99designs
Bug with `struct_fields_always_pointers` config
Hi there,
I've been using the lib for some time now and I'm quite sure that the struct_fields_always_pointers doesn't do anything.
I mean, I've tried setting it to true for passing all the generated structs to non-pointers fields but it didn't do anything.
Please see the following setup for reproduction.
gqlgen.yml:
# Where are all the schema files located? globs are supported eg src/**/*.graphqls
schema:
- graph/*.graphqls
# Where should the generated server code go?
exec:
package: graph
layout: single-file # Only other option is "follow-schema," ie multi-file.
# Only for single-file layout:
filename: graph/generated.go
# Only for follow-schema layout:
# dir: graph
# filename_template: "{name}.generated.go"
# Optional: Maximum number of goroutines in concurrency to use per child resolvers(default: unlimited)
# worker_limit: 1000
# Uncomment to enable federation
# federation:
# filename: graph/federation.go
# package: graph
# version: 2
# options:
# computed_requires: true
# Where should any generated models go?
model:
filename: graph/model/models_gen.go
package: model
# Optional: Pass in a path to a new gotpl template to use for generating the models
# model_template: [your/path/model.gotpl]
# Where should the resolver implementations go?
resolver:
package: graph
layout: follow-schema # Only other option is "single-file."
# Only for single-file layout:
# filename: graph/resolver.go
# Only for follow-schema layout:
dir: graph
filename_template: "{name}.resolvers.go"
# Optional: turn on to not generate template comments above resolvers
# omit_template_comment: false
# Optional: Pass in a path to a new gotpl template to use for generating resolvers
# resolver_template: [your/path/resolver.gotpl]
# Optional: turn on to avoid rewriting existing resolver(s) when generating
# preserve_resolver: false
# Optional: turn on use ` + "`" + `gqlgen:"fieldName"` + "`" + ` tags in your models
# struct_tag: json
# Optional: turn on to use []Thing instead of []*Thing
# omit_slice_element_pointers: false
# Optional: turn on to omit Is<Name>() methods to interface and unions
# omit_interface_checks: true
# Optional: turn on to skip generation of ComplexityRoot struct content and Complexity function
# omit_complexity: false
# Optional: turn on to not generate any file notice comments in generated files
# omit_gqlgen_file_notice: false
# Optional: turn on to exclude the gqlgen version in the generated file notice. No effect if `omit_gqlgen_file_notice` is true.
# omit_gqlgen_version_in_file_notice: false
# Optional: turn on to exclude root models such as Query and Mutation from the generated models file.
# omit_root_models: false
# Optional: turn on to exclude resolver fields from the generated models file.
# omit_resolver_fields: false
# Optional: turn off to make struct-type struct fields not use pointers
# e.g. type Thing struct { FieldA OtherThing } instead of { FieldA *OtherThing }
struct_fields_always_pointers: false
# Optional: turn off to make resolvers return values instead of pointers for structs
# resolvers_always_return_pointers: true
# Optional: turn on to return pointers instead of values in unmarshalInput
# return_pointers_in_unmarshalinput: false
# Optional: wrap nullable input fields with Omittable
# nullable_input_omittable: true
# Optional: set to speed up generation time by not performing a final validation pass.
# skip_validation: true
# Optional: set to skip running `go mod tidy` when generating server code
# skip_mod_tidy: true
# Optional: if this is set to true, argument directives that
# decorate a field with a null value will still be called.
#
# This enables argumment directives to not just mutate
# argument values but to set them even if they're null.
call_argument_directives_with_null: true
# Optional: set build tags that will be used to load packages
# go_build_tags:
# - private
# - enterprise
# Optional: set to modify the initialisms regarded for Go names
# go_initialisms:
# replace_defaults: false # if true, the default initialisms will get dropped in favor of the new ones instead of being added
# initialisms: # List of initialisms to for Go names
# - 'CC'
# - 'BCC'
# gqlgen will search for any type names in the schema in these go packages
# if they match it will use them, otherwise it will generate them.
autobind:
# - "github.com/lbf38/gql_repro/graph/model"
# This section declares type mapping between the GraphQL and go type systems
#
# The first line in each type will be used as defaults for resolver arguments and
# modelgen, the others will be allowed when binding to fields. Configure them to
# your liking
models:
ID:
model:
- github.com/99designs/gqlgen/graphql.ID
- github.com/99designs/gqlgen/graphql.Int
- github.com/99designs/gqlgen/graphql.Int64
- github.com/99designs/gqlgen/graphql.Int32
# gqlgen provides a default GraphQL UUID convenience wrapper for github.com/google/uuid
# but you can override this to provide your own GraphQL UUID implementation
UUID:
model:
- github.com/99designs/gqlgen/graphql.UUID
# The GraphQL spec explicitly states that the Int type is a signed 32-bit
# integer. Using Go int or int64 to represent it can lead to unexpected
# behavior, and some GraphQL tools like Apollo Router will fail when
# communicating numbers that overflow 32-bits.
#
# You may choose to use the custom, built-in Int64 scalar to represent 64-bit
# integers, or ignore the spec and bind Int to graphql.Int / graphql.Int64
# (the default behavior of gqlgen). This is fine in simple use cases when you
# do not need to worry about interoperability and only expect small numbers.
Int:
model:
- github.com/99designs/gqlgen/graphql.Int32
Int64:
model:
- github.com/99designs/gqlgen/graphql.Int
- github.com/99designs/gqlgen/graphql.Int64
schema.graphql:
# GraphQL schema example
#
# https://gqlgen.com/getting-started/
type Todo {
id: ID!
text: String
done: Boolean
user: User
}
type User {
id: ID!
name: String!
}
type Query {
todos: [Todo!]!
}
input NewTodo {
text: String!
userId: String!
}
type Mutation {
createTodo(input: NewTodo!): Todo!
}
models_gen.go:
// Code generated by github.com/99designs/gqlgen, DO NOT EDIT.
package model
type Mutation struct {
}
type NewTodo struct {
Text string `json:"text"`
UserID string `json:"userId"`
}
type Query struct {
}
type Todo struct {
ID string `json:"id"`
Text *string `json:"text,omitempty"`
Done *bool `json:"done,omitempty"`
User *User `json:"user,omitempty"`
}
type User struct {
ID string `json:"id"`
Name string `json:"name"`
}
Please observe that some fields aren't required in the GraphQL schema (Todo type). However, the generated model still have pointer fields in the Todo model.
Other issues references:
- #1683
- #3121
- #2512
Thanks, if you are able to figure out why this is happening and have a PR I'm happy to look at it.
@LBF38 -- I am the original author of the change that added this config option. It is not intended to prevent the use of pointers globally (i.e. for optional fields in the schema). It was intended to prevent the use of pointers for non-optional struct fields.
Here is the original PR for additional context: https://github.com/99designs/gqlgen/pull/2174
type Todo struct {
ID string `json:"id"`
Text *string `json:"text,omitempty"`
Done *bool `json:"done,omitempty"`
User *User `json:"user,omitempty"`
}
I guess using the pointer for those non-nilable types is reasonable because they cannot represent the null value if the pointer is removed.
@ianling Thanks for your reply ! Your explanation is crystal clear. I've now configured my project w/ custom models so the pointers doesn't bother me anymore, as I'm using GORM too.
I understand the need for this config and why it has been added. After reading your original PR, I think you should add this to the project's documentation. This will help future users better understand this configuration option.