spirit of the test 15:52
lizmat except maybe now it's a runtime rather than a compile time error ?
ab5tract well, that's why it's failing the reality of the test :) 15:53
lizmat m: my Int $a = Inf
camelia ===SORRY!=== Error while compiling <tmp>
Cannot assign a literal of type Num (Inf) to a variable of type Int.
You can declare the variable to be of type Real, or try to coerce the
value with Inf.Int or Int(Inf).
at <tmp>:1
------> m…
lizmat m: my Int $a = val("Inf").Num 15:54
camelia Type check failed in assignment to $a; expected Int but got Num (Inf)
in block <unit> at <tmp> line 1
lizmat so the sorry is not creating a panic?
ab5tract yup, that's the distinction there
nope, the sorry is just never thrown and it is caught later on 15:55
lizmat ah... hmm
ab5tract and the tests are checking the specific syntax error class, so those fail 15:58
something about github.com/ab5tract/rakudo/blob/2d...kumod#L736 16:00
m: dd NaN ~~ Numeric
camelia Bool::True
ab5tract actually, $vartype there should be set to Int, IIUC 16:15
ugexe main indeed has a green CI again 17:18
lizmat ah, looks like Github is not delivering webhooks atm 17:52
18:00 reportable6 left, reportable6 joined
ab5tract lizmat: my PR is fixed. now +4 to spectest :) 18:05
lizmat whee!
so ready to merge ?
18:06 Geth left, Geth joined
Geth rakudo/main: c21f4b6b40 | (Elizabeth Mattijsen)++ | src/Raku/ast/operator-properties.rakumod
Add reserved operator knowledge to OperatotProperties

  - is-reserved-operator return whether given op name is reserved
  - reserved-operator-hint return hint for reserved operator name
rakudo/main: ceb3d4078c | (Elizabeth Mattijsen)++ | src/Raku/Actions.nqp
RakuAST: add ::Action.OperatorProperties

to be able to access the OperatorProperties class in the grammar. At least for now, until we have a better solution
lizmat wow, about 30 minutes delay
ab5tract yeah, it should be ready to merge 18:28
I assume nine is okay with the use of the resolver to grab Numeric 18:30
lizmat yeah, that should be ok: if you have a resolver handy, why not use it? 18:31
I've merged, but it seems Github has a 30 minute delay in delivering webhooks atm
so it will be a while before you see it here
on Twitter: "We’ve mitigated the issue causing Webhooks latency and are monitoring processing of the backlog of queued requests." 18:33
942 +4 :-) 18:40
Geth rakudo/main: 978e2d7898 | ab5tract++ (committed using GitHub Web editor) | 4 files
RakuAST: Fix various Role issues

This allows statements like `role R {}; my R $r = 42` to fail at run-time, as expected.
Fixes the following code:
   role R { has $.n }
   class C does R { has R $.r .= new(:$!n) }
   my C $c .= new(:n(1));
   dd $c.r.n
lizmat only 15 minutes delay :-)
ab5tract it seems you squash merged it. I was being very careful with the commits, as there were things like adding :ver<>:auth<>:api<> along the way 18:55
stuff that was mostly orthogonal to roles in the specific case 18:56
lizmat sorry want me to revert and merge again?
ab5tract no big deal 18:57
lizmat sorry again :-( 18:58
ab5tract I don't know if it's worth rewriting history over :)
it's all good. next time!
lizmat ok, will keep in mind (hopefully) 18:59
ab5tract I'll try to remember to mention it, as well
well, on second thought.. how big a deal would it be to revert and merge again? 19:00
lizmat not very
ab5tract if it's not a mega hassle, I think future code writers might appreciate it
Geth rakudo/revert-5366-fix-roles-basic: 96b84ac43b | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | 4 files
Revert "RakuAST: Fix various Role issues"

This reverts commit 978e2d789870885b8b37ec4da4f6808fe900be7a.
rakudo: lizmat++ created pull request #5367:
Revert "RakuAST: Fix t/spec/S14-roles/basic.t "
lizmat hmmm.. that was not what I was going for :-(
ok, much more complicated than I thought :-( 19:02
ab5tract then: next time :) 19:03
19:06 linkable6 left, evalable6 left 19:08 linkable6 joined 19:09 evalable6 joined 19:22 melezhik joined
gfldex m: &fail andthen :continue; 19:25
camelia WARNINGS for <tmp>:
Useless use of ":continue" in sink context (line 1)
gfldex m: quietly &fail andthen :continue;
camelia WARNINGS for <tmp>:
Useless use of ":continue" in sink context (line 1)
19:28 melezhik left
ab5tract gfldex: it doesn't give that warning with RAKUDO_RAKUAST=1 .. bug or feature? 19:32
gfldex I would say that is a bug. 19:33
We got those sink context warnings for a reason.
ab5tract indeed
but atm it doesn't have a single complaint about your poetry 19:34
gfldex As I understand it, those sink context warnings where put into place to indicate precedence mistakes with chaind infix operators. 19:38
So they also may be LTA. 19:39
vrurg ugexe: do we have a semi-legal way of pulling out particular dependencies from a precompilation? 20:00
ugexe i dont believe so. i'm not sure it could even know its dependencies since they could potentially not even come about until runtime 20:08
i guess if it was possible to get at its $?DISTRIBUTION object you could sort of do it that way, but i don't think thats quite what you're looking for 20:10
vrurg Nah... I would be happy to get just those which are known at the compile time. 20:15
After all, the bytecode loaders knows them anyway.
ugexe if they know it would be because the sha1 is known, but sha1 is only one direction 20:16
so im not sure it can be determined
vrurg Makes sense. OK then. 20:17
It was my last hope to try optimizing app compilation time. Full precomp with all the dependencies takes ~16mins. 20:18
ugexe oh maybe they can be, although i dunno if its legal 20:19
stuff like that is in the repo files
so maybe just naively parse out the CompUnit::DependencySpecification 20:20
i think the parts are \0 separated
vrurg They are. Hm, at first I hoped to avoid that, but why not, after all? 20:22
Thanks! 20:23
ugexe i wonder if it makes sense to have some sort of api to do it 20:24
vrurg I think it does. Xliff would most certainly be happy about it. It seems like me too now. ;) 20:26
ugexe for instance with RakuAST i wonder if the same functionality could be done on source code
so then all Repository objects could have some dependencies-for-spec(...) or some such 20:27
the precomp repository objects would have the same method, except it would determine them by e.g. reading and parsing that file
that should hide most implementation details 20:28
vrurg Sounds like a plan to me. RakuAST may not know all the compile-time dependecies because some might occur in BEGIN phasers via, say, `require`. But then they are available before precomp is produced. 20:33
ugexe m: say CompUnit::PrecompilationDependency::File.deserialize("BF820EA236973A1CB5A7F26B2F7AA2FAA9F9BA0A\0site#sources/BF820EA236973A1CB5A7F26B2F7AA2FAA9F9BA0A\0837991E401FD609CA0F33912991916EF0B2FF465\0CompUnit::DependencySpecification.new(:short-name<XXX>,:version-matcher<0.1.0>,:auth-matcher<github:ugexe>,:api-matcher<0>)").spec 20:41
camelia CompUnit::DependencySpecification.new(:short-name<XXX>,:version-matcher<0.1.0>,:auth-matcher<github:ugexe>,:api-matcher<0>)
ugexe so that method can parse the precomp dependency lines 20:42
its still doing an EVAL though :/
vrurg And this is what I call 'semi-legal'. :) It would be preferable to have just keys alone in the precomp. I was even considering using JSON at some point. But that's too much for the task. 20:45
22:23 statisfiable6 left, evalable6 left, tellable6 left, nativecallable6 left, greppable6 left, reportable6 left, sourceable6 left, committable6 left, shareable6 left, releasable6 left, coverable6 left, bisectable6 left, squashable6 left, unicodable6 left, benchable6 left, linkable6 left, notable6 left, bloatable6 left, quotable6 left 22:24 linkable6 joined, bisectable6 joined, squashable6 joined, shareable6 joined, evalable6 joined, unicodable6 joined 22:25 coverable6 joined, reportable6 joined, quotable6 joined, notable6 joined, committable6 joined, greppable6 joined, sourceable6 joined 22:26 nativecallable6 joined, bloatable6 joined, benchable6 joined, statisfiable6 joined, releasable6 joined 22:27 tellable6 joined 22:53 Altai-man left 23:55 japhb left