Revision: Thu, 18 Apr 2024 09:24:13 GMT

Many To Many

A relation of type 'Many To Many' provides a more complex connection with the ability to use an intermediate entity for the connection. The relation requires the through option with similar rules as target.

'Many To Many' relations are, in fact, two relations combined together. This relation requires an intermediate (pivot) entity to connect the source and target entities. Example: many users have many tags, many posts have many favorites.

The relation provides access to an intermediate object on all the steps, including creation, update and query building.

Definition

To define a 'Many To Many' relation using the annotated entities' extension, use (attention, make sure to create pivot entity):

php
use Cycle\Annotated\Annotation\Relation\ManyToMany;
use Cycle\Annotated\Annotation\Entity;

#[Entity]
class User
{
    // ...

    #[ManyToMany(target: Tag::class, through: UserTag::class)]
    protected array $tags;
    
    public function getTags(): array
    {
        return $this->tags;
    }
    
    public function addTag(Tag $tag): void
    {
        $this->tags[] = $tag;
    }
    
    public function removeTag(Tag $tag): void
    {
        $this->tags = array_filter($this->tags, static fn(Tag $t) => $t !== $tag);
    }
}
php
use Cycle\Annotated\Annotation\Column;
use Cycle\Annotated\Annotation\Entity;

#[Entity]
class UserTag
{
    #[Column(type: 'primary')]
    private int $id;
}
php
use Cycle\Annotated\Annotation\Relation\ManyToMany;
use Cycle\Annotated\Annotation\Column;
use Cycle\Annotated\Annotation\Entity;

#[Entity]
class Tag
{
    #[Column(type: 'primary')]
    private int $id;

    #[Column(type: 'string')]
    private string $name;

    public function __construct(string $name)
    {
        $this->name = $name;
    }

    public function getName(): string
    {
        return $this->name;
    }
}

By default, ORM will generate FK and indexes in through entity using the role and primary keys of the linked objects. Following values are available for the configuration:

Option Value Comment
load lazy/eager Relation load approach. Defaults to lazy
cascade bool Automatically save related data with parent entity. Defaults to false
innerKey string Inner key name in source entity. Defaults to a primary key
outerKey string Outer key name in target entity. Defaults to a primary key
throughInnerKey string Key name connected to the innerKey of source entity. Defaults to {sourceRole}_{innerKey}
throughOuterKey string Key name connected to the outerKey of a related entity. Defaults to {targetRole}_{outerKey}
throughWhere array Where conditions applied to through entity
where array Where conditions applied to a related entity
orderBy array Additional sorting rules
fkCreate bool Set to true to automatically create FK on throughInnerKey and throughOuterKey. Defaults to true
fkAction CASCADE, NO ACTION, SET NULL FK onDelete and onUpdate action. Defaults to SET NULL
fkOnDelete CASCADE, NO ACTION, SET NULL FK onDelete action. It has higher priority than {$fkAction}. Defaults to @see {$fkAction}
indexCreate bool Create index on [throughInnerKey, throughOuterKey]. Defaults to true
collection string Collection type that will contain loaded entities. By defaults uses Cycle\ORM\Collection\ArrayCollectionFactory. Read more about relation collections.

You can keep your pivot entity empty, the only requirement is to have defined a primary key.

Usage

To associate two entities using Many To Many relation, use proper way depended on collection type you use. In our example we use default collection factory Cycle\ORM\Collection\ArrayCollectionFactory:

Read more about relation collections here.

php
$user = new User();
$user->setName("Antony");
$user->addTag(new Tag("tag a"));

$manager = new \Cycle\ORM\EntityManager($orm);
$manager->persist($user);
$manager->run();

Disassociation will remove the UserTag entity, and not the Tag entity.

php
$user->removeTag($tag);

Loading

Use the method load of your Select object to preload data of related and pivot entities:

php
$users = $orm->getRepository(User::class)
    ->select()
    ->load('tags')
    ->fetchAll();

Once loaded, you can access the related entity data using the collection:

php
$users = $orm->getRepository(User::class)
    ->select()
    ->load('tags')
    ->fetchAll();

foreach ($users as $u) {
    print_r($u->getTags()->toArray());
}

Accessing Pivot Entity

If you use Cycle\ORM\Collection\DoctrineCollectionFactory for 'Many To Many' relation, you have the ability to access the pivot entity's data using the Cycle\ORM\Collection\Pivoted\PivotedCollectionInterface object. You can do that using the getPivot method:

php
use Cycle\Annotated\Annotation\Relation\ManyToMany;
use Cycle\Annotated\Annotation\Entity;
use Cycle\ORM\Collection\Pivoted\PivotedCollection;

#[Entity]
class User
{
    // ...
    
    #[ManyToMany(target: Tag::class, through: UserTag::class, collection: 'doctrine')]
    public PivotedCollection $tags;
    
    public function __construct() 
    {
        $this->tags = new PivotedCollection();
    }
}
php
$users = $orm->getRepository(User::class)
    ->select()
    ->load('tags')
    ->fetchAll();

foreach ($users as $user) {
    foreach ($user->tags as $tag) {
         print_r($tag);
         print_r($user->tags->getPivot($tag));
    }
}

You can change the values of this entity as they will be persisted with the parent entity. This approach allows you to easier control the association between parent and related entities.

For example, we can add a new property to our UserTag:

php
use Cycle\Annotated\Annotation\Relation\ManyToMany;
use Cycle\Annotated\Annotation\Entity;
use Cycle\Annotated\Annotation\Column;

#[Entity]
class UserTag
{
    #[Column(type: 'primary')]
    private int $id;

    #[Column(type: 'datetime', default: null)]
    private \DateTimeInterface $created_at;

    public function __construct(\DateTimeInterface $d)
    {
        $this->created_at = $d;
    }
}

Now we can assign this entity to the newly created connection:

php
$u = new User();
$u->setName("Antony");

$tag = new Tag("tag a");

$u->tags->add($tag);
$u->tags->setPivot($tag, new UserTag(new \DateTimeImmutable()));

$t->persist($u);
$t->run();

Filtering

Similar to Has Many the entity query can be filtered using the with method:

php
$users = $orm->getRepository(User::class)
    ->select()
    ->distinct()
    ->with('tags')
    ->fetchAll();

You can filter the entity results using the where method on related properties:

php
$users = $orm->getRepository(User::class)
    ->select()
    ->distinct()
    ->where('tags.name', 'tag a')
    ->fetchAll();

Following SQL will be produced:

sql
SELECT DISTINCT `user`.`id`   AS `c0`,
                `user`.`name` AS `c1`
FROM `users` AS `user`
         INNER JOIN `user_tags` AS `user_tags_pivot`
                    ON `user_tags_pivot`.`user_id` = `user`.`id`
         INNER JOIN `tags` AS `user_tags`
                    ON `user_tags`.`id` = `user_tags_pivot`.`tag_id`
WHERE `user_tags`.`name` = 'tag a'

Chain Filtering

Pivot entity data is available for filtering as well, you must use the keyword @ to access it.

php
$hour = new \DateInterval("PT40M");

$users = $orm->getRepository(User::class)
    ->select()
    ->distinct()
    ->where('tags.@.created_at', '>', (new \DateTimeImmutable())->sub($hour))
    ->fetchAll();

You can also load/filter the relations assigned to the pivot entity.

php
$users = $orm->getRepository(User::class)
    ->select()
    ->distinct()
    ->where('tags.@.subRelation.value', $value)
    ->fetchAll();

Cross-database Many To Many relations are not supported yet.

Complex Loading

You can load related data using conditions and sorts applied to the pivot table using the option load.

For example, we can have the following entities:

  • category (id, title)
  • photo (id, url)
  • photo_to_category (photo_id, category_id, position)
php
$categories = $orm->getRepository('category')->select();

We can now load categories with photos inside them ordered by photo_to_category position using a WHERE IN or JOIN query:

php
$result = $categories->load('photos', [
    'load' => function (\Cycle\ORM\Select\QueryBuilder $q) {
        $q->orderBy('@.@.position'); // @ current relation (photos), @.@ current relation pivot (photo_to_category)
    }
])->fetchAll();

The produced SQL:

sql
SELECT "category"."id"    AS "c0",
       "category"."title" AS "c1"
FROM "categories" AS "category"

SQL #2:

sql
SELECT "l_category_photos_pivot"."id"          AS "c0",
       "l_category_photos_pivot"."position"    AS "c1",
       "l_category_photos_pivot"."photo_id"    AS "c2",
       "l_category_photos_pivot"."category_id" AS "c3",
       "category_photos"."id"                  AS "c4",
       "category_photos"."url"                 AS "c5"
FROM "photos" AS "category_photos"
         INNER JOIN "photo_category_positions" AS "l_category_photos_pivot"
                    ON "l_category_photos_pivot"."photo_id" = "category_photos"."id"
WHERE "l_category_photos_pivot"."category_id" IN (1, 2, 3, 4)
ORDER BY "l_category_photos_pivot"."position" ASC

We can force the ORM to use a single query to pull the data (useful for more complex conditions):

php
$result = $categories->load('photos', [
     'method' => \Cycle\ORM\Select::SINGLE_QUERY,
     'load'   => function (\Cycle\ORM\Select\QueryBuilder $q) {
         $q->orderBy('@.@.position');  // @ current relation (photos), @.@ current relation pivot (photo_to_category)
     }
])->orderBy('id')->fetchAll();

SQL:

sql
SELECT "category"."id"                           AS "c0",
       "category"."title"                        AS "c1",
       "l_l_category_photos_pivot"."id"          AS "c2",
       "l_l_category_photos_pivot"."position"    AS "c3",
       "l_l_category_photos_pivot"."photo_id"    AS "c4",
       "l_l_category_photos_pivot"."category_id" AS "c5",
       "l_category_photos"."id"                  AS "c6",
       "l_category_photos"."url"                 AS "c7"
FROM "categories" AS "category"
         LEFT JOIN "photo_category_positions" AS "l_l_category_photos_pivot"
                   ON "l_l_category_photos_pivot"."category_id" = "category"."id"
         INNER JOIN "photos" AS "l_category_photos"
                    ON "l_category_photos"."id" = "l_l_category_photos_pivot"."photo_id"
ORDER BY "category"."id" ASC, "l_l_category_photos_pivot"."position" ASC
Edit this page