What's the use case of per-call JSON? I'm not aware of considerations that would vary within the same app...

we are using couch_potato and tire in our application. couch_potato uses JSON.create_id = 'ruby_class' to store the type of documents in the couch. tire on the other hand uses the ruby_class attribute in an incompatibile way to store index metadata. so currently we are swapping the json engine for specific calls like this:

    def state
      # WORKAROUND (ps) json-gem in combination with JSON.create_id = 'ruby_class' fails in here
      engine = MultiJson.engine
      MultiJson.engine = :yajl
      @response = Tire::Configuration.client.get("#{Tire::Configuration.url}/_cluster/state")
      MultiJson.decode(@response.body)
    ensure
      MultiJson.engine = engine
    end

i know that we could also use the Yajl::Parser directly, but i thought it would be neat to do it on a per call basis.

That makes sense, and an :engine option seems to be the right way to do it. Can you update your fork to be compatible with the latest HEAD?

hope i resolved the conflicts properly :]

@mbleigh do you still want to pull this?

Yes I do, can you fix it to latest once more? Sorry for the delay.

started from scratch on latest master