View Source Absinthe.Schema behaviour (absinthe v1.7.6)
Build GraphQL Schemas
Custom Schema Manipulation (in progress)
In Absinthe 1.5 and up schemas are built using the same process by which queries are
executed. All the macros in this module and in Notation
build up an intermediary tree of structs in the
%Absinthe.Blueprint{}
namespace, which we generally call "Blueprint structs".
At the top you've got a %Blueprint{}
struct which holds onto some schema
definitions that look a bit like this:
%Blueprint.Schema.SchemaDefinition{
type_definitions: [
%Blueprint.Schema.ObjectTypeDefinition{identifier: :query, ...},
%Blueprint.Schema.ObjectTypeDefinition{identifier: :mutation, ...},
%Blueprint.Schema.ObjectTypeDefinition{identifier: :user, ...},
%Blueprint.Schema.EnumTypeDefinition{identifier: :sort_order, ...},
]
}
You can see what your schema's blueprint looks like by calling
__absinthe_blueprint__
on any schema or type definition module.
defmodule MyAppWeb.Schema do
use Absinthe.Schema
query do
end
end
> MyAppWeb.Schema.__absinthe_blueprint__
#=> %Absinthe.Blueprint{...}
These blueprints are manipulated by phases, which validate and ultimately construct a schema. This pipeline of phases you can hook into like you do for queries.
defmodule MyAppWeb.Schema do
use Absinthe.Schema
@pipeline_modifier MyAppWeb.CustomSchemaPhase
query do
end
end
defmodule MyAppWeb.CustomSchemaPhase do
alias Absinthe.{Phase, Pipeline, Blueprint}
# Add this module to the pipeline of phases
# to run on the schema
def pipeline(pipeline) do
Pipeline.insert_after(pipeline, Phase.Schema.TypeImports, __MODULE__)
end
# Here's the blueprint of the schema, do whatever you want with it.
def run(blueprint, _) do
{:ok, blueprint}
end
end
The blueprint structs are pretty complex, but if you ever want to figure out how to construct something in blueprints you can always just create the thing in the normal AST and then look at the output. Let's see what interfaces look like for example:
defmodule Foo do
use Absinthe.Schema.Notation
interface :named do
field :name, :string
end
end
Foo.__absinthe_blueprint__ #=> ...
Summary
Callbacks
Used to set some values in the context that it may need in order to run.
Used to hydrate the schema with dynamic attributes.
Used to apply middleware on all or a group of fields based on pattern matching.
Used to define the list of plugins to run before and after resolution.
Functions
Get all concrete types for union, interface, or object
List all directives on a schema
List all implementors of an interface on a schema
Run the introspection query on a schema.
Get all introspection types
Defines a root Mutation object
Defines a root Query object
Get all types that are referenced by an operation
Replace the default middleware.
Defines a root Subscription object
Converts a schema to an SDL string
List all types on a schema
Get all types that are used by an operation
Types
@type t() :: module()
Callbacks
Used to set some values in the context that it may need in order to run.
Examples
Setup dataloader:
def context(context) do
loader =
Dataloader.new
|> Dataloader.add_source(Blog, Blog.data())
Map.put(context, :loader, loader)
end
@callback hydrate( node :: Absinthe.Blueprint.Schema.t(), ancestors :: [Absinthe.Blueprint.Schema.t()] ) :: Absinthe.Schema.Hydrator.hydration()
Used to hydrate the schema with dynamic attributes.
While this is normally used to add resolvers, etc, to schemas
defined using import_sdl/1
and import_sdl/2
, it can also be
used in schemas defined using other macros.
The function is passed the blueprint definition node as the first argument and its ancestors in a list (with its parent node as the head) as its second argument.
See the Absinthe.Phase.Schema.Hydrate
implementation of
Absinthe.Schema.Hydrator
callbacks to see what hydration
values can be returned.
Examples
Add a resolver for a field:
def hydrate(%Absinthe.Blueprint.Schema.FieldDefinition{identifier: :health}, [%Absinthe.Blueprint.Schema.ObjectTypeDefinition{identifier: :query} | _]) do
{:resolve, &__MODULE__.health/3}
end
# Resolver implementation:
def health(_, _, _), do: {:ok, "alive!"}
Note that the values provided must be macro-escapable; notably, anonymous functions cannot be used.
You can, of course, omit the struct names for brevity:
def hydrate(%{identifier: :health}, [%{identifier: :query} | _]) do
{:resolve, &__MODULE__.health/3}
end
Add a description to a type:
def hydrate(%Absinthe.Blueprint.Schema.ObjectTypeDefinition{identifier: :user}, _) do
{:description, "A user"}
end
If you define hydrate/2
, don't forget to include a fallback, e.g.:
def hydrate(_node, _ancestors), do: []
@callback middleware( [Absinthe.Middleware.spec(), ...], Absinthe.Type.Field.t(), Absinthe.Type.Object.t() ) :: [Absinthe.Middleware.spec(), ...]
Used to apply middleware on all or a group of fields based on pattern matching.
It is passed the existing middleware for a field, the field itself, and the object that the field is a part of.
Examples
Adding a HandleChangesetError
middleware only to mutations:
# if it's a field for the mutation object, add this middleware to the end
def middleware(middleware, _field, %{identifier: :mutation}) do
middleware ++ [MyAppWeb.Middleware.HandleChangesetErrors]
end
# if it's any other object keep things as is
def middleware(middleware, _field, _object), do: middleware
@callback plugins() :: [Absinthe.Plugin.t()]
Used to define the list of plugins to run before and after resolution.
Plugins are modules that implement the Absinthe.Plugin
behaviour. These modules
have the opportunity to run callbacks before and after the resolution of the entire
document, and have access to the resolution accumulator.
Plugins must be specified by the schema, so that Absinthe can make sure they are all given a chance to run prior to resolution.
Functions
@spec apply_modifiers(Absinthe.Pipeline.t(), t()) :: Absinthe.Pipeline.t()
@spec concrete_types(t(), Absinthe.Type.t()) :: [Absinthe.Type.t()]
Get all concrete types for union, interface, or object
@spec directives(t()) :: [Absinthe.Type.Directive.t()]
List all directives on a schema
@spec implementors(t(), Absinthe.Type.identifier_t() | Absinthe.Type.Interface.t()) :: [ Absinthe.Type.Object.t() ]
List all implementors of an interface on a schema
@spec introspect(schema :: t(), opts :: Absinthe.run_opts()) :: Absinthe.run_result()
Run the introspection query on a schema.
Convenience function.
@spec introspection_types(t()) :: [Absinthe.Type.t()]
Get all introspection types
Defines a root Mutation object
mutation do
field :create_user, :user do
arg :name, non_null(:string)
arg :email, non_null(:string)
resolve &MyApp.Web.BlogResolvers.create_user/2
end
end
Defines a root Query object
@spec referenced_types(t()) :: [Absinthe.Type.t()]
Get all types that are referenced by an operation
Replace the default middleware.
Examples
Replace the default for all fields with a string lookup instead of an atom lookup:
def middleware(middleware, field, object) do
new_middleware = {Absinthe.Middleware.MapGet, to_string(field.identifier)}
middleware
|> Absinthe.Schema.replace_default(new_middleware, field, object)
end
subscription(raw_attrs \\ [name: "RootSubscriptionType"], list)
View Source (macro)Defines a root Subscription object
Subscriptions in GraphQL let a client submit a document to the server that outlines what data they want to receive in the event of particular updates.
For a full walk through of how to setup your project with subscriptions and
Phoenix
see the Absinthe.Phoenix
project moduledoc.
When you push a mutation, you can have selections on that mutation result to get back data you need, IE
mutation {
createUser(accountId: 1, name: "bob") {
id
account { name }
}
}
However, what if you want to know when OTHER people create a new user, so that your UI can update as well. This is the point of subscriptions.
subscription {
newUsers {
id
account { name }
}
}
The job of the subscription macros then is to give you the tools to connect
subscription documents with the values that will drive them. In the last example
we would get all users for all accounts, but you could imagine wanting just
newUsers(accountId: 2)
.
In your schema you articulate the interests of a subscription via the config
macro:
subscription do
field :new_users, :user do
arg :account_id, non_null(:id)
config fn args, _info ->
{:ok, topic: args.account_id}
end
end
end
The topic can be any term. You can broadcast a value manually to this subscription by doing
Absinthe.Subscription.publish(pubsub, user, [new_users: user.account_id])
It's pretty common to want to associate particular mutations as the triggers for one or more subscriptions, so Absinthe provides some macros to help with that too.
subscription do
field :new_users, :user do
arg :account_id, non_null(:id)
config fn args, _info ->
{:ok, topic: args.account_id}
end
trigger :create_user, topic: fn user ->
user.account_id
end
end
end
The idea with a trigger is that it takes either a single mutation :create_user
or a list of mutations [:create_user, :blah_user, ...]
and a topic function.
This function returns a value that is used to lookup documents on the basis of
the topic they returned from the config
macro.
Note that a subscription field can have trigger
as many trigger blocks as you
need, in the event that different groups of mutations return different results
that require different topic functions.
Converts a schema to an SDL string
Per the spec, only types that are actually referenced directly or transitively from the root query, subscription, or mutation objects are included.
Example
Absinthe.Schema.to_sdl(MyAppWeb.Schema)
"schema {
query {...}
}"
@spec types(t()) :: [Absinthe.Type.t()]
List all types on a schema
@spec used_types(t()) :: [Absinthe.Type.t()]
Get all types that are used by an operation