Add apolloBatchIndex and apolloBatchSize fields on the context object

[n1mmy]

This is so that users can detect when clients have requested batching. This will be useful for performance monitoring (see https://github.com/apollostack/optics-agent-js/issues/47)

[glasser]

Do we want to include this even when there's no batching? ie, is there any point to distinguishing between "sent a single query" and "sent a size-1 batch"? (I don't know the answer, just though it was worth asking.)

[n1mmy]

Oh, right, I suggested that to @helfer in person and he agreed it was a good idea, then I forgot to implement it. Will add unless someone suggests a reason why it would be bad to distinguish.

[DxCx]

Few comments i want to raise:

1. Batching is being handled (Copy/Paste mostly) in 3 different files.
   i think we should add core module for batchedRunQuery, getting the common logic there.
2. I'm not sure context is the right place to add this metadata, context is an object created by the user (Talking about the server developer, not the final user) for the user. modifying it for the user (without the user explicitly asks so) might bring chaos, because imagine that this is just 1 feature added, but later on more and more people will start overloading context after you.
3. if you decided to use context anyway, i think you should add metadata key to the context, and only then append the info.
4. we can trick the schema and hook resolvers for appending metadata to info object (which is by definition created internally for the user to use) but this is abit more work and more open for flaws..

would like to hear more thoughts :)

[glasser]

I'm not sure I follow exactly what you're suggesting, but it makes me think that we should just add a single apollo or apolloContext field to the context and hang things like these numbers off of it.

[helfer]

@DxCx I agree, see also #176.

As for the key we use on the context, I think we should stick to just apolloBatchIndex if that's the only one we have for now, but if we add batchSize as well, then let's "namespace" it in a apolloContext field as @DxCx suggested.

[n1mmy]

The current PR has both context.apolloBatchIndex and context.apolloBatchSize. Happy to change to context.apolloContext.batchIndex and context.apolloContext.batchSize or something like that if that's what people prefer.

@DxCx The situation is I'm trying to find a way to distinguish which query a resolver is running in the context of, for https://github.com/apollostack/optics-agent-js. context is the only reliable way I found to get that information into resolvers. Unfortunately, this interacts poorly with apollo-server transport level batching, as that uses the same context object for each request. I agree overwriting fields of the user-provider context is kinda hacky, but I couldn't think of anything better, unfortunately. Other ideas welcome!

[helfer]

@n1mmy okay, then let's create an apolloContext object on the context.

In the medium to long term I think we should move batching into the core, which will also simplify the integrations.

[n1mmy]

Closing in favor of #182