This is a pure PHP realization of the GraphQL protocol based on the working draft of the official GraphQL Specification located on https://github.com/facebook/graphql.
GraphQL is a modern replacement of the almost obsolete REST approach to present API. It's been almost 16 years since the REST idea was found in 2000 by Roy Fielding. With all credit to everything we accomplished using REST it's time to change for something better. GraphQL advanced in many ways and has fundamental improvements over the old good REST:
- self-checks embedded on the ground level of your backend architecture
- reusable API for different client versions and devices, i.e. no more need in maintaining "/v1" and "/v2"
- a complete new level of distinguishing of the backend and frontend logic
- easily generated documentation and incredibly intuitive way to explore created API
- once your architecture is complete – most client-based changes does not require backend modifications
It could be hard to believe but give it a try and you'll be rewarded with much better architecture and so much easier to support code.
Current package is and will be trying to be kept up to date with the latest revision of the official GraphQL Specification which is now of April 2016.
Symfony bundle is available by the link – http://github.com/Youshido/GraphqlBundle
- Getting Started
- Installation
- Example – Creating Blog Schema
- Query Documents
- Type System
- Building your schema
- Useful information
You should be better off starting with some examples, and "Star Wars" become a somewhat "Hello world" example for the GraphQL implementations. We have that too and if you're looking for just that go directly by this link – Star Wars example. On the other hand based on the feedback we prepared a step-by-step guide for those who wants to get up to speed bit by bit.
You should simply install this package using the composer. If you're not familiar with it you should check out their manual.
Add the following package to your composer.json
(or simply create a new file with this content):
{
"require": {
"youshido/graphql": "*"
}
}
Run composer update
.
Alternatively you can do the following sequence:
mkdir graphql-test && cd graphql-test
composer init
composer require youshido/graphql
Having the previous step done you're ready to create your first GraphQL Schema and check if you've successful installation.
Your simple schema will accept one request currentTime
and response with a formatted time string.
you can find this example in the examples directory – 01_sandbox.
Create an index.php
file with the following content:
<?php
namespace Sandbox;
use Youshido\GraphQL\Execution\Processor;
use Youshido\GraphQL\Schema\Schema;
use Youshido\GraphQL\Type\Object\ObjectType;
use Youshido\GraphQL\Type\Scalar\StringType;
require_once 'vendor/autoload.php';
$processor = new Processor(new Schema([
'query' => new ObjectType([
'name' => 'RootQueryType',
'fields' => [
'currentTime' => [
'type' => new StringType(),
'resolve' => function() {
return date('Y-m-d H:ia');
}
]
]
])
]));
$processor->processPayload('{ currentTime }');
echo json_encode($processor->getResponseData()) . "\n";
Now you can run php index.php
and if everything was set up correctly you should see the response with your current time:
{
data: { currentTime: "2016-05-01 19:27pm" }
}
If you're having any troubles – here are some ideas to troubleshoot them:
- check that you have the latest composer version (
composer self-update
) - make sure your
index.php
file has been created in the same directory that you havevendor
folder in (presumably it'sgraphql-test
folder) - last but not least, check that you have php-cli installed and running and it's version >= 5.5 (
php -v
)
Also, you can always run a script from examples
folder. Simply run php vendor/youshido/GraphQL/examples/01_sandbox/index.php
.
For our learning example we'll architect a GraphQL Schema for the Blog.
To keep it simple we'll just have Users who can write Posts and leave Comments. Take a look at the example query that returns title and summary of the latest Post:
latestPost {
title,
summary
}
As you can see, GraphQL query is a simple text query structured very much similar to the json format.
Supposedly our server should reply with a json response structured the following way:
{
data: {
latestPost: {
title: "This is a post title",
summary: "This is a post summary"
}
}
}
Let's go ahead and create a backend that can handle that.
You'll probably be using our package along with your favorite framework (we have a Symfony version here). But for the purpose of the current example we'll keep it all examples as plain php code. (Complete example of the Blog schema available by the following link https://github.com/Youshido/GraphQL/examples/02_Blog)
We'll take a quick look on different approaches you can use to define your schema. Each of them has it's own pros and cons, inline approach might seem to be easier and faster when object oriented gives you more flexibility and freedom as your project grows. We believe that you should architect your schema using the OOP approach and use inline for small implementations where it's appropriate.
Let's start by creating the Post type. For now we'll have only two fields – a title and a summary:
inline-schema.php
<?php
use Youshido\GraphQL\Type\Object\ObjectType;
use Youshido\GraphQL\Type\Scalar\StringType;
// creating a root query structure
$rootQueryType = new ObjectType([
// name for the root query type doesn't matter, by the convention it's RootQueryType
'name' => 'RootQueryType',
'fields' => [
'latestPost' => [
'type' => new ObjectType([ // the Post type will be extended from the generic ObjectType
'name' => 'Post', // name of our type – "Post"
'fields' => [
'title' => new StringType(), // defining the "title" field, type - String
'summary' => new StringType(), // defining the "summary" field, also a String type
],
]),
'resolve' => function () { // this is a resolve function
return [ // for now it returns a static array with data
"title" => "New approach in API has been revealed",
"summary" => "This post describes a new approach to create and maintain APIs",
];
}
]
]
]);
We should also create an endpoint to work with our schema so we can actually test what we do. Later on we'll add an ability to handle the requests from the client.
index.php
<?php
namespace BlogTest;
use Youshido\GraphQL\Execution\Processor;
use Youshido\GraphQL\Schema\Schema;
use Youshido\GraphQL\Type\Object\ObjectType;
use Youshido\GraphQL\Validator\ResolveValidator\ResolveValidator;
require_once __DIR__ . '/vendor/autoload.php';
require_once __DIR__ . '/inline-schema.php'; // including our schema
/** @var ObjectType $rootQueryType */
$processor = new Processor(new Schema([
'query' => $rootQueryType
]));
$payload = '{ latestPost { title, summary } }';
$processor->processPayload($payload);
echo json_encode($processor->getResponseData()) . "\n";
To check if everything is working well, simply execute index.php – php index.php
You should see a json encoded object latestPost
inside the data
section:
{
data: {
latestPost: {
title: "New approach in API has been revealed",
summary: "This post describes a new approach to create and maintain APIs"
}
}
}
You can try to play with the code by removing one field from the request or by changing the resolve function.
We're going to create a separate class for the PostType
and then use it to build our GraphQL Schema.
To keep everything structured we're going to put this and all our future classes to the Schema
folder.
Create a file Schema/PostType.php
and put the following content there:
<?php
namespace Examples\Blog\Schema;
use Youshido\GraphQL\Type\NonNullType;
use Youshido\GraphQL\Type\Object\AbstractObjectType;
use Youshido\GraphQL\Type\Scalar\IdType;
use Youshido\GraphQL\Type\Scalar\StringType;
class PostType extends AbstractObjectType // extending abstract Object type
{
public function build($config) // implementing an abstract function where you build your type
{
$config->addField('title', new StringType()) // adding title field of type String
->addField('summary', new StringType()); // adding summary field of type String
}
public function getName()
{
return "Post"; // important to use the real name here, it will be used later in the Schema
}
}
In order to make it work we need to update our index.php
as well:
<?php
namespace BlogTest;
use Youshido\GraphQL\Execution\Processor;
use Youshido\GraphQL\Schema\Schema;
use Youshido\GraphQL\Type\Object\ObjectType;
require_once __DIR__ . '/vendor/autoload.php';
require_once __DIR__ . '/Schema/PostType.php'; // including PostType definition
$rootQueryType = new ObjectType([
'name' => 'RootQueryType',
'fields' => [
// you can specify fields for your query as an array
'latestPost' => [
'type' => new PostType(),
'resolve' => function ($value, $args, $info) // implementing resolve function
{
return [
"title" => "New approach in API has been revealed",
"summary" => "This post describes a new approach to create and maintain APIs",
];
}
]
]
]);
$processor = new Processor(new Schema([
'query' => $rootQueryType
]));
$payload = '{ latestPost { title, summary } }';
$processor->processPayload($payload);
echo json_encode($processor->getResponseData()) . "\n";
Ensure everything is working properly by running php index.php
. You should see the same response you saw for the inline approach.
You can make a complete separate class for the field to incapsulate the resolve function, you'll have to extend AbstractField
class to do so:
class LatestPostField extends AbstractField
{
public function getType()
{
return new PostType();
}
public function resolve($value, array $args, ResolveInfo $info)
{
return [
"title" => "New approach in API has been revealed",
"summary" => "This post describes a new approach to create and maintain APIs",
];
}
}
We will use this class later in our Schema by adding it's instance to fields
definition:
$rootQueryType = new ObjectType([
'name' => 'RootQueryType',
'fields' => [
new LatestPostField()
]
]);
We would recommend to stick to object oriented approach for the several reasons (that matter the most for the GraphQL specifically):
- it makes your Types reusable
- adds an ability to refactor your schema using IDEs
- autocomplete to help you avoid typos
- much easier to navigate through your Schema
Inline approach is usually used to bootstrap and explore your ideas or to create simple parts of the schema where creating classes might be an overkill. With the inline approach you can be fast and agile in creating mock-data server to test your frontend or mobile client.
Use valid Names We highly recommend to get familiar with the official GraphQL Specification Remember that valid identifier in GraphQL should follow the pattern
/[_A-Za-z][_0-9A-Za-z]*/
. That means any identifier should consist of a latin letter, underscore, or a digit and cannot start with a digit. Names are case sensitive
We'll continue to develop on the Blog Schema to explore all the details of GraphQL.
Query Document in terms of GraphQL describes a complete request received by GraphQL service. It contains list of Operations and Fragments. Both are fully supported by our PHP library. There are two types of Operations in GraphQL:
- Query – a read only request that is not supposed to do any changes on the server
- Mutation – a request that changes data on the server followed by a data fetch
You've already seen latestPost
and currentTime
queries in the examples above, so let's define a simple Mutation that will provide an API to Like the Post.
Here's an example of the request sent and expected result of the simplest possible mutation for our task:
request
mutation {
likePost(id: 5)
}
response
{
data: { likePost: 2 }
}
Any Operation needs to have a response type and in this case the mutation type is
Int
Note, that the response type of this mutation is a scalar Int
.
Of course in real life you'll more likely have a response of type Post
for such mutation, but we're going to create a simple example above and even keep it inside the index.php
:
<?php
namespace BlogTest;
use Examples\Blog\Schema\PostType;
use Examples\Blog\Schema\LatestPostField;
use Youshido\GraphQL\Execution\Processor;
use Youshido\GraphQL\Schema\Schema;
use Youshido\GraphQL\Type\NonNullType;
use Youshido\GraphQL\Type\Object\ObjectType;
use Youshido\GraphQL\Type\Scalar\IntType;
require_once __DIR__ . '/vendor/autoload.php';
require_once __DIR__ . '/Schema/PostType.php'; // including PostType definition
require_once __DIR__ . '/Schema/LatestPostField.php'; // including field definition
$rootQueryType = new ObjectType([
'name' => 'RootQueryType',
'fields' => [
'latestPost' => new LatestPostField()
]
]);
$rootMutationType = new ObjectType([
'name' => 'RootMutationType',
'fields' => [
// here's our likePost mutation
'likePost' => [
// we specify the output type – simple Int, since it doesn't have a structure
// mutation name will be used as a key for the result
'type' => new IntType(),
// we set the argument for our mutation, in our case it's an Int
// with a composition of NonNull
'args' => [
'id' => new NonNullType(new IntType())
],
// simple resolve function that always returns 2
'resolve' => function () {
return 2;
},
]
]
]);
$processor = new Processor(new Schema([
'query' => $rootQueryType,
'mutation' => $rootMutationType,
]));
$payload = 'mutation { likePost(id:5) }';
$processor->processPayload($payload);
echo json_encode($processor->getResponseData()) . "\n";
If you run php index.php
you should see a valid response:
{"data":{"likePost":2}}
Now, let's make our GraphQL Schema a little more complex by adding a likeCount
field to the PostType
:
<?php
// add it after the last ->addField in your build function
->addField('likeCount', new IntType())
// update the resolve function:
public function resolve($value, $args, $info)
{
$id = !empty($args['id']) ? $args['id'] : null;
return [
"title" => "Post " . $id . " title",
"summary" => "This new GraphQL library for PHP works really well",
"likeCount" => 2
];
}
Let's also change our mutation type from the IntType
to the PostType
and update the resolve
function to be complaint with the the new type we set:
<?php
$rootMutationType = new ObjectType([
'name' => 'RootMutationType',
'fields' => [
'likePost' => [
'type' => new PostType(),
'args' => [
'id' => new NonNullType(new IntType())
],
'resolve' => function ($value, $args, $type) {
// adding like count code goes here
return [
'title' => 'Title for the post #' . $args['id'],
'summary' => 'We can now get a richer response from the mutation',
'likeCount' => 2
];
},
]
]
]);
As you can see we're repeating ourselves with the resolve function. Since we already have one that can return PostType
structure, we can utilize it by using the 3rd argument of the resolve
function – it's output type:
$rootMutationType = new ObjectType([
'name' => 'RootMutationType',
'fields' => [
'likePost' => [
'type' => new PostType(),
'args' => [
'id' => new NonNullType(new IntType())
],
'resolve' => function ($value, $args, $type) {
// adding like count code goes here
return $type->resolve($value, $args, $type);
},
]
]
]);
Finally, run this code to make sure everything is good so far – php index.php
:
{
"data": {
"likePost": {
"title":"Title for the post #5",
"summary":"We can now get a richer response from the mutation",
"likeCount":2
}
}
}
Now you have a basic understanding of how queries and mutations are structured and ready to move on to the details of the GraphQL Type System and PHP-specific features of the GraphQL service architecture.
Type is an atom of definition in GraphQL Schema. Every field, object, or argument has a type. As a result – GraphQL is a strongly typed language.
There are common types and types defined specifically for the application, in our case we'll have types like Post
, User
, Comment
and so on.
GraphQL has variety of build in types that are usually used to build your own custom types.
List of GraphQL Scalar types:
- Int
- Float
- String
- Boolean
- Id (serialized as String per spec)
In addition, we implemented some types that might be useful and which we're considering to be scalar as well:
- Timestamp
- Date
- DateTime
- DateTimeTz
If you will ever need to define a new Scalar type, you can do that by extending from the AbstractScalarType
class.
usage of scalar types will be shown in combination with other types along the way
Every entity in your business logic will probably have a class that represents it's type. That class must be either extended from the AbstractObjectType
or created as an instance of ObjectType
.
In our blog example we used ObjectType
to create an inline Post
type and in the object oriented example we extended the AbstractObjectType
to create a PostType
class.
Let's look closer on the structure of the ObjectType
and pay attention to it's fields configuration.
We'll start with the inline approach:
<?php
$postType = new ObjectType([
// you have to specify a valid name
'name' => 'Post',
// fields are represented as an array
'fields' => [
// here you have a complex field with a lot of parameters defined
'title' => [
'type' => new StringType(), // string type
'description' => 'This field contains a post title', // description
'isDeprecated' => true, // marked as deprecated
'deprecationReason' => 'field title is now deprecated', // explain the reason
'args' => [
'truncated' => new BooleanType() // add an optional argument
],
'resolve' => function ($value, $args, $info) {
// using argument defined above to modify a field value
return (!empty($args['truncated'])) ? explode(' ', $value)[0] . '...' : $value;
}
],
// if field just has a type, you can use a short declaration syntax like this
'summary' => new StringType(),
'likeCount' => new IntType(),
],
// arguments for the query
'args' => [
'id' => new IntType()
],
// resolve function for the query
'resolve' => function ($value, $args, $info) {
return [
'title' => 'Title for the latest Post',
'summary' => 'Post summary',
'likeCount' => 2,
];
},
]);
And in comparison take a look at the Object oriented version with all the same fields:
<?php
class PostType extends AbstractObjectType
{
public function getName()
{
return "Post";
}
public function build(TypeConfigInterface $config)
{
$config
->addField('title', [
'type' => new NonNullType(new StringType()),
'description' => 'This field contains a post title',
'isDeprecated' => true,
'deprecationReason' => 'field title is now deprecated',
'args' => [
'truncated' => new BooleanType()
],
'resolve' => function ($value, $args) {
return (!empty($args['truncated'])) ? explode(' ', $value)[0] . '...' : $value;
}
])
->addField('summary', new StringType())
->addField('likeCount', new IntType());
$config->addArgument('id', new IntType());
}
public function resolve($value, $args, $info)
{
return [
"title" => "Title for the latest Post",
"summary" => "Post summary",
"likeCount" => 2
];
}
}
Once again, it's not that a difference between two approaches but having a separate class for the Type will give you some freedom and flexibility along the way.
GraphQL supports Interfaces
. You can define Interface and use it as a Type
of an item in the List
, or use Interface to make sure that specific objects certainly have fields you need.
Each InterfaceType
needs to have at least one defined field and resolveType
function. That function will be used to determine what exact Type
of the object GraphQL resolver is currently working on.
Let's create a ContentBlockInterface
that can represent a piece of content for the web page that have a title
and a summary
.
<?php
/**
* ContentBlockInterface.php
*/
namespace Examples\Blog\Schema;
use Youshido\GraphQL\Type\Config\TypeConfigInterface;
use Youshido\GraphQL\Type\NonNullType;
use Youshido\GraphQL\Type\Object\AbstractInterfaceType;
use Youshido\GraphQL\Type\Scalar\StringType;
class ContentBlockInterface extends AbstractInterfaceType
{
public function build(TypeConfigInterface $config)
{
$config->addField('title', new NonNullType(new StringType()));
$config->addField('summary', new StringType());
}
public function resolveType($object) {
// since there's only one type right now this interface will always resolve PostType
return new PostType();
}
}
Most often you'll be using only the build
function to define fields and/or arguments that need to be implemented.
In order to add this Interface to the PostType
we have to override it's getInterfaces
method:
<?php
/**
* PostType.php
*/
namespace Examples\Blog\Schema;
use Youshido\GraphQL\Type\Config\TypeConfigInterface;
use Youshido\GraphQL\Type\NonNullType;
use Youshido\GraphQL\Type\Object\AbstractObjectType;
use Youshido\GraphQL\Type\Scalar\BooleanType;
use Youshido\GraphQL\Type\Scalar\IntType;
use Youshido\GraphQL\Type\Scalar\StringType;
class PostType extends AbstractObjectType
{
public function build(TypeConfigInterface $config)
{
$config
->addField('title', new StringType())
->addField('summary', new StringType())
->addField('likeCount', new IntType());
}
public function getInterfaces()
{
return [new ContentBlockInterface()];
}
public function resolve($value, $args, $info)
{
return [
"title" => "Post title from the PostType class",
"summary" => "This new GraphQL library for PHP works really well",
"likeCount" => 2
];
}
}
As you might have noticed there's no getName
method in both Interface and Type classes – that's a simplified approach available when you want to have your name exactly the same as the class name without the Type
at the end.
If you run the script as it is right now – php index.php
, you should get an error:
{"errors":[{"message":"Implementation of ContentBlockInterface is invalid for the field title"}]}
You've got this error because the title
field definition in the PostType
is different from the one described in the ContentBlockInterface
.
To fix it we have to declare fields that exist in the Interface
with the same names and same types in out PostType
.
We already have title
but it's a nullable field so we have to change it by adding a non-null wrapper – new NonNullType(new StringType())
.
You can check the result by executing index.php script again, you should get the usual response.
GraphQL Enums are the variation on the Scalar type, which represents one of the predefined values. Enums serialize as a string: the name of the represented value but can be associated with a numeric (as an example) value.
To show you how Enums work we're going to create a new class - PostStatus
:
<?php
/**
* PostStatus.php
*/
namespace Examples\Blog\Schema;
use Youshido\GraphQL\Type\Enum\AbstractEnumType;
class PostStatus extends AbstractEnumType
{
public function getValues()
{
return [
[
'value' => 0,
'name' => 'DRAFT',
],
[
'value' => 1,
'name' => 'PUBLISHED',
]
];
}
}
Now when you have this class created you can add a status field to our PostType
:
<?php
// add field to the build function of the PostType class
->addField('status', new PostStatus())
// and update the resolve function
return [
"title" => "Post title from the PostType class",
"summary" => "This new GraphQL library for PHP works really well",
"status" => 1,
"likeCount" => 2
];
Call the status
field in your request:
$payload = '{ latestPost { title, status, likeCount } }';
You should get a result similar to the following:
{"data":{"latestPost":{"title":"Post title from the PostType class","status":"PUBLISHED","likeCount":2}}}
GraphQL Unions represent an object type that could be resolved as one of a specified GraphQL Object types.
To get you an idea of what this is we'll create a new query field that will return a list of unions (and get to the ListType
after it).
You can consider Union as a combined type that is needed mostly when you want to have a list of different objects
Imaging that you have a page and you need to get all content blocks for this page. Let content block be either Post
or Banner
.
Create a BannerType
:
<?php
/**
* BannerType.php
*/
namespace Examples\Blog\Schema;
use Youshido\GraphQL\Type\Config\TypeConfigInterface;
use Youshido\GraphQL\Type\Object\AbstractObjectType;
use Youshido\GraphQL\Type\Scalar\StringType;
class BannerType extends AbstractObjectType
{
public function build(TypeConfigInterface $config)
{
$config
->addField('title', new StringType())
->addField('imageLink', new StringType());
}
}
Now let's combine the Banner
type and the Post
type to create a ContentBlockUnion
that will extend an AbstractUnionType
.
Each UnionType
needs to define a list of types it unites by implementing the getTypes
method and the resolveType
method to resolve object that will be returned for each instance of the Union
.
<?php
/**
* ContentBlockUnion.php
*/
namespace Examples\Blog\Schema;
use Youshido\GraphQL\Type\Object\AbstractUnionType;
class ContentBlockUnion extends AbstractUnionType
{
public function getTypes()
{
return [new PostType(), new BannerType()];
}
public function resolveType($object)
{
// we simple look if there's a "post" inside the object id that it's a PostType otherwise it's a BannerType
return empty($object['id']) ? null : (strpos($object['id'], 'post') !== false ? new PostType() : new BannerType());
}
}
We're also going to create a simple DataProvider
that will give us test data to operate with:
<?php
/**
* DataProvider.php
*/
namespace Examples\Blog\Schema;
class DataProvider
{
public static function getPost($id)
{
return [
"id" => "post-" . $id,
"title" => "Post " . $id . " title",
"summary" => "This new GraphQL library for PHP works really well",
"status" => 1,
"likeCount" => 2
];
}
public static function getBanner($id)
{
return [
'id' => "banner-" . $id,
'title' => "Banner " . $id,
'imageLink' => "banner" . $id . ".jpg"
];
}
}
Now, we're ready to update our Schema and include ContentBlockUnion
into it.
As we're getting our schema bigger we'd like to extract it to a separate file:
<?php
/**
* BlogSchema.php
*/
namespace Examples\Blog\Schema;
use Youshido\GraphQL\AbstractSchema;
use Youshido\GraphQL\Type\Config\Schema\SchemaConfig;
use Youshido\GraphQL\Type\ListType\ListType;
class BlogSchema extends AbstractSchema
{
public function build(SchemaConfig $config)
{
$config->getQuery()->addFields([
'latestPost' => new PostType(),
'randomBanner' => [
'type' => new BannerType(),
'resolve' => function () {
return DataProvider::getBanner(rand(1, 10));
}
],
'pageContentUnion' => [
'type' => new ListType(new ContentBlockUnion()),
'resolve' => function () {
return [DataProvider::getPost(1), DataProvider::getBanner(1)];
}
]
]);
$config->getMutation()->addFields([
'likePost' => new LikePost()
]);
}
}
Having this separate schema file you should update your index.php
to look like this:
<?php
namespace BlogTest;
use Youshido\GraphQL\Execution\Processor;
use Youshido\GraphQL\Schema\Schema;
require_once __DIR__ . '/schema-bootstrap.php';
$schema = new BlogSchema();
$processor = new Processor($schema);
$payload = '{ pageContentUnion { ... on Post { title } ... on Banner { title, imageLink } } }';
$processor->processPayload($payload);
echo json_encode($processor->getResponseData()) . "\n";
For the convenience of use we've created the schema-bootstrap.php
:
<?php
namespace BlogTest;
require_once __DIR__ . '/vendor/autoload.php';
require_once __DIR__ . '/Schema/DataProvider.php';
require_once __DIR__ . '/Schema/PostType.php';
require_once __DIR__ . '/Schema/PostStatus.php';
require_once __DIR__ . '/Schema/ContentBlockInterface.php';
require_once __DIR__ . '/Schema/LikePost.php';
require_once __DIR__ . '/Schema/BannerType.php';
require_once __DIR__ . '/Schema/ContentBlockUnion.php';
require_once __DIR__ . '/Schema/BlogSchema.php';
Due to the GraphQL syntax you have to specify fields for each type of object you're getting in the union request, if you're not familiar with it read more at [official documentation(https://facebook.github.io/graphql/#sec-Unions)] If everything was done right you should see the following response:
{
"data": {
"pageContentUnion":[
{"title":"Post 1 title","summary":"This new GraphQL library for PHP works really well"},
{"title":"Banner 1","imageLink":"banner1.jpg"}
]
}
}
Also, you might want to check out how to use GraphiQL tool to get a better visualization of what you're doing here.
As you've seen in the previous example ListType
is used to create a list of any items that are or extend GraphQL type.
List type can be also created by using InterfaceType
as an item which gives you flexibility in defining your schema.
Let's go ahead and add ListType
field to our BlogSchema.
<?php
/**
* BlogSchema.php
*/
namespace Examples\Blog\Schema;
use Youshido\GraphQL\AbstractSchema;
use Youshido\GraphQL\Type\Config\Schema\SchemaConfig;
use Youshido\GraphQL\Type\ListType\ListType;
class BlogSchema extends AbstractSchema
{
public function build(SchemaConfig $config)
{
$config->getQuery()->addFields([
'latestPost' => new PostType(),
'randomBanner' => [
'type' => new BannerType(),
'resolve' => function () {
return DataProvider::getBanner(rand(1, 10));
}
],
'pageContentUnion' => [
'type' => new ListType(new ContentBlockUnion()),
'resolve' => function () {
return [DataProvider::getPost(1), DataProvider::getBanner(1)];
}
],
'pageContentInterfaced' => [
'type' => new ListType(new ContentBlockInterface()),
'resolve' => function () {
return [DataProvider::getPost(2), DataProvider::getBanner(3)];
}
]
]);
$config->getMutation()->addFields([
'likePost' => new LikePost()
]);
}
}
We've added a pageContentInterfaced
field that have a ListType
with items of ContentBlockInterface
type. Resolve function returns list of one Post
and one Banner
objects.
To test it we'll modify our payload to the following one:
<?php
$payload = '{ pageContentInterface { title} }';
Be aware, because BannerType
doesn't implement ContentBlockInterface
you would get an error:
{ "errors": [ "message": "Type Banner does not implement ContentBlockInterface" } ]}
To fix this we just need to add ContentBlockInterface
by implementing getInterfaces
method and adding the proper field definitions to our BannerType
:
<?php
/**
* BannerType.php
*/
namespace Examples\Blog\Schema;
use Youshido\GraphQL\Type\Config\TypeConfigInterface;
use Youshido\GraphQL\Type\NonNullType;
use Youshido\GraphQL\Type\Object\AbstractObjectType;
use Youshido\GraphQL\Type\Scalar\StringType;
class BannerType extends AbstractObjectType
{
public function build($config)
{
$config
->addField('title', new NonNullType(new StringType()))
->addField('summary', new StringType())
->addField('imageLink', new StringType());
}
public function getInterfaces()
{
return [new ContentBlockInterface()];
}
}
Send the request again and you'll get a nice response with titles of the both Post and Banner:
{
"data": {
"pageContentInterface":[
{"title":"Post 2 title"},
{"title":"Banner 3"}
]
}
}
So far we've been working mostly on the requests that does not require you to send any kind of data other than a simple Int
, but in real life you'll have a lot of requests (mutations) where you'll be sending to server all kind of forms – login, registration, create post and so on.
In order to properly handle and validate that data GraphQL type system provides an InputObjectType
class.
By default all the
Scalar
types are inputs but if you want to have a single more complicated input type you need to extend anInputObjectType
.
Let's develop a PostInputType
that could be used to create a new Post in our system.
<?php
/**
* PostInputType.php
*/
namespace Examples\Blog\Schema;
use Youshido\GraphQL\Type\Config\InputTypeConfigInterface;
use Youshido\GraphQL\Type\NonNullType;
use Youshido\GraphQL\Type\InputObject\AbstractInputObjectType;
use Youshido\GraphQL\Type\Scalar\StringType;
class PostInputType extends AbstractInputObjectType
{
public function build($config)
{
$config
->addField('title', new NonNullType(new StringType()))
->addField('summary', new StringType());
}
}
This InputType
could be used to create a new mutation (we can do it in the BlogSchema::build
for testing):
<?php
// BlogSchema->build() method
$config->getMutation()->addFields([
'likePost' => new LikePost(),
'createPost' => [
'type' => new PostType(),
'args' => [
'post' => new PostInputType(),
'author' => new StringType()
],
'resolve' => function($value, array $args, ResolveInfo $info) {
// code for creating a new post goes here
// we simple use our DataProvider for now
$post = DataProvider::getPost(10);
if (!empty($args['post']['title'])) $post['title'] = $args['post']['title'];
return $post;
}
]
]);
Try to execute the following mutation so you can see the result:
mutation {
createPost(author: "Alex", post: {title: "Hey, this is my new post", summary: "my post" }) {
title
}
}
result:
{"data":{"createPost":{"title":"Hey, this is my new post"}}}
The best way to see the result of your queries/mutations and to inspect the Schema is to use a GraphiQL tool
NonNullType
is really simple to use – consider it as a wrapper that can insure that your field / argument is required and being passed to the resolve function.
We have used NonNullType
couple of times already so we'll just show you useful methods that that could be called on NonNullType
objects:
getNullableType()
getNamedType()
These two can return you a type that was wrapped up in the NonNullType
so you can get it's fields, arguments or name.
It's always a good idea to give your heads up about any possible errors as soon as possible, better on the development stage. For this purpose specifically we made a lot of Abstract classes that will force you to implement the right methods to reduce amount of errors or if you're lucky enough – to have no errors at all.
If you want to implement a new type consider extending the following classes:
- AbstractType
- AbstractScalarType
- AbstractObjectType
- AbstractMutationObjectType
- AbstractInputObjectType
- AbstractInterfaceType
- AbstractEnumType
- AbstractListType
- AbstractUnionType
- AbstractSchemaType
You can create a mutation by extending AbstractObjectType
or by creating a new field of ObjectType
inside your Schema::build
method.
It is crucial for the class to have a getType
method returning the actual OutputType of your mutation but it couldn't be implemented as abstract method, so we created a wrapper class called AbstractMutationObjectType
.
This abstract class can help you to not forget about OutputType
by forcing you to implement a method getOutputType
that will eventually be used by internal getType
method.
This section will be updating on a regular basis with the useful links and references that might help you to quicker become a better GraphQL developer.
To improve our testing experience even more we suggest to start using GraphiQL client, that's included in our examples. It's a JavaScript GraphQL Schema Explorer.
To use it – run the server.sh
from the examples/02_blog/
folder and open the examples/GraphiQL/index.html
file in your browser.
You'll see a nice looking editor that has an autocomplete function and contains all information about your current Schema on the right side in the Docs sidebar: