There is no easy way to use an EXPORT sub to export a long named spaces class

[FCO]

If I try to do something like:

my \bla = Metamodel::ClassHOW.new_type: :name<Bla::Ble::Bli::Blo::Blu>;
bla.^compose;

sub EXPORT(--> Map()) {
    "Bla::Ble::Bli::Blo::Blu" => bla
}

When I use that, it adds the whole name as a key on ::.keys, which doesn't work.

If I try something like:

module Bla::Ble::Bli::Blo {}
Bla::Ble::Bli::Blo::<Blu> := bla;

It tends to conflict with modules declared outside that file.

I belive it should be an easy way of doing that.

[raiph]

Hi @FCO

Can you provide a link to the discussion or document that demonstrated or led to the conclusion in the title? I'm asking because I'm hoping to at least understand the issue. :)

If there's no such discussion / document link, or even if there is, please consider the following code:

# Bla/Ble/Bli/Blo/Blu.rakumod
my \foo = Metamodel::ClassHOW.new_type: :name<Bla::Ble::Bli::Blo::Blu>;
foo.^compose;

sub EXPORT(--> Map()) {
  package Bla::Ble::Bli::Blo::Blu {}
  Blu => foo;
}

# Bla/Ble/Bly.rakumod
unit class Bla::Ble::Bly;

# main.raku
use lib '.';
use Bla::Ble::Bli::Blo::Blu;
use Bla::Ble::Bly;

say Bla::Ble::.keys; # (Bly Bli)

say Blu; # (Blu)

My guess is that the package Bla::Ble::Bli::Blo::Blu {} is the problem in that I've manually written it but your scenario requires that it's programmatically generated. Is that indeed a problem? If so, is it the essence of the problem or is there more to it?

[FCO]

The discussion started here: https://irclogs.raku.org/raku-dev/2025-02-17.html#09:52-0004

But yes, I want to access Blu calling the complete name (Bla::Ble::Bli::Blo::Blu) and not only the small name.

On your example, the exported Bla::Ble::Bli::Blo::Blu is not the created class (what was the intention).

Here is an example: https://glot.io/snippets/h5vzljy8k0

[raiph]

@FCO

Thanks for the links. :)

• I should of course not have declared Bla::Ble::Bli::Blo::Blu in the keyword declaration of a package. I should instead have declared Bla::Ble::Bli::Blo, i.e. dropping the ::Blu. Just as you did in your opening comment. Sorry about that mistake.

• I should of course have then had a follow on line Bla::Ble::Bli::Blo::<Blu> := foo;. Again, just as you did in your opening comment. Sorry about that mistake too.

With those two corrections, I have tried but all but failed to reproduce a problem. Here is where I've decided to stop: https://glot.io/snippets/h5wug9yupc

That said at the end I concluded there were three things that might be what you are focused on:

• I am writing literal package declarations, with declaration keywords, rather than generating packages using metaprogramming. I noted this in my first comment. Perhaps this is a part of your problem?

• You wrote "It tends to conflict with modules declared outside that file" about code you showed that used a literal package declaration. I've declared modules in separate files (i.e. that are neither the main.raku nor the generated class file Bla::Ble::Bli::Blo::Blu), and given them the same prefixes (i.e. Bla:: or Bla::Ble:: or Bla::Ble::Bli:: or Bla::Ble::Bli::Blo or Bla::Ble::Bli::Blo::Blu), and reordered the use statements. I saw nothing go wrong. I got the blu i want displayed as expected for a say Blu; statement no matter what variants I tried with many files/compunits. (To be clear I did not push far. I didn't have modules useing other modules with deep nesting etc.) What did you mean by "tends to conflict with modules declared outside that file."? What does "tends" mean? What does "conflict" mean? Most importantly, can you provide a test case that fails in some way?

• I found one issue that hints at a possible source of problems. When I used package instead of module as the declarator of a namespace (one or more packages deep) I had to also declare at least one type whose short name (and thus symbol that got added to the lexical symbol table) matched at least one of the namespace levels of the package declaration, otherwise a runtime error occurred. For example, when I switched the module Bla::Ble::Bli::Blo {} line to package Bla::Ble::Bli::Blo {} I had to precede or follow it with, say, subset Bar; or grammar Bar::Ble {} or whatever. If I did, everything was fine. But if I did not, then I got a Could not find symbol '&Blu' in 'Bla::Ble::Bli::Blo' error instead of the blu i want.

Hope the above provides some progress in at least understanding the problem.

[FCO]

So, my error was not define the module inside the EXPORT sub?

I think you solution fixes my problem.

Thank you!

[raiph]

@FCO,

So, my error was not define the module inside the EXPORT sub?

I don't know. I was just trying to explore within the limits of my understanding of what seemed like it could work, perhaps should work, perhaps would work. That said, I'm hoping I'm on the right track.

I think you solution fixes my problem.

Hopefully! 🤞

---

Couple things to clear up:

• Maybe something like BEGIN (my $pkgname = 'foo::bar'; "module $pkgname \{}".EVAL); is necessary to dynamically code gen the compile time module declaration.

• I found issues related to the oddity I noted that led to Could not find symbol '&Blu' errors. Here's a comment I just added to one of them.