04:43 hurufu joined 05:20 hurufu left, hurufu joined 05:24 hurufu left 05:25 hurufu joined 06:12 [Tux] left 06:14 [Tux] joined 08:43 lizmat left 08:44 lizmat joined 09:23 apogee_ntv left 09:26 apogee_ntv joined
Geth rakudo/main: ffecffdaef | (Elizabeth Mattijsen)++ | 3 files
Remove cases of Mu.POPULATE being called during setting compilation

The idea being that in some future, Mu.POPULATE won't need to exist anymore because each class will get its own method POPULATE at class composition. Which in turn would make the BUILDPLAN obsolete.		 09:27
10:21 apogee_ntv left 10:22 apogee_ntv joined
[Tux] Rakudo v2026.05-37-gffecffdae (v6.d) on MoarVM 2026.05-3-g236fcdf7c
csv-ip5xs0.269 - 0.274
csv-ip5xs-201.071 - 1.359
csv-parser1.073 - 1.081
csv-test-xs-200.112 - 0.116
test1.728 - 1.817
test-t0.480 - 0.487
test-t --race0.297 - 0.308
test-t-205.869 - 6.276
test-t-20 --race1.420 - 1.422
 11:09
csv-test-xs 0.014 - 0.016
tux.nl/Talks/CSV6/speed4-20.html / tux.nl/Talks/CSV6/speed4.html tux.nl/Talks/CSV6/speed.log
11:26 apogee_ntv left 11:27 apogee_ntv joined 11:38 hurufu left, apogee_ntv left 11:47 hurufu joined
Geth rakudo/lizmat-7: 106348dfdf | (Elizabeth Mattijsen)++ | 2 files
RakuAST: streamline Nodify lookup

Instead of calling .Nodify with name parts, always call it with a fully qualified class name as a string (minus the RakuAST:: part). This allows a simple cache to be built for these class lookups.
Also introduce a .Nodify method in the Grammar to more easily (and recognizable) places where class lookups are done.
This appears to take off about 1 second from stage parse of the c setting in RakuAST (33 -> 32 seconds), at least for me.		 12:14
rakudo: lizmat++ created pull request #6209:
RakuAST: streamline Nodify lookup
rakudo/lizmat-7: 7f42a520b8 | (Elizabeth Mattijsen)++ | src/Raku/Actions.nqp
RakuAST: Fix a few stragglers that were missed

And which showed up in CI		 13:19
rakudo/lizmat-7: 5d61a1afef | (Elizabeth Mattijsen)++ | src/Raku/Grammar.nqp
RakuAST: revert to .r interface

These roles do not consume the Raku::Common class, so need to use the old "r" interface.
japhb lizmat: dev.to/lizmat/moving-printf-format...rward-1m3p refers to '2023.06' (rather than '2026.06') in the last paragraph 16:31
lizmat japhb: that's because it's from 3 years ago 19:14
most of the work was done then, but basically stalled because of the synthetic RakuAST precomp issue
lizmat now wonders why RakuAST::Trait::Is needs a RakuAST::Name instead of just a str 19:15
aah... because of something like "is IO::Path" 19:17
check
ugexe: so I think we have a chicken/egg problem with creating accessors in RakuAST 19:56
Package::Attachable.PRODUCE-META-ATTACHABLES
so: where it says create method POPULATE here 19:57
I can create the AST for such a method from the info in $!attached-attributes
but how do I get a meta-object from that to do a $how.add_method with ? 19:58
so it feels like it is too late, as everything else has been resolved already at that point, and the method doesn't take a resolver 19:59
21:50 hurufu left, hurufu joined
ugexe Take a look at RakuAST::Method::AttributeAccessor and the way CompilerServices.generate_accessor drives it 22:31
22:55 hurufu left
ugexe I assume the step you're missing is calling .to-begin-time($resolver, $context) before .meta-object 23:56

Powered by Cro and the Raku® Programming Language.

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

Thank you!