Generate clojure.spec with GraphQL and extend GraphQL with clojure.spec.
Table of Contents
Serene can generate specs for an entire GraphQL API in one line of code.
Let's say you have a project called "Serenity" and a GraphQL API available at "http://localhost:3000/graphql"
:
(ns serenity.now
(:require
[clojure.spec.alpha :as s]
[paren.serene :as serene]
[paren.serene.schema :as schema]))
;; Define specs
(serene/def-specs (schema/fetch "http://localhost:3000/graphql"))
;; Use specs
(s/valid? ::User {:firstName "Frank" :lastName "Costanza"}) ;=> true
It is our experience that GraphQL is superior to REST for most APIs used for web and mobile applications. We also think that clojure.spec provides a good balance of expressiveness and strictness.
GraphQL's type system provides a point of leverage for API providers and consumers. Because GraphQL schemas are introspectable, GraphQL tooling tends to be very powerful. Some languages, like OCaml/Reason, can even validate queries and response code at compile time.
If other languages can leverage GraphQL to this extent, Clojure should be able to as well. Serene aims to address this.
Spec names are keywords that are prefixed and namespaced by their position in the schema.
For the example below, let's assume a prefix of :gql
, though the prefix is customizable.
# Built-in scalars are defined: :gql/Boolean, :gql/Float , :gql/ID , :gql/Int, :gql/String
scalar Email # :gql/Email
enum Mood { # :gql/Mood
SERENE # :gql.Mood/SERENE
ANNOYED # :gql.Mood/ANNOYED
ANGRY # :gql.Mood/ANGRY
}
type User { # :gql/User
id: ID! # :gql.User/id
username: String! # :gql.User/username
email: Email! # :gql.User/email
mood: Mood # :gql.User/mood
}
type Mutation { # :gql/Mutation
createUser(
username: String!, # :gql.Mutation.createUser/username
email: Email! # :gql.Mutation.createUser/email
mood: Mood # :gql.Mutation.createUser/mood
# :gql.Mutation.createUser/&args is an anonymous `s/keys` spec for args map
): User! # :gql.Mutation/createUser
}
paren.serene/def-specs
Note: All arguments to def-specs
are eval
ed.
def-specs
is a macro that will define specs for a GraphQL schema.
It takes compilation options as an optional second argument.
Compilation options are documented below.
(serene/def-specs gql-schema options)
paren.serene/spit-specs
spit-specs
is like def-specs
, but outputs s/def
forms to a file.
The file path and namespace are the first two arguments to spit-specs
.
(serene/spit-specs "src/api/specs.cljc" 'api.specs gql-schema options)
paren.serene.schema/fetch
fetch
takes a GraphQL server endpoint and optional configuration.
(schema/fetch "https://api.github.com/graphql" {:headers {"Authorization" (str "bearer " gh-access-token)}})
paren.serene.schema/query
query
is this GraphQL introspection query string.
fetch
works by asking the HTTP GraphQL server to execute query
.
You can use query
directly if your GraphQL API is not accessible via HTTP.
:extend
:extend
is a function or map of spec names to spec forms. If a spec form is returned, it will be combined with default specs using s/and
.
For example, if you have a custom Keyword
scalar you could use the following to add custom scalar validation:
(serene/def-specs gql-schema {:extend {:Keyword `keyword?}})
:alias
:alias
is a function or map which receives unprefixed spec names and returns aliases for those names.
(serene/def-specs gql-schema {:alias {:Query #{:api/query :api.query/root}
:Query/node :api.query/get-node}})
This would cause both :api/query
and :api.query/root
to be defined as aliases of the :Query
type spec and would cause :api.query/get-node
to be defined as an alias of the :Query/node
field spec.
:prefix
:prefix
is a wrapper around :alias
for the common case of altering default *ns*
prefixes.
For example, instead of having a long namespace prefix, you might want to prefix your specs with :gql
:
(serene/def-specs gql-schema {:prefix :gql})
This will produce specs like :gql/Query
, :gql.Query/node
, etc.
:gen-object-fields
:gen-object-fields
will cause test.check generators for object types to generate all fields,
even though all object fields are optional.
This is necessary if you are using test.check to generate data for object, interface, or union types because all object fields are optional (clients can query for any combination of fields).
However, generating data for map specs where all keys are optional can be frustrating because you often end up with empty or nearly empty maps. It is also not possible to always generate all fields, because objects can be cyclic, so you have to stop after some pre-determined level.
:gen-object-fields
solves this by always generating all fields up to n
where n
is a configurable depth that defaults to s/*recursion-limit*
.
;; modify all objects to generate `s/*recursion-limit*` levels deep
(schema/def-specs gql-schema {:gen-object-fields true})
;; modify only `Query` to generate 5 levels deep
(schema/def-specs gql-schema {:gen-object-fields {:Query 5}})
All compilation options are implemented and documented in paren.serene.compiler.transducers
.
Custom compilation options can be added in the same way that default options are provided.
Serene works in much the same way as GraphiQL and other GraphQL tools; it uses GraphQL's introspection capabilities. GraphQL schemas are introspectable, meaning that you can query a running API to determine all of the capabilities of that API.
Serene uses this introspection query, which is conveniently defined as paren.serene/introspection-query
, to generate specs that match your API.
If clojure.spec is alpha, then Serene is extra alpha.
Consider everything to be an implementation detail unless it is explicitly documented here.
Serene uses Break Versioning.
Copyright © Paren, LLC
Distributed under the Eclipse Public License version 2.0.
Can you improve this documentation?Edit on GitHub
cljdoc is a website building & hosting documentation for Clojure/Script libraries
× close