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.
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 tickets with “polymorphic” in title:
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.
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?
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?
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.
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
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?
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.
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.
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.
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.
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.
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.
