Top level arrays and fragment caching

[mbrandonw]

I've added a method to Jbuilder so that we can use the fragment caching pattern from ActionView (see issue #22), but along the way I also had to make a change to array!.

My change to array! allows one to simply set the array directly, similar to how set! allows one to set a key/value pair directly:

json.array! [1,2,3]  # => "[1,2,3]"

This change is pretty useful by itself. But, with this I created a method cache! that has the same signature as the helper cache in ActionView and allows the following pattern:

json.cache! ['v1', @person], :expires_in => 10.minutes do |json|
  json.extract! @person, :name, :age
end

There is still some work to be done (write tests, update readme, and right now I'm just using Rails.cache for the cache store so i guess that needs to be changed), but is this something people are interested in pulling?

[maletor]

Can't be sure of exact reproduction steps or if valid bug, but rendering a partial within a partial that calls Paperclip::Attachment#expiring_url throws an error in dalli.

Marshalling error for key 'jbuilder/0.303715376742858': singleton can't be dumped
You are trying to cache a Ruby object which cannot be serialized to memcached.
/Users/maletor/.rvm/gems/ree-1.8.7-2012.02/gems/dalli-1.1.5/lib/dalli/server.rb:266:in `dump'

[rgarver]

Is there are reason not to bypass the double serialization from the cache to jbuilder and then again to output? Can you drop the raw string in a wrapper class that has a to_json that returns the string unaltered and attach it to the json at that point?

I haven't had a chance to test this approach out so I'd genuinely like to know if this works.

[mbrandonw]

Good point. That would require some deep changes to jbuilder. I don't have the time to explore that, and I doubt the performance gains would be worth it. I've been using this caching for a few weeks now and pretty happy with it.

[rgarver]

Got it. I may take a stab at it sometime this week.

[eviltrout]

I'm very interested in this patch! Let me know if I can help out in any way (tests, documentation, etc.)

[maletor]

This should write to Rails.logger like it does in ERB.

Write fragment 'blah'
Read fragment 'blah'

[alanpca]

@eviltrout sounds like tests and documentation is necessary. I just stumbled upon this today and am pretty interested in some of the problems outlined above.

[maletor]

#35 (comment)

Pretty sure this was a result of #34.

[maletor]

@alanpca is right. Some tests, documentation and probably an implementation of the Read/Write fragment log and this will be looking really good.

@rgarver did you ever get a chance?

[rgarver]

@maletor I did make some progress on this, but didn't have enough time to really dig in to tests and docs. I just polished a corner case for my purposes this evening. My version lives here:

https://github.com/rgarver/jbuilder/tree/feature/add-caching

[maletor]

@rgarver looks really good. Might need a test for trying to de/serialize something with a Proc in it. What do you think?

[dhh]

I like this. Please update against master and we can merge.

[rgarver]

I'm in the middle of a long-distance move so sorry for the delayed response. I'll get my branch updated to master as soon as I can post a new pull request.

[maletor]

@rgarver it looks like it's a fast forward for your branch.

[rgarver]

See pull request #51 has my branch merged up with the latest master.