Polymorphic URL helpers documentation and fixes

[Mislav Marohnić]

Yesterday I answered a question regarding polymorphic URL helpers on Core ML and noticed that the module has no documentation. I've documented it and rewritten unit tests using Mocha. I also optimized some of the code slightly (nothing major, though). The patch also includes two fixes by Geoff Buesing, who has done awesome work in this area in the past.

http://dev.rubyonrails.org/ticket/10883

Check the documentation, please yell if anything is wrong:
http://dev.rubyonrails.org/attachment/ticket/10883/PolymorphicRoutes.html?format=raw

What's still broken? (I decided not to fix everything in one go.)

1. Hash argument: polymorphic_url(:id => @article). Test is commented out (like before).
   
2. Should this work: polymorphic_url(@article, :format => :pdf) ? Currently it doesn't. I've included a failing tests, it's commented out.

Open ActionPack tickets with "polymorphic" in title:
http://dev.rubyonrails.org/query?status=new&status=assigned&status=reopened&component=ActionPack&summary=%7Epolymorphic&order=priority

This one needs a discussion of its own: http://dev.rubyonrails.org/ticket/9621

Thanks.
- Mislav

[Michael Koziarski]

On Jan 22, 2008 9:28 AM, Mislav Marohnić <mislav....@gmail.com> wrote:
> Yesterday I answered a question regarding polymorphic URL helpers on Core ML
> and noticed that the module has no documentation. I've documented it and
> rewritten unit tests using Mocha. I also optimized some of the code slightly
> (nothing major, though). The patch also includes two fixes by Geoff Buesing,
> who has done awesome work in this area in the past.
>
> http://dev.rubyonrails.org/ticket/10883
>
> Check the documentation, please yell if anything is wrong:
>
> http://dev.rubyonrails.org/attachment/ticket/10883/PolymorphicRoutes.html?format=raw
>
> What's still broken? (I decided not to fix everything in one go.)
> Hash argument: polymorphic_url(:id => @article). Test is commented out (like
> before).
> Should this work: polymorphic_url(@article, :format => :pdf) ? Currently it
> doesn't. I've included a failing tests, it's commented out.Open ActionPack

I don't think it should work with these 'advanced' hash arguments. do
they do the right thing when passing through to the optimised named
route generators? Do we want to support them going forwards?

> This one needs a discussion of its own:
> http://dev.rubyonrails.org/ticket/9621

I think that's just an abuse of inheritance and I'm not sure we could
do anything performant and reasonable to support that case, and the
case where AssetsController is meant to handle all the different
subclasses?

--
Cheers

Koz

[Mislav Marohnić]

2008/1/22 Michael Koziarski <mic...@koziarski.com>:

> What's still broken? (I decided not to fix everything in one go.)
> Hash argument: polymorphic_url(:id => @article). Test is commented out (like
> before).
> Should this work: polymorphic_url(@article, :format => :pdf) ? Currently it
> doesn't. I've included a failing tests, it's commented out.Open ActionPack

I don't think it should work with these 'advanced' hash arguments.  do
they do the right thing when passing through to the optimised named
route generators?  Do we want to support them going forwards?

In my experience, formatted_foo_path(@foo, :pdf) and formatted_foo_path(@foo, :format => :pdf) are identical. Still, my patch only adds support for formatted_polymorphic_path([@foo, :pdf]), and I've put the hash argument up for discussion.

> http://dev.rubyonrails.org/ticket/9621

I think that's just an abuse of inheritance and I'm not sure we could
do anything performant and reasonable to support that case, and the
case where AssetsController is meant to handle all the different
subclasses?

I agree with such thinking; I'd only like these issues regarding polymorphic helpers resolved. Make the API concrete, reject features we don't want etc. Thanks Koz

P.S. somehow you avoided to comment on my patch :D

[Trevor Squires]

On 1/21/08, Michael Koziarski <mic...@koziarski.com> wrote:

On Jan 22, 2008 9:28 AM, Mislav Marohnić <mislav....@gmail.com> wrote:
>
> Should this work: polymorphic_url(@article, :format => :pdf) ? Currently it
> doesn't. I've included a failing tests, it's commented out.Open ActionPack

I don't think it should work with these 'advanced' hash arguments.  do
they do the right thing when passing through to the optimised named
route generators?  Do we want to support them going forwards?

Koz,

I've often wondered why (apart from specific implementation reasons) we need a formatted_xxx helper to cater for formats at all.

Bearing in mind that the generated helper code (including optimization stuff) could just look at the size of the hash and whether it's only member is :format, is there a reason the generated helpers can't simply distinguish between these signatures?

album_path(an_album)

album_path(an_album, :format => :xml)

albums_path()

albums_path(:format => :xml)

Somehow that seems easier on the eyes than:

album_path(an_album)

formatted_album_path(an_album, :xml)
albums_path()

formatted_albums_path(:xml)

 

I think it's because formatted_ seems to just add noise to the method name.  The format isn't (to me) the most important thing, it's more of an afterthought.  Keeping it's specification exclusively in the params mirrors that imho.

Regards,

Trevor
--
--
Trevor Squires
http://somethinglearned.com

[Chris Cruft]

I'm frequently in need of generating a formatted polymorphic url.
Currently, it's done ugly because the formatted_polymorphic_url code
is broken (see ticket 8782 and Mislav's #2 above). I would like to
see that helper working-but if the job can be done with something
simpler, I'm all for it. Trevor's suggestion is intriguing -leave the
"formatted" prefix off and use the signature to determine if
formatting is required. That would be ideal.

On Jan 21, 8:35 pm, "Trevor Squires" <tre...@protocool.com> wrote:
> On 1/21/08, Michael Koziarski <mich...@koziarski.com> wrote:
>
> > On Jan 22, 2008 9:28 AM, Mislav Marohnić <mislav.maroh...@gmail.com>

[Michael Koziarski]

> I've often wondered why (apart from specific implementation reasons) we need
> a formatted_xxx helper to cater for formats at all.

I don't believe there's a reason beyond the (considerable)
implementation specific reasons :). Changing that with the current
routing implementation would probably be a little tricky without
making everyone generate urls like:

/albums/1.html

> I think it's because formatted_ seems to just add noise to the method name.
> The format isn't (to me) the most important thing, it's more of an
> afterthought. Keeping it's specification exclusively in the params mirrors
> that imho.

While it's a little odd, I don't see an easy migration path that's
suitable for a point release.

--
Cheers

Koz

[Mislav Marohnić]

OK, so we're keeping formatted_foo helpers for now. But what of my patch?
http://dev.rubyonrails.org/ticket/10883

It fixes issues. It adds tons of documentation. It provides more test coverage. Is there anything wrong with it? Anything that someone would like to put up for discussion, or to have me change it? Thanks.

- Mislav

[Michael Koziarski]

> OK, so we're keeping formatted_foo helpers for now. But what of my patch?
>
> http://dev.rubyonrails.org/ticket/10883
>
> It fixes issues. It adds tons of documentation. It provides more test
> coverage. Is there anything wrong with it? Anything that someone would like
> to put up for discussion, or to have me change it? Thanks.

I'm happy with the documentation but not sure I buy the need for
#8720? The intention of those helpers is that they automatically
figure out the right url to go to... #8782 seems fine and the
documentation seems awesome.

--
Cheers

Koz

[Mislav Marohnić]

On Jan 25, 2008 9:22 PM, Michael Koziarski <mic...@koziarski.com> wrote:

I'm happy with the documentation but not sure I buy the need for
#8720?

Why match the API of polymorphic_path with polymorphic_url? Because there is no need for these two to have different signatures, especially since generated helpers for resources (foo_url vs. foo_path) have the same parameters. I think consistency is key here, and since polymorphic_path should be more used in general, it should take the parameters option. Otherwise you would have to use polymophic_url with :routing_type => :path.