Coverage increased (+0.2%) to 92.317% when pulling 6bd89f2 on TangoAgency:batch into d348ec8 on graphql-python:master.

Not sure why the UnboundLocalError suddenly occurred in my build on Travis. master tests probably never checked the else branch of if execution_result condition. Anyway, the issue is now fixed.

Looks great!!

Thanks for such a great PR!

First of all, thanks a lot for adding this.
A question/request though.
As I understand, this is made for a specific client implementation of batched queries...??

I'm using apollo-client which also supports query batching but it expects the results to come back the same way as "normal" queries, in a data prop.
From what I can see the response format for a batched query looks like this:

                response = {
                     'id': id,
                     'payload': response,
                     'status': status_code,
                 }

where response['data'] is overwritten with this 'new' response instead.

I did make this work by sub classing the GraphQLView and basically removed those lines but it seems a bit cumbersome since I have to copy/paste this entire function:
def get_response(self, request, data, show_graphiql=False):

and make the changes in there.

Would it be an option to make the response format more pluggable in some way...perhaps like a function that we can overwrite instead that takes the execution result and returns the response, and that would default to response['data']. Then one could sub class the GraphQLView and just overwrite that function to return a different response format if so desired.

@jole78 try setting response transform function as batch property of GraphQLView and then transforming the actual response using that function in the code you linked.

I would also suggest making the current implementation default to not break existing code.