fix: skip include logic handling multiple paths to same node

[boopathi]

• fix misbehaving spreading fragments #166

TL;DR

To use the new behaviour, use the new CompilerOption -

compileQuery(
  schema, ast, operationName,
  {
    useExperimentalPathBasedSkipInclude: true
  }
)

Description

When there are multiple paths to the same field Node (fragments), there can be multiple skip/include requirements for the same node. This breaks our assumption in the existing implementation where we compute the skip/include for a particular field node and use the same for all paths.

For example,

query {
  paths {
    one {
      ...frag @skip(if: true)
    }
    two {
      ...frag @skip(if: false)
    }
  }
}
fragment frag on Item {
  target
}

Here, in this query, the field target has two paths - paths.one.target, and paths.two.target. But, in the current implementation we compute __internalShouldInclude for a particular field node - so this gets reused for all paths to the field Node - so is the reason for the bug.

This PR fixes it by remodelling the internal representation. This PR uses a path specific skip/include logic and while generating code, it uses the path specific logic. We represent this as an object - __internalShouldIncludePaths[path.in.dot.notation] = SkipIncludeCode for that path.

[boopathi]

👍

[ayruslore]

👍

[Joneser]

Do you think it would be possible to merge this in order to unblock #177 ?

[boopathi]

👍

[oporkka]

👍