Association Mapping =================== This chapter introduces association mappings which are used to explain references between objects and are mapped to a relational database using foreign keys. Instead of working with the foreign keys directly you will always work with references to objects: - A reference to a single object is represented by a foreign key. - A collection of objects is represented by many foreign keys pointing to the object holding the collection This chapter is split into three different sections. - A list of all the possible association mapping use-cases is given. - :ref:`association_mapping_defaults` are explained that simplify the use-case examples. - :ref:`collections` are introduced that contain entities in assoactions. To master assocations you should also learn about :doc:`owning and inverse sides of associations ` One-To-One, Unidirectional -------------------------- A unidirectional one-to-one association is very common. Here is an example of a ``Product`` that has one ``Shipping`` object associated to it. The ``Shipping`` side does not reference back to the ``Product`` so it is unidirectional. .. configuration-block:: .. code-block:: php .. code-block:: yaml Product: type: entity oneToOne: shipping: targetEntity: Shipping joinColumn: name: shipping_id referencedColumnName: id Note that the @JoinColumn is not really necessary in this example, as the defaults would be the same. Generated MySQL Schema: .. code-block:: sql CREATE TABLE Product ( id INT AUTO_INCREMENT NOT NULL, shipping_id INT DEFAULT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; CREATE TABLE Shipping ( id INT AUTO_INCREMENT NOT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; ALTER TABLE Product ADD FOREIGN KEY (shipping_id) REFERENCES Shipping(id); One-To-One, Bidirectional ------------------------- Here is a one-to-one relationship between a ``Customer`` and a ``Cart``. The ``Cart`` has a reference back to the ``Customer`` so it is bidirectional. .. configuration-block:: .. code-block:: php .. code-block:: yaml Customer: oneToOne: cart: targetEntity: Cart mappedBy: customer Cart: oneToOne: customer: targetEntity: Customer inversedBy: cart joinColumn: name: customer_id referencedColumnName: id Note that the @JoinColumn is not really necessary in this example, as the defaults would be the same. Generated MySQL Schema: .. code-block:: sql CREATE TABLE Cart ( id INT AUTO_INCREMENT NOT NULL, customer_id INT DEFAULT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; CREATE TABLE Customer ( id INT AUTO_INCREMENT NOT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; ALTER TABLE Cart ADD FOREIGN KEY (customer_id) REFERENCES Customer(id); See how the foreign key is defined on the owning side of the relation, the table ``Cart``. One-To-One, Self-referencing ---------------------------- You can easily have self referencing one-to-one relationships like below. .. code-block:: php phonenumbers = new \Doctrine\Common\Collections\ArrayCollection(); } // ... } /** @Entity **/ class Phonenumber { // ... } .. code-block:: xml .. code-block:: yaml User: type: entity manyToMany: phonenumbers: targetEntity: Phonenumber joinTable: name: users_phonenumbers joinColumns: user_id: referencedColumnName: id inverseJoinColumns phonenumber_id: referencedColumnName: id unique: true .. note:: One-To-Many uni-directional relations with join-table only work using the @ManyToMany annotation and a unique-constraint. Generates the following MySQL Schema: .. code-block:: sql CREATE TABLE User ( id INT AUTO_INCREMENT NOT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; CREATE TABLE users_phonenumbers ( user_id INT NOT NULL, phonenumber_id INT NOT NULL, UNIQUE INDEX users_phonenumbers_phonenumber_id_uniq (phonenumber_id), PRIMARY KEY(user_id, phonenumber_id) ) ENGINE = InnoDB; CREATE TABLE Phonenumber ( id INT AUTO_INCREMENT NOT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; ALTER TABLE users_phonenumbers ADD FOREIGN KEY (user_id) REFERENCES User(id); ALTER TABLE users_phonenumbers ADD FOREIGN KEY (phonenumber_id) REFERENCES Phonenumber(id); Many-To-One, Unidirectional --------------------------- You can easily implement a many-to-one unidirectional association with the following: .. configuration-block:: .. code-block:: php .. code-block:: yaml User: type: entity manyToOne: address: targetEntity: Address .. note:: The above ``@JoinColumn`` is optional as it would default to ``address_id`` and ``id`` anyways. You can omit it and let it use the defaults. Generated MySQL Schema: .. code-block:: sql CREATE TABLE User ( id INT AUTO_INCREMENT NOT NULL, address_id INT DEFAULT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; CREATE TABLE Address ( id INT AUTO_INCREMENT NOT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; ALTER TABLE User ADD FOREIGN KEY (address_id) REFERENCES Address(id); One-To-Many, Bidirectional -------------------------- Bidirectional one-to-many associations are very common. The following code shows an example with a Product and a Feature class: .. configuration-block:: .. code-block:: php features = new \Doctrine\Common\Collections\ArrayCollection(); } } /** @Entity **/ class Feature { // ... /** * @ManyToOne(targetEntity="Product", inversedBy="features") * @JoinColumn(name="product_id", referencedColumnName="id") **/ private $product; // ... } .. code-block:: xml Note that the @JoinColumn is not really necessary in this example, as the defaults would be the same. Generated MySQL Schema: .. code-block:: sql CREATE TABLE Product ( id INT AUTO_INCREMENT NOT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; CREATE TABLE Feature ( id INT AUTO_INCREMENT NOT NULL, product_id INT DEFAULT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; ALTER TABLE Feature ADD FOREIGN KEY (product_id) REFERENCES Product(id); One-To-Many, Self-referencing ----------------------------- You can also setup a one-to-many association that is self-referencing. In this example we setup a hierarchy of ``Category`` objects by creating a self referencing relationship. This effectively models a hierarchy of categories and from the database perspective is known as an adjacency list approach. .. configuration-block:: .. code-block:: php children = new \Doctrine\Common\Collections\ArrayCollection(); } } .. code-block:: xml .. code-block:: yaml Category: type: entity oneToMany: children: targetEntity: Category mappedBy: parent manyToOne: parent: targetEntity: Category inversedBy: children Note that the @JoinColumn is not really necessary in this example, as the defaults would be the same. Generated MySQL Schema: .. code-block:: sql CREATE TABLE Category ( id INT AUTO_INCREMENT NOT NULL, parent_id INT DEFAULT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; ALTER TABLE Category ADD FOREIGN KEY (parent_id) REFERENCES Category(id); Many-To-Many, Unidirectional ---------------------------- Real many-to-many associations are less common. The following example shows a unidirectional association between User and Group entities: .. configuration-block:: .. code-block:: php groups = new \Doctrine\Common\Collections\ArrayCollection(); } } /** @Entity **/ class Group { // ... } .. code-block:: xml .. code-block:: yaml User: type: entity manyToMany: groups: targetEntity: Group joinTable: name: users_groups joinColumns: user_id: referencedColumnName: id inverseJoinColumns: group_id: referencedColumnName: id Generated MySQL Schema: .. code-block:: sql CREATE TABLE User ( id INT AUTO_INCREMENT NOT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; CREATE TABLE users_groups ( user_id INT NOT NULL, group_id INT NOT NULL, PRIMARY KEY(user_id, group_id) ) ENGINE = InnoDB; CREATE TABLE Group ( id INT AUTO_INCREMENT NOT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; ALTER TABLE users_groups ADD FOREIGN KEY (user_id) REFERENCES User(id); ALTER TABLE users_groups ADD FOREIGN KEY (group_id) REFERENCES Group(id); .. note:: Why are many-to-many associations less common? Because frequently you want to associate additional attributes with an association, in which case you introduce an association class. Consequently, the direct many-to-many association disappears and is replaced by one-to-many/many-to-one associations between the 3 participating classes. Many-To-Many, Bidirectional --------------------------- Here is a similar many-to-many relationship as above except this one is bidirectional. .. configuration-block:: .. code-block:: php groups = new \Doctrine\Common\Collections\ArrayCollection(); } // ... } /** @Entity **/ class Group { // ... /** * @ManyToMany(targetEntity="User", mappedBy="groups") **/ private $users; public function __construct() { $this->users = new \Doctrine\Common\Collections\ArrayCollection(); } // ... } .. code-block:: xml .. code-block:: yaml User: type: entity manyToMany: groups: targetEntity: Group inversedBy: users joinTable: name: users_groups joinColumns: user_id: referencedColumnName: id inverseJoinColumns: group_id: referencedColumnName: id Group: type: entity manyToMany: users: targetEntity: User mappedBy: groups The MySQL schema is exactly the same as for the Many-To-Many uni-directional case above. Picking Owning and Inverse Side ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ For Many-To-Many associations you can chose which entity is the owning and which the inverse side. There is a very simple semantic rule to decide which side is more suitable to be the owning side from a developers perspective. You only have to ask yourself, which entity is responsible for the connection management and pick that as the owning side. Take an example of two entities ``Article`` and ``Tag``. Whenever you want to connect an Article to a Tag and vice-versa, it is mostly the Article that is responsible for this relation. Whenever you add a new article, you want to connect it with existing or new tags. Your create Article form will probably support this notion and allow to specify the tags directly. This is why you should pick the Article as owning side, as it makes the code more understandable: .. code-block:: php addArticle($this); // synchronously updating inverse side $this->tags[] = $tag; } } class Tag { private $articles; public function addArticle(Article $article) { $this->articles[] = $article; } } This allows to group the tag adding on the ``Article`` side of the association: .. code-block:: php addTag($tagA); $article->addTag($tagB); Many-To-Many, Self-referencing ------------------------------ You can even have a self-referencing many-to-many association. A common scenario is where a ``User`` has friends and the target entity of that relationship is a ``User`` so it is self referencing. In this example it is bidirectional so ``User`` has a field named ``$friendsWithMe`` and ``$myFriends``. .. code-block:: php friendsWithMe = new \Doctrine\Common\Collections\ArrayCollection(); $this->myFriends = new \Doctrine\Common\Collections\ArrayCollection(); } // ... } Generated MySQL Schema: .. code-block:: sql CREATE TABLE User ( id INT AUTO_INCREMENT NOT NULL, PRIMARY KEY(id) ) ENGINE = InnoDB; CREATE TABLE friends ( user_id INT NOT NULL, friend_user_id INT NOT NULL, PRIMARY KEY(user_id, friend_user_id) ) ENGINE = InnoDB; ALTER TABLE friends ADD FOREIGN KEY (user_id) REFERENCES User(id); ALTER TABLE friends ADD FOREIGN KEY (friend_user_id) REFERENCES User(id); .. _association_mapping_defaults: Mapping Defaults ---------------- Before we introduce all the association mappings in detail, you should note that the @JoinColumn and @JoinTable definitions are usually optional and have sensible default values. The defaults for a join column in a one-to-one/many-to-one association is as follows: :: name: "_id" referencedColumnName: "id" As an example, consider this mapping: .. configuration-block:: .. code-block:: php .. code-block:: yaml Product: type: entity oneToOne: shipping: targetEntity: Shipping This is essentially the same as the following, more verbose, mapping: .. configuration-block:: .. code-block:: php .. code-block:: yaml Product: type: entity oneToOne: shipping: targetEntity: Shipping joinColumn: name: shipping_id referencedColumnName: id The @JoinTable definition used for many-to-many mappings has similar defaults. As an example, consider this mapping: .. configuration-block:: .. code-block:: php .. code-block:: yaml User: type: entity manyToMany: groups: targetEntity: Group This is essentially the same as the following, more verbose, mapping: .. configuration-block:: .. code-block:: php .. code-block:: yaml User: type: entity manyToMany: groups: targetEntity: Group joinTable: name: User_Group joinColumns: User_id: referencedColumnName: id inverseJoinColumns: Group_id: referencedColumnName: id In that case, the name of the join table defaults to a combination of the simple, unqualified class names of the participating classes, separated by an underscore character. The names of the join columns default to the simple, unqualified class name of the targeted class followed by "\_id". The referencedColumnName always defaults to "id", just as in one-to-one or many-to-one mappings. If you accept these defaults, you can reduce the mapping code to a minimum. .. _collections: Collections ----------- In all the examples of many-valued associations in this manual we will make use of a ``Collection`` interface and a corresponding default implementation ``ArrayCollection`` that are defined in the ``Doctrine\Common\Collections`` namespace. Why do we need that? Doesn't that couple my domain model to Doctrine? Unfortunately, PHP arrays, while being great for many things, do not make up for good collections of business objects, especially not in the context of an ORM. The reason is that plain PHP arrays can not be transparently extended / instrumented in PHP code, which is necessary for a lot of advanced ORM features. The classes / interfaces that come closest to an OO collection are ArrayAccess and ArrayObject but until instances of these types can be used in all places where a plain array can be used (something that may happen in PHP6) their usability is fairly limited. You "can" type-hint on ``ArrayAccess`` instead of ``Collection``, since the Collection interface extends ``ArrayAccess``, but this will severely limit you in the way you can work with the collection, because the ``ArrayAccess`` API is (intentionally) very primitive and more importantly because you can not pass this collection to all the useful PHP array functions, which makes it very hard to work with. .. warning:: The Collection interface and ArrayCollection class, like everything else in the Doctrine namespace, are neither part of the ORM, nor the DBAL, it is a plain PHP class that has no outside dependencies apart from dependencies on PHP itself (and the SPL). Therefore using this class in your domain classes and elsewhere does not introduce a coupling to the persistence layer. The Collection class, like everything else in the Common namespace, is not part of the persistence layer. You could even copy that class over to your project if you want to remove Doctrine from your project and all your domain classes will work the same as before. Initializing Collections ------------------------ You have to be careful when using entity fields that contain a collection of related entities. Say we have a User entity that contains a collection of groups: .. code-block:: php groups; } } With this code alone the ``$groups`` field only contains an instance of ``Doctrine\Common\Collections\Collection`` if the user is retrieved from Doctrine, however not after you instantiated a fresh instance of the User. When your user entity is still new ``$groups`` will obviously be null. This is why we recommend to initialize all collection fields to an empty ``ArrayCollection`` in your entities constructor: .. code-block:: php groups = new ArrayCollection(); } public function getGroups() { return $this->groups; } } Now the following code will be working even if the Entity hasn't been associated with an EntityManager yet: .. code-block:: php find('Group', $groupId); $user = new User(); $user->getGroups()->add($group); Runtime vs Development Mapping Validation ----------------------------------------- For performance reasons Doctrine 2 has to skip some of the necessary validation of association mappings. You have to execute this validation in your development workflow to verify the associations are correctly defined. You can either use the Doctrine Command Line Tool: .. code-block:: php doctrine orm:validate-schema Or you can trigger the validation manually: .. code-block:: php validateMapping(); if (count($errors) > 0) { // Lots of errors! echo implode("\n\n", $errors); } If the mapping is invalid the errors array contains a positive number of elements with error messages. .. warning:: One mapping option that is not validated is the use of the referenced column name. It has to point to the equivalent primary key otherwise Doctrine will not work. .. note:: One common error is to use a backlash in front of the fully-qualified class-name. Whenever a FQCN is represented inside a string (such as in your mapping definitions) you have to drop the prefix backslash. PHP does this with ``get_class()`` or Reflection methods for backwards compatibility reasons.