Re: Brainstorming: quoted identifier operator

On Wed, 7 Jan 2026 at 12:44, Philippe Bruhat (BooK) <[email protected]>
wrote:

> On Wed, Jan 07, 2026 at 12:22:47PM +0100, Christian Walde wrote:
> > On 07/01/2026 12:16, Dave Mitchell wrote:
> > > On Wed, Jan 07, 2026 at 10:56:03AM +0100, Branislav Zahradník wrote:
> > > > just another idea based on my everyday work - let's defined new
> quoting
> > > > operator: quoted identifier.
> > > ...
> > >
> > > You don't say what this quoting operator actually does.
> >
> > The proposed example is:
> >
> > sub qi (this is nicely readable test name) {}
> >
> > This tells me the intent is to be able to declare sub names similar to
> how
> > prototypes are handled, i.e. with literally anything in them.
>

Yes, anything (though primary usage is for use case where current
identifier is transformation of real text - to avoid code duplication)


>
> How would the sub be invoked, though?
>
> Like this?
>
>   qi(this is nicely readable test name)( @args )
>

Yes, qi() will act as as single token (bareword).


>
> Surely not this, right?
>
>   qi(this is nicely readable test name)->( @args )
>

Of course, this will remain as it is - invoke coderef returned by function


>
> We already have syntax for declaring and calling subs with unusual names:
>
>     no strict 'refs';
>     *{"some readable name"} = sub { say "foo" };
>     &{"some readable name"}();
>
> The fact it doesn't look so great might actually be a good thing...
>

As I tried to mentioned (if I failed I'm sorry), I can see effective usage
only integration
scenarios, where primary identifier of behaviour is a string
(non-identifier).

what I can imagine, is for example introduction of new prototype I
(identifier), and capability to
declare for example:
subtest (i&)

so one will be able to write:
subtest qi (validation errors) {
}

(but that is for future, I don't have vision yet how to implement this)


> Identifiers are everywhere, including variables, does that mean we could
> do this?
>
>   $qi(my favorite variables have long, descriptive names)
>

I didn't think about this (part of brainstorming is pointing it out).

For sigil symbols I'd restrict such usage to braces syntax: ${ qi (my
variable) }.


>
> Also, as far as I know, the only variables names that do not have to be
> proper identifiers are package variables (because of the way packages
> work), no? Which means this wouldn't work for lexical variables.
>

As you mentioned packages: consistent implementation will (in my eyes) may
lead to change of treating
namespaces in Perl core. For example:
Foo::qi (bar::baz) is for me different package than Foo::qi (bar)::baz

I'm already experimenting with usage of pad to replace GV
(based on my older PR replacing sigils in pad with symbol identifier, with
new symbol identifier "namespace")


>
> Also, would qi interpolate?
>
> --
>  Philippe Bruhat (BooK)
>
>  People are all unique- but some are more unique than others.
>                                     (Moral from Groo The Wanderer #22
> (Epic))
>