Required input struct generated as pointer

[Junaid-Ahm]

What happened?

For a required object field in another schema object (happens with both type or input), it declares the required object field as pointer to the struct - see field Details declared as *UserDetailsInput in declaration of UpdateUserInput. It should have been declared as UserDetailsInput instead.

The models_gen.go contains

type UpdateUserInput struct {
	ID      string            `json:"id"`
	Name    string            `json:"name"`
	Details *UserDetailsInput `json:"details"`
}

type UserDetailsInput struct {
	X string  `json:"x"`
	Y string  `json:"y"`
	Z *string `json:"z"`
}

What did you expect?

If the object field is required, it should be declared without pointer. In other words, it should only be declared as pointer if the object field is optional.

The models_gen.go should contain

type UpdateUserInput struct {
	ID      string           `json:"id"`
	Name    string           `json:"name"`
	Details UserDetailsInput `json:"details"`
}

Minimal graphql.schema and models to reproduce

type Mutation {
    updateUser(input: UpdateUserInput!): User!
}

input UserDetailsInput {
    x: String!
    y: String!
    z: String
}

input UpdateUserInput {
    id: ID!
    name: String!
    details: UserDetailsInput!
}

versions

• go run github.com/99designs/gqlgen version: v0.17.4
• go version: 1.18

[frederikhors]

Already discussed hundreds of times. Search through issues and discussions.

[ianling]

@frederikhors I searched and could not find anything related to this specific issue. Can you link directly to a past discussion of this? I am running into the same behavior and I don't understand why it is designed this way.

[ianling]

Found it! #701

To summarize, it seems this was due to the need to check for cyclical relationships between structs. It was quicker and easier to simply turn all structs within structs into pointers. The more complicated approach would have been to try to detect cyclical relationships by traversing the struct's fields, but it's understandable that the "always pointers" approach was taken instead, as it avoids any potential bugs in the struct traversal code.

[ianling]

I add a config option for this in #2174