Use functions for custom mutations and queries

[Secretmapper]

The current customMutations and customQueries are powerful, but the types aren't exposed so it is impossible to use the generated graphQLTypes on customQueries/Mutations.

example:

const customMutations = {
  addMaterialChapter: mutationWithClientMutationId({
    name: 'addMaterialChapter',
    inputFields: {
      id: { type: new GraphQLNonNull(GraphQLID) },
      title: { type: GraphQLString },
      content: { type: GraphQLString },
    },
    outputFields: {
      changedMaterial: { type: ??? }                                // no type
    },
    mutateAndGetPayload: async ({ id, title }) => {
      // resolver
    }
  })
}

export default getSchema([Material], {hooks, customMutations})

is not possible since the changedMaterial doesn't have the necessary type, since the type will only be exposed once getSchema is called and by then it'd already be too late.

What I propose is to pass in a function for both customMutations and Queries, so that the types can be passed dynamically:

const customMutations = (types) => ({
  addMaterialChapter: mutationWithClientMutationId({
    name: 'addMaterialChapter',
    inputFields: {
      //resolve
    },
    outputFields: {
      changedMaterial: { type: types.Material }                     // type obtained from types passed
    },
    mutateAndGetPayload: async ({ id, title }) => {
      //resolve
    }
  })
})

I realize this is a breaking change, (and I'm not even sure if the above is actually impossible without my PR), so wanted to see what you think @tothandras . If you think the approach is good I'll write tests for it.

[Secretmapper]

@tothandras Ah, that's even better, thanks!

I'm about to hit the sack but I'll submit another commit tomorrow with tests.

[tothandras]

@Secretmapper I had a few minutes :) hope you don't mind! https://github.com/tothandras/graffiti-mongoose/commit/0cb244fa1f877dde398e9f48b7ad70b64075d2ec Can you try it out?

[Secretmapper]

@tothandras Thanks!

Another thing I forgot, we also needed to expose the processId function. This way the user can write his/her own resolver using the ids generated by graffiti-mongoose. I've also added in the function checks!

[tothandras]

Sure, if you need that, but we are about to process every field, because it seems we would need that for #112! :) Also, take a look at my commit, we need to convert the object types to input types.