03:22 kjp left 03:29 kjp joined
patrickb Assuming in a module there are many Rakudoc files (in whatever folders). I'd like to provide a way to put those files in some order. What would be a good way to specify that order? 10:45
Some ideas:
- Sort by file name (maybe only as fallback if no explicit order was specified) 10:46
- A top level JSON file ("CHAPTERS.json") containing a sole list of file paths. 10:47
- A new META6.json entry, provides-docs, containing a list of filenames. (Inspired by gist.github.com/ugexe/8caac54436b4...d6fb4fb31) 10:53
lizmat yes! 10:54
patrickb :-P
lizmat the Monday after FOSDEM I will be at a conference: repos.openssf.org/principles-for-p...urity.html 10:55
and will probably come up with some more entries for META6.json (or some other distro file, although I don't see a reason to create a separate file for that yet) sjn ideas ?
Geth rakudo: rc-chuah++ created pull request #6066:
MSI, LICENSE & README: Update All Ranged Copyrights To 2026: 		11:25
11:46 finanalyst joined
Geth rakudo/main: 12 commits pushed by (RC Chuah)++, (Patrick Böker)++
review: github.com/rakudo/rakudo/compare/9...2709248ec3		 11:51
14:45 rakkable left 14:46 rakkable joined
ab5tract I feel like I'm taking crazy pills.. now matter how I try to attach a new RakuAST::VarDeclaration::Implicit subclass to the AST tree, I'm not getting any traction. 15:36
for example, adding via PRODUCE-IMPLICIT-DECLARATIONS fails to result in IMPL-QAST-DECL being called 15:37
timo i'm just guessing here; can you double-check the method resolution order? 15:38
is there roles involved or something?
ab5tract nope, no roles (as RakuAST dialect is just a jazzed up NQP) 15:40
timo nqp has roles too i thought
ab5tract hmm, well everything in RakuAST at least is done via class / is 15:41
timo my idea is bust, then
should we try diagnosing together?
ab5tract it's just confusing because PRODUCE-IMPLICIT-DECLARATIONS usually "injects" the declaration node just fine (IMPL-QAST-DECL is called at the very least, then also IMPL-EXPR-QAST etc) 15:43
but in this case it is not affecting the QAST output at all :/ 15:44
timo can you tell apart whether the function isn't called at all or if it just results in the same output?
sjn lizmat: happy to share some notes with you on metadata ideas 15:45
or explore, if needed :-)
ab5tract my handy little print statements are never emitted, so that's an extra confirmation in addition to the fact that none of the QAST I'm trying to add is present
lizmat sjn: there's github.com/Raku/problem-solving/issues/491 15:47
ab5tract A new version of META would be very appreciated. version strings should be deprecated in favor of proper version objects 15:48
This simple change would be a huge improvement for anything that needs to parse META6.json
Which otherwise requires implementing Raku's adverb syntax 15:49
ugexe you can already write dependencies without having to write e.g. :ver<1.0>
ab5tract yes, I want to push that further by not allowing the adverb form at all 15:50
the biggest obstacle to META6.json handling is the multiple shapes that each thingy can take 15:51
timo without more details i'm not sure i can give any helpful pointers
ugexe i would argue the biggest problem with meta6.json handling is the mutating of meta6 data on install 15:53
for example `provides`
timo could the `$!implicit-declarations-cache` already be filled by the time you're hoping your code would run, so get-implicit-declarations on some RakuAST::ImplicitDeclarations derived node doesn't run your code?
ugexe but parsing the adverbs would be a close second :)
(see github.com/rakudo/rakudo/pull/2778) 15:55
ab5tract timo: ah nice catch. I've checked again and the PRODUCE-IMPLICIT-DECLARATIONS is called 16:10
which I interpret to mean that the cache is not the issue 16:11
ugexe: there's also the way that dependencies will change shape when there's architecture specific stuff going on 16:12
I haven't bothered to fix this in the intellij plugin 16:13
ugexe i dont see how that can be avoided 16:14
ab5tract If dependencies were always objects, then it would just be an optional field within a well-defined object 16:19
as opposed to jumping between a list of strings and objects 16:20
ugexe i'm not following how that applies to the architecture comment. assume someoen writes a meta6.json using the existing object dependency definitions. how does the architecture comment apply? 16:21
i ask because what you're describing sounds like github.com/ugexe/zef/blob/2c6ab9ff...st#L20-L30 16:22
oh you mean that "name" is an object in my link, but it could also be (and normally would be) a string 16:23
ab5tract ah yeah, that's exactly what I meant 16:40
I was mis-remembering the structure a bit there 16:56
21:06 librasteve_ joined 22:44 finanalyst left

Powered by Cro and the Raku® Programming Language.

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

Thank you!