From @masak

Due to the structurally recursive nature of C<.perl>, the fact that
Rakudo r34997 doesn't give newly defined classes a default C<.perl>
method upon definition is very keenly felt e.g. when doing C<.perl> on
a list of instances of the new class.

$ perl6 -e '(class {}).new.perl'Method 'perl' not found for invocant of class ''
[...]
$ perl6 -e 'class A {}; A.new.perl'
Method 'perl' not found for invocant of class 'A'
[...]

Two questions naturally arise​: the first is why this method isn't
inherited from Object? The second is what such a default C<.perl>
should output.

I have a vague feeling that similar tickets have been submitted
before. But this question needs to be put forward until the spec is
clarified (or whatever is the blocker) -- missing methods is one of
the few things that can ruin a scripting session.

From @pmichaud

On Tue, Jan 06, 2009 at 02​:50​:07AM -0800, Carl Mäsak wrote​:

Due to the structurally recursive nature of C<.perl>, the fact that
Rakudo r34997 doesn't give newly defined classes a default C<.perl>
method upon definition is very keenly felt e.g. when doing C<.perl> on
a list of instances of the new class.

$ perl6 -e '(class {}).new.perl'Method 'perl' not found for invocant of class ''
[...]
$ perl6 -e 'class A {}; A.new.perl'
Method 'perl' not found for invocant of class 'A'
[...]

Two questions naturally arise​: the first is why this method isn't
inherited from Object? The second is what such a default C<.perl>
should output.

Rakudo doesn't implement a .perl in Object because of the second
question -- we don't know what that default C<.perl> should output.

But if you'd like to add a temporary ".perl" method to Object
that does something more reasonable (however you define it),
that's certainly ok.

In the meantime I'm marking this ticket as a [Spec] ticket.

Pm

The RT System itself - Status changed from 'new' to 'open'

From @masak

Pm (>)​:

But if you'd like to add a temporary ".perl" method to Object
that does something more reasonable (however you define it),
that's certainly ok.

<jnthn> masak​: What about a TheClass.new(...) which sets up attributes
to their current values?
<masak> jnthn​: you know, that's actually not half bad. :)
<masak> (why didn't I think of that?)

From @moritz

On Tue Jan 06 07​:47​:30 2009, masak wrote​:

Pm (>)​:

But if you'd like to add a temporary ".perl" method to Object
that does something more reasonable (however you define it),
that's certainly ok.

<jnthn> masak​: What about a TheClass.new(...) which sets up attributes
to their current values?
<masak> jnthn​: you know, that's actually not half bad. :)
<masak> (why didn't I think of that?)

To "steal" some prior art from pugs​:

pugs> class A { has $.b; has @​.c; }; A.new(​:b<3> :c<4 5>).perl
"A.new((\"b\" => \"3\"), (\"c\" => (\"4\", \"5\")))"

For the case of cyclic structures, we could emit things like​:

do {
  my $a = A.new();
  # this only works is $.b is declared as 'is rw'
  $a.b = $a;
}

I don't think it'll work without some kind of special syntax (that
allows setting of private attributes from the outside later on).

But for now the non-cyclic case would be *very* helpful if implemented
somehow.

Cheers,
Moritz

From @bbkr

[15​:30] <bbkr> is this ticket still valid?
http://rt.perl.org/rt3/Ticket/Display.html?id=62002 looks like it was
created before Mu object was added to spec.
[15​:31] <moritz_> bbkr​: well, it's now s​:g/Object/Mu/
[15​:31] <moritz_> bbkr​: but all in all we still need a good default
.perl method
[15​:34] <bbkr> moritz_​: thanks. but since both code samples work then
the main problem in ticked - .perl method not available in user defined
class - is solved (despite output it generates)
[15​:34] <bbkr> ticket*
[15​:34] <moritz_> bbkr​: if you say so... then somebody will open a
ticket about Any.perl being useless
[15​:34] <moritz_> I don't mind either way

leaving unchanged then

From @pmichaud

Test added to S02-names-vars/perl.t, closing ticket.

Pm

@pmichaud - Status changed from 'open' to 'resolved'