Bug #68394 for MooseX-Role-ConstructorRoleApplication: Please explain differences from MooseX::Traits in Pod

Mon May 23 08:14:49 2011 bobtfish [...] bobtfish.net - Ticket created

Subject:	Please explain differences from MooseX::Traits in Pod
Date:	Mon, 23 May 2011 13:16:23 +0100
To:	bug-MooseX-Role-ConstructorRoleApplication [...] rt.cpan.org
From:	Tomas Doran <bobtfish [...] bobtfish.net>

Hi. From an initial look, your module appears to be a re-implementation of MooseX::Traits. Can you explain in the Pod of your module why you should use your module instead of MooseX::Traits? Cheers t0m

Mon May 23 08:25:46 2011 FRASE [...] cpan.org - Correspondence added

On Mon May 23 08:14:49 2011, BOBTFISH wrote: Show quoted text

> Hi. > > From an initial look, your module appears to be a re-implementation > of MooseX::Traits. Can you explain in the Pod of your module why you > should use your module instead of MooseX::Traits? > > Cheers > t0m >

Sheesh... how the heck did I remain oblivious to MooseX::Traits all this time? In light of this, MooseX::Role::ConstructorRoleApplication is going to unceremoniously depart from this world. Thank you for enlightening me.

Mon May 23 08:25:50 2011 The RT System itself - Status changed from 'new' to 'open'

Tue May 24 01:03:47 2011 FRASE [...] cpan.org - Correspondence added

On Mon May 23 08:14:49 2011, BOBTFISH wrote: Show quoted text

> Hi. > > From an initial look, your module appears to be a re-implementation > of MooseX::Traits. Can you explain in the Pod of your module why you > should use your module instead of MooseX::Traits? > > Cheers > t0m >

After a closer look, it seems the existance of CRA is warranted. I have updated the Pod to explain differences and use cases in 0.03.

Tue May 24 01:03:48 2011 FRASE [...] cpan.org - Status changed from 'open' to 'resolved'

Tue May 24 03:27:16 2011 bobtfish [...] bobtfish.net - Correspondence added

Thanks for the quick response! Show quoted text

> After a closer look, it seems the existance of CRA is warranted. I have > updated the Pod to explain differences and use cases in 0.03.

I'm sorry to be a pain in the ass, but this really doesn't explain it to me. You say: Use C<MooseX::Role::ConstructorRoleApplication> if you're a fan of L<Test::More>'s C<new_ok> or for other reasons require a single call to C<new> rather than a chain of calls (as would be necessary with C<MooseX::Traits>). Otherwise, use C<MooseX::Traits>. But MooseX::Traits provides a ->new_with_traits method which internally calls ->with, then ->new on the result.. Isn't that the functionality you are noting as different?

Tue May 24 03:27:17 2011 The RT System itself - Status changed from 'resolved' to 'open'

Tue May 24 07:42:16 2011 FRASE [...] cpan.org - Correspondence added

On Tue May 24 03:27:16 2011, BOBTFISH wrote: Show quoted text

> Thanks for the quick response! >

> > After a closer look, it seems the existance of CRA is warranted. I have > > updated the Pod to explain differences and use cases in 0.03.

> > I'm sorry to be a pain in the ass, but this really doesn't explain it

to me. Show quoted text

>

No pain at all. Show quoted text

> You say: > > Use C<MooseX::Role::ConstructorRoleApplication> if > you're a fan of L<Test::More>'s C<new_ok> or for other reasons require a > single call to C<new> rather than a chain of calls (as would be necessary > with C<MooseX::Traits>). Otherwise, use C<MooseX::Traits>. > > But MooseX::Traits provides a ->new_with_traits method which internally > calls ->with, then ->new on the result.. Isn't that the functionality > you are noting as different?

Good point. Further clarification is needed. The (sole) advantage, then, is that you can use it with code (other modules, e.g. Test::More) that calls ->new on a given package. I can not see any way that you could use new_ok (as an example I am familiar with; there may be others) with MooseX::Traits. Then again, I could be completely blind to an obvious solution, in which case I welcome further enlightenment.

Tue May 24 08:35:27 2011 bobtfish [...] bobtfish.net - Correspondence added

On Tue May 24 07:42:16 2011, FRASE wrote: Show quoted text

> Good point. Further clarification is needed. The (sole) advantage, > then, is that you can use it with code (other modules, e.g. Test::More) > that calls ->new on a given package. I can not see any way that you > could use new_ok (as an example I am familiar with; there may be > others) with MooseX::Traits.

Here you're correct. This is entirely a design decision from MooseX::Traits, as overloading new to do something other than pure construction was/is considered bad practice, and most of the time whatever you're doing can easily be done a different way (e.g. writing your own new_ok function that called ->new_with_traits if present) is trivial.

Wed May 25 02:40:16 2011 FRASE [...] cpan.org - Correspondence added

On Tue May 24 08:35:27 2011, BOBTFISH wrote: Show quoted text

> On Tue May 24 07:42:16 2011, FRASE wrote:

> > Good point. Further clarification is needed. The (sole) advantage, > > then, is that you can use it with code (other modules, e.g. Test::More) > > that calls ->new on a given package. I can not see any way that you > > could use new_ok (as an example I am familiar with; there may be > > others) with MooseX::Traits.

> > Here you're correct. > > This is entirely a design decision from MooseX::Traits, as overloading > new to do something other than pure construction was/is considered bad > practice, and most of the time whatever you're doing can easily be done > a different way (e.g. writing your own new_ok function that called > ->new_with_traits if present) is trivial.

Well, one ends up writing either a replacement for new_ok (or some other affected function) or a MooseX::Traits-ish role. I have updated the POD once more; if you feel further discussion or clarification is needed, let me know your suggestions.

Wed May 25 02:40:17 2011 FRASE [...] cpan.org - Status changed from 'open' to 'resolved'

Wed May 25 07:32:52 2011 hdp [...] cpan.org - Correspondence added

On Tue May 24 08:35:27 2011, BOBTFISH wrote: Show quoted text

> On Tue May 24 07:42:16 2011, FRASE wrote:

> > Good point. Further clarification is needed. The (sole) advantage, > > then, is that you can use it with code (other modules, e.g. Test::More) > > that calls ->new on a given package. I can not see any way that you > > could use new_ok (as an example I am familiar with; there may be > > others) with MooseX::Traits.

> > Here you're correct.

Not for that particular example, anyway. my $obj = new_ok($class->with_traits(...), \@args, $test_name); with_traits just returns a new class, so there's no reason you can't call new_ok on it.

Wed May 25 07:32:54 2011 The RT System itself - Status changed from 'resolved' to 'open'

Wed May 25 09:02:41 2011 FRASE [...] cpan.org - Correspondence added

On Wed May 25 07:32:52 2011, HDP wrote: Show quoted text

> On Tue May 24 08:35:27 2011, BOBTFISH wrote:

> > On Tue May 24 07:42:16 2011, FRASE wrote:

> > > Good point. Further clarification is needed. The (sole) advantage, > > > then, is that you can use it with code (other modules, e.g.

Test::More) Show quoted text

> > > that calls ->new on a given package. I can not see any way that you > > > could use new_ok (as an example I am familiar with; there may be > > > others) with MooseX::Traits.

> > > > Here you're correct.

> > Not for that particular example, anyway. > > my $obj = new_ok($class->with_traits(...), \@args, $test_name); > > with_traits just returns a new class, so there's no reason you can't > call new_ok on it.

The anonymous class returned by with_traits() does not appear in the class' @ISA, so new_ok fails (as would any other operation calling $obj->isa($result_of_with_traits).

Wed May 25 13:57:20 2011 hdp [...] cpan.org - Correspondence added

On Wed May 25 09:02:41 2011, FRASE wrote: Show quoted text

> > my $obj = new_ok($class->with_traits(...), \@args, $test_name); > > > > with_traits just returns a new class, so there's no reason you can't > > call new_ok on it.

> > The anonymous class returned by with_traits() does not appear in the > class' @ISA, so new_ok fails (as would any other operation calling > $obj->isa($result_of_with_traits).

Either we've miscommunicated or you didn't try it: { package Stuffy; use Moose::Role; } { package Stuff; use Moose; with 'MooseX::Traits'; } use Test::More; my $obj = new_ok(Stuff->with_traits('Stuffy')); done_testing; # output: ok 1 - The object isa MooseX::Traits::__ANON__::SERIAL::1 new_ok doesn't care that the anonymous class isn't in Stuff's @ISA, because new_ok doesn't know anything about Stuff, because you didn't pass it Stuff as an argument; you passed in the automatically generated class name from with_traits. with_traits does return a new anonymous class each time you call it, which may be contributing to your confusion -- that is, this second test would fail: isa_ok($obj, Stuff->with_traits('Stuffy')) I think that's arguably a bug (and there is an RT ticket for it), but it's unlikely to be a real limitation in any case. Checking isa('Stuff') and/or does('Stuffy') is the right thing to do, not checking isa($autogenerated_class). If you still have a case in mind where with_traits is insufficient, can you please show code?

Wed May 25 19:43:55 2011 FRASE [...] cpan.org - Correspondence added

On Wed May 25 13:57:20 2011, HDP wrote: Show quoted text

> On Wed May 25 09:02:41 2011, FRASE wrote:

> > > my $obj = new_ok($class->with_traits(...), \@args, $test_name); > > > > > > with_traits just returns a new class, so there's no reason you can't > > > call new_ok on it.

> > > > The anonymous class returned by with_traits() does not appear in the > > class' @ISA, so new_ok fails (as would any other operation calling > > $obj->isa($result_of_with_traits).

> > Either we've miscommunicated or you didn't try it: > > { > package Stuffy; > use Moose::Role; > } > { > package Stuff; > > use Moose; > with 'MooseX::Traits'; > } > > use Test::More; > > my $obj = new_ok(Stuff->with_traits('Stuffy')); > done_testing; > > # output: > ok 1 - The object isa MooseX::Traits::__ANON__::SERIAL::1 > > new_ok doesn't care that the anonymous class isn't in Stuff's @ISA, > because new_ok doesn't know anything about Stuff, because you didn't > pass it Stuff as an argument; you passed in the automatically generated > class name from with_traits. > > with_traits does return a new anonymous class each time you call it, > which may be contributing to your confusion -- that is, this second test > would fail: > > isa_ok($obj, Stuff->with_traits('Stuffy')) > > I think that's arguably a bug (and there is an RT ticket for it), but > it's unlikely to be a real limitation in any case. Checking > isa('Stuff') and/or does('Stuffy') is the right thing to do, not > checking isa($autogenerated_class). > > If you still have a case in mind where with_traits is insufficient, can > you please show code?

You're right; it turned out to be a design decision in our code that caused new_ok to fail. After much to-ing and fro-ing, it's back to the original conclusion: MooseX::Role::ConstructorRoleApplication is headed for the bit bucket. Thanks all for your advice.

Wed May 25 19:43:56 2011 FRASE [...] cpan.org - Status changed from 'open' to 'resolved'