EPIC: Directive support

[vektah]

Here to track the status of directives

:warning: these features are subject to change - if you are using any of these features make sure you have tests on your directives that actually run queries against a graph. :warning:

Execution directives

location	implemented	obj is
QUERY	:heavy_check_mark: https://github.com/99designs/gqlgen/pull/744	*ast.OperationDefinition
MUTATION	:heavy_check_mark: https://github.com/99designs/gqlgen/pull/744	*ast.OperationDefinition
SUBSCRIPTION	:heavy_check_mark: https://github.com/99designs/gqlgen/pull/744	*ast.OperationDefinition
FIELD	:heavy_check_mark: https://github.com/99designs/gqlgen/pull/756	parent type
FRAGMENT_DEFINITION	:no_entry:
FRAGMENT_SPREAD	:no_entry:
INLINE_FRAGMENT	:no_entry:

Schema directives

location	implemented	obj is
SCHEMA	:no_entry:
SCALAR	:no_entry:
OBJECT	:heavy_check_mark: https://github.com/99designs/gqlgen/pull/1117	parent type
FIELD_DEFINITION	:heavy_check_mark:	parent type
ARGUMENT_DEFINITION	:heavy_check_mark: https://github.com/99designs/gqlgen/pull/460	argument bag
INTERFACE	:no_entry:
UNION	:no_entry:
ENUM	:no_entry:
ENUM_VALUE	:no_entry:
INPUT_OBJECT	:no_entry:
INPUT_FIELD_DEFINITION	:heavy_check_mark:	parent input

Schema field directives on subscriptions

obj is always nil for any root (Query, Mutation, Subscription) next() returns a channel of the field type

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[dbadura]

Hello, what happen with directive on ARGUMENT_DEFINITION?

[dtrckd]

Hi, I've commented on a potential problem with OBJECT directive, here.

[logrusorgru]

Hi, there.

There are a few questions about directives. Commonly most about INPUT_FIELD_DEFINITION. It is marked with a check mark and its input object is a parent object.

1. The gqlgen generator doesn't allow directives with INPUT_FIELD_DEFINITION on input fields. But allows them on the input fields with FIELD_DEFINITION.
2. Object received in a directive handler is map[string]interface{}, not a Go object. Is it by design? Why not Go object?
3. The following directive produce invalid server that panics just before the directive called

directive @allowValueType (field: String!, allow: [ValueTypeAllow!]) on
	INPUT_FIELD_DEFINITION |
	FIELD_DEFINITION # <- have to add

enum ValueType {
	ONE
	TWO
}

input ValueTypeAllow {
	valueType: ValueType!
	allow:     Boolean!
}

input InputValue {
	name:      String!
	valueType: ValueType! @allowValueType (field: "valueType", allow: [
		ValueTypeAllow{valueType: ONE, allow: true},
		ValueTypeAllow{valueType: TWO, allow: true}
	])
}

However, it's okay with the 3rd point if the directive receives strings for example

directive @allowValueType (field: String!, allow: [String!]) on
	INPUT_FIELD_DEFINITION |
	FIELD_DEFINITION # <- have to add

input InputValue {
	name:      String!
	valueType: ValueType! @allowValueType (field: "valueType", allow: [
		"ONE:true",
		"TWO:true"
	])
}

There is repository that represents all of them https://github.com/logrusorgru/gqlif . It can be cloned and executed with go run main.go. Then, trying to execute

mutation a($name: String!, $value: InputValue!) {
    set (name: $name, value: $value) {name valueType}
}

with

{"name": "pew-pew", "value": {"name": "one", "valueType": "ONE"}}

You will got the panic.

Best regards.

[dangolbeeker]

Have you found a patch yet?