Directive precedence should be explicit, not implicit

[savaki]

The draft states that if both the @Skip and @include directives are present, the @Skip directive should take precede. This strikes me as confusing as it opens the door to others creating directives with implicit precedence.

Wouldn't it be simpler and less confusing to say precedence is left to write in the order it was written?

If fragments are involved, then we could linearize the dependencies by saying depth first. Then there are no ambiguities.

[leebyron]

Thanks for the recommendation.

The goal is to be as simple as possible, so I appreciate the argument. Simplicity also led to the current behavior, though primarily simplicity of implementation. Adding the additional step of order-dependence makes this more complicated.

There is also a desire for the directives to be un-ordered so they can be safely represented as a Set rather than a List internally, this can be a simplifying force for implementations and removes a potential area for mistakes by accidental re-ordering.

I have another suggestion that I'm interested in getting your feedback on, hopefully it solves your concerns:

As currently specced:

1. If selection provides the directive @skip, let skipDirective be that directive.
   1. If skipDirective‘s if argument is true, continue with the next selection in selectionSet.
2. If selection provides the directive @include, let includeDirective be that directive.
   1. If includeDirective‘s if argument is false, continue with the next selection in selectionSet.

We get this truth table if both skip and include are present:

	@skip(if: true)	@skip(if: false)	Not provided
@include(if: false)	skipped	skipped	skipped
@include(if: true)	skipped	included	included
Not provided	skipped	included	included

Based on this, I think we can reframe the relationship as written in the spec to be a formal AND. That would have an identical truth table, but would not contain any implicitly ordered relationship between the two nor would it rely on any explicit ordering (due to the commutative property).

Another nice outcome of this is that the spec could potentially be written to allow multiple @skip or @include directives, all AND'd together, so any single @skip(if: true) or @include(if: false) in the list would result in a skip. I'm not sure if this is a good idea or not yet (it could be a bag of worms to allow for multiple directives) but I thought it was interesting to bring up.

[leebyron]

Closing since this note has been added to the spec about there being no precedence in this case: http://facebook.github.io/graphql/#sec--include