Some notes I have made after reading the spec.

It's not. Well, not really - its an internal property of the type that's exposed through introspection - much in the same way that deprecation is.

We might want to consider doing this - that is have a property on GraphQLInputObjectType like we do for deprecation. Looking at the JS code I am pretty sure that what they do as well

OneOf Input Objects are a special variant of Input Objects where the type system
asserts that exactly one of the fields must be set and non-null, all others
being omitted.

One option is to have GraphQLInputObjectType here as a new type. But...... it feels like to much of a stretch

In schema introspection, the __Type.isOneOf field will return {true} for OneOf
Input Objects, and {false} for all other Input Objects.

This backs up the idea that is a variant of GraphQLInputObjectType and not a new class.

If the value of the single entry in the coerced unordered map is {null} an
error must be thrown.

This is incongruous in some regards - that they are all nullable but at runtime they are not. But it's the nature of this situation - we must make sure we cater for this at validation times and perhaps at coercing time.

A design question is - should we co-erce to a null value say and then allow validation to catch this bad input data? If we do this then paths like ENF that coerce but DONT run validation will not catch this situation. This is something to consider when testing coverage.

The graphql-js impl does it at coercing time - not at query valdiation tiome

https://github.com/graphql/graphql-js/pull/3513/files#diff-7a6a8382bb37b3db453772d9e1b3fdbcc04725efe79157f34ca65b3a21478841R145

If the Input Object is a OneOf Input Object then:
1. The type of the input field must be nullable.
2. The input field must not have a default value.

default values is something to watch out for here in testing.

isOneOf must return {true} when representing a OneOf Input Object, {false}
otherwise.

The Introspection code needs to be updated to handle this. This is yet more conformation that oneOf is a propery of GraphQLInputObjectType and not a new type.

IntrospectionQueryBuilder is going to need work - should we include the new directive by default? I think so BUT should we allow an option to turn it off. Will there be clients who choke in the new attr in __Type ??

IntrospectionResultToSchema needs work or at least tests to show it round tripping

Implement @oneOf directive #3266

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions