class GraphQL::Schema::RelayClassicMutation
Mutations that extend this base class get some conventions added for free:
-
An argument called `clientMutationId` is always added, but it's not passed to the resolve method. The value is re-inserted to the response. (It's for client libraries to manage optimistic updates.)
-
The returned object type always has a field called `clientMutationId` to support that.
-
The mutation accepts one argument called `input`, `argument`s defined in the mutation class are added to that input object, which is generated by the mutation.
These conventions were first specified by Relay
Classic, but they come in handy:
-
`clientMutationId` supports optimistic updates and cache rollbacks on the client
-
using a single `input:` argument makes it easy to post whole JSON objects to the mutation using one
GraphQL
variable (`$input`) instead of making a separate variable for each argument.
@see {GraphQL::Schema::Mutation} for an example, it's basically the same.
Public Class Methods
Also apply this argument to the input type:
GraphQL::Schema::Resolver::argument
# File lib/graphql/schema/relay_classic_mutation.rb, line 86 def argument(*args, **kwargs, &block) it = input_type # make sure any inherited arguments are already added to it arg = super # This definition might be overriding something inherited; # if it is, remove the inherited definition so it's not confused at runtime as having multiple definitions prev_args = it.own_arguments[arg.graphql_name] case prev_args when GraphQL::Schema::Argument if prev_args.owner != self it.own_arguments.delete(arg.graphql_name) end when Array prev_args.reject! { |a| a.owner != self } if prev_args.empty? it.own_arguments.delete(arg.graphql_name) end end it.add_argument(arg) arg end
Extend {Schema::Mutation.field_options} to add the `input` argument
GraphQL::Schema::Resolver::field_options
# File lib/graphql/schema/relay_classic_mutation.rb, line 129 def field_options sig = super # Arguments were added at the root, but they should be nested sig[:arguments].clear sig[:arguments][:input] = { type: input_type, required: true, description: "Parameters for #{graphql_name}" } sig end
The base class for generated input object types @param new_class [Class] The base class to use for generating input object definitions @return [Class] The base class for this mutation's generated input object (default is {GraphQL::Schema::InputObject})
# File lib/graphql/schema/relay_classic_mutation.rb, line 112 def input_object_class(new_class = nil) if new_class @input_object_class = new_class end @input_object_class || (superclass.respond_to?(:input_object_class) ? superclass.input_object_class : GraphQL::Schema::InputObject) end
@param new_input_type [Class, nil] If provided, it configures this mutation to accept `new_input_type` instead of generating an input type @return [Class] The generated {Schema::InputObject} class for this mutation's `input`
# File lib/graphql/schema/relay_classic_mutation.rb, line 121 def input_type(new_input_type = nil) if new_input_type @input_type = new_input_type end @input_type ||= generate_input_type end
Private Class Methods
Generate the input type for the `input:` argument To customize how input objects are generated, override this method @return [Class] a subclass of {.input_object_class}
# File lib/graphql/schema/relay_classic_mutation.rb, line 142 def generate_input_type mutation_args = all_argument_definitions mutation_name = graphql_name mutation_class = self Class.new(input_object_class) do graphql_name("#{mutation_name}Input") description("Autogenerated input type of #{mutation_name}") mutation(mutation_class) # these might be inherited: mutation_args.each do |arg| add_argument(arg) end argument :client_mutation_id, String, "A unique identifier for the client performing the mutation.", required: false end end
Public Instance Methods
Override {GraphQL::Schema::Resolver#resolve_with_support} to delete `client_mutation_id` from the kwargs.
GraphQL::Schema::Resolver#resolve_with_support
# File lib/graphql/schema/relay_classic_mutation.rb, line 31 def resolve_with_support(**inputs) # Without the interpreter, the inputs are unwrapped by an instrumenter. # But when using the interpreter, no instrumenters are applied. if context.interpreter? input = inputs[:input].to_kwargs new_extras = field ? field.extras : [] all_extras = self.class.extras + new_extras # Transfer these from the top-level hash to the # shortcutted `input:` object all_extras.each do |ext| # It's possible that the `extra` was not passed along by this point, # don't re-add it if it wasn't given here. if inputs.key?(ext) input[ext] = inputs[ext] end end else input = inputs end if input # This is handled by Relay::Mutation::Resolve, a bit hacky, but here we are. input_kwargs = input.to_h client_mutation_id = input_kwargs.delete(:client_mutation_id) else # Relay Classic Mutations with no `argument`s # don't require `input:` input_kwargs = {} end return_value = if input_kwargs.any? super(**input_kwargs) else super() end # Again, this is done by an instrumenter when using non-interpreter execution. if context.interpreter? context.schema.after_lazy(return_value) do |return_hash| # It might be an error if return_hash.is_a?(Hash) return_hash[:client_mutation_id] = client_mutation_id end return_hash end else return_value end end