ansaurus

Question

What is the best RESTful way to pass string collections as parameters?

Answer 1

A:

Well you could always POST them.

Byron Whitlock 2009-09-18 21:01:53

REST uses POST for a special purpose (creating a new resource), so if he is trying to GET a resource, he wouldn't use POST...

SingleShot 2009-09-18 21:03:05

He actually doesn't say what he is doing though :-)

SingleShot 2009-09-18 21:04:23

@SingleShot good point, well made :-) Basically I am trying to retrieve a collection of resources filtered by the parameters. They are actually locations - so filtered by tag, origin point, distance, etc. So yes, it should be a GET. I want to be able to cache it tho.

frankodwyer 2009-09-18 21:20:11

Answer 2

+2 A:

Let's say you are trying to GET a resource, perhaps a "List of Questions" and you want to filter the list based on a set of "Tags" (What a great idea!).

Your URL should be the path to the resource. Perhaps /questions would be a good URL. I know you don't want to use URL parameters for your tags, but on a GET, that's what they are for - filtering or customizing the view of a resource. So to filter for a few subjects of interest rather than getting all questions, your URL might look like:

/questions?tag=ruby-on-rails&tag=restful&tag=url-routing

I believe that is the RESTful way. Alternatives may be palatable, but I think they would be violating the "purist" form of REST.

SingleShot 2009-09-18 21:09:45

my problem with the query string here is that rails doesn't seem to take any account of it when caching to the page cache. In other words, /questions?tag=foo and /questions?tag=bar will go to the same location in the page cache: /questions

frankodwyer 2009-09-18 21:24:29

I will have to deal with this at some point so went off and dug a little. Searching reveals a common theme: use a rails plug-in that allows you to customize your cache keys (allowing params to be part of the key). The question is, does doing so violate REST or is the caching problem rails-specific? Here's the link to the plugin: http://agilewebdevelopment.com/plugins/action_cache

SingleShot 2009-09-18 21:46:49

for action caching, I've been able to get that going using vanilla rails like so: caches_action :controller, :expires_in=>1.hour, :cache_path => Proc.new { |c| "control/#{c.params[:bar]}/#{c.params[:foo]}".gsub(/ /,'') }This fudges the parameters into the cache path and makes sure that distinct requests are cached separately. However I'm not sure what version of rails this requires - I'm using 2.3.2 . Plus, I don't know if similar can be done in relation to the page cache.

frankodwyer 2009-09-18 22:44:07

Answer 3

+1 A:

Use the query string. It's there for exactly this purpose: passing encoded arguments to a resource.

Your version using path elements obscures the fact that /controller/tagged is the actual resource being queried, with the tags as parameters to the request.

To see why the path approach isn't RESTful, consider the following two URLs:

http://example.com/controller/tagged/foo/bar

http://example.com/controller/tagged/bar/foo

By the description you gave, these should refer to the same virtual collection of tagged items. However, in a RESTful system, URLs refer unambiguously to one and only one resource. Your path-based addressing associates many URLs with a single result.

rcoder 2009-09-18 21:11:03

my problem with the query string (at least the last time I looked at using it), is that it is ignored for caching. I need the responses to go to the page cache.

frankodwyer 2009-09-18 21:15:43

btw +1 for the observation that two urls map to the same resource collection - hadn't thought of that. Not sure it is a problem in practice but it would be nice to avoid it.

frankodwyer 2009-09-18 21:26:13

In that case, you may need to manage the cache yourself in your controller, instead of caching at the page level. Try something like `cache(:key => "tagged/#{params[:tag].join(',')}") do ...` in your controller action, capturing the result of the query and template rendering necessary to return the list of tagged resources.

rcoder 2009-09-18 21:27:52

Answer 4

+1 A:

Sam Ruby and Leonard Richardson recommend in their book Restful Web Services to separate non-hierarchical data by commas or semicolons (and query variables for algorithmic resources).

jug 2009-09-18 21:21:50

can you elaborate on 'and query variables for algorithmic resources'? Didn't get that - any examples?

frankodwyer 2009-09-18 21:27:06

Searches for example should be better addressed as www.google.com/search?q=jellyfish than as www.google.com/search/jellyfish

jug 2009-09-18 21:45:22

ansaurus

tags:

views:

answers:

What is the best RESTful way to pass string collections as parameters?

related questions