Support users embedding fql snippets in objects and arrays

[ptpaterson]

Ticket(s): FE-3501

resolves #259

Problem

Query instances (i.e. results from the fql function) cannot be encoded when embedded in objects or arrays.

The QueryValue type is incorrectly used for types over the wire.

Solution

The database now accepts a query interpolation fragments for array and object, where the existing fql fragment can now be nested under the array and object fragments. Add new encodeInterpolation method to encode with the new fragments.

Query arguments do not use interpolation, so they still need the @object tag. Encoding arguments still uses the existing encode method.

Update types to better distinguish between

• Arguments to be encoded
  • allow Query within objects and arrays
  • QueryValue should remain a valid QueryArgument to ensure round-trip-ability
• Encoded values sent over the wire
  • Updated fragment types and remove QueryValue from fragments
• Decoded values
  • Specify QueryValue type only where we decode values from Fauna.

Result

Queries can be encoded in QueryInterpolations.

Queries can be embedded within JS arrays and objects.

Type changes are corrections to the internal implementation. No user-facing changes were made to types (except QueryArgument)

Out of scope

n/a

Testing

Updated existing tests and added new ones for embedding queries in arrays and objects.

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[pnwpedro]

One small fix to a comment please. Looks good otherwise

[pnwpedro]

I think we should revisit the type/class organization at some point. We're increasing the strength of our circular dependencies. wire-protocol and tagged-type both now have more dependencies in query-builder, and query-builder already has dependencies in each.

We don't need to address in this PR

[ptpaterson]

I had some reorganization of types staged but I put everything back. I was worried about

1. breaking import paths, and
2. complicating the diff for this PR.

On further consideration of no. 1, that doesn't actually matter because we bundle the driver and only expose the exports from src/index and that's it.

So I guess my point is: agreed! We should come back to this.

[pnwpedro]

I don't think this comment is correct

[ptpaterson]

Thanks for catching the copy-pasta! I've updated in 7bb285a. How's that?