Try using a field extension for 1.9 interpreter

[rmosolgo]

This doesn't quite work yet, and I think it's because of the changes in rmosolgo/graphql-ruby#1984. (Previously, there were cases when execution would proceed eagerly down the tree, but in that PR, it was changed to always resolve each level of depth before proceeding to further depths.)

I think this means that the interpreter has to hardcode handling for mutation root fields, which isn't too surprising. I'll take a look. Fixed by rmosolgo/graphql-ruby#2097

This adds a CI build for 1.9.0.pre4 and and implementation that uses a FieldExtension. I think it will work pretty well, the only problem will be if people do:

# DONT DO THIS, IT WOULD BE BAD 👻
use GraphQL::Batch
mutation(Types::Mutation)

It will still work for now because plugins are applied after other configurations. But in the future, there will be no "build" step, and instead, the schema will be constructed just as configured in the source, so it might start breaking.

We could override self.mutation(t) to raise an error, to avoid that configuration error. Would you like to?

(Also, I realize you may prefer a different path, depending on the conversation over at #93, but I thought I'd code one up just to see how it played out.)

[swalkinshaw]

I'm in favour of is even if it ends up being a shorter term solution 👍It would be good to have a version that supports 1.9 before we consider a larger re-org or unification?

[rmosolgo]

Same here, I think it'd be great to release a graphql-batch version that will support 1.9.0 so that people can update OK. (I've been doing the same for graphql-client.) Anything else I can do on this one, or does it look OK?

[rmosolgo]

👍 Thanks for taking a look! I applied that suggestion and cleaned up the now-needless require at the end of the file.