25 KiB

Raw Blame History

Active Record

Active Record implements the Active Record design pattern. The premise behind Active Record is that an individual ActiveRecord object is associated with a specific row in a database table. The object's attributes are mapped to the columns of the corresponding table. Referencing an Active Record attribute is equivalent to accessing the corresponding table column for that record.

As an example, say that the Customer ActiveRecord class is associated with the tbl_customer table. This would mean that the class's name attribute is automatically mapped to the name column in tbl_customer. Thanks to Active Record, assuming the variable $customer is an object of type Customer, to get the value of the name column for the table row, you can use the expression $customer->name. In this example, Active Record is providing an object-oriented interface for accessing data stored in the database. But Active Record provides much more functionality than this.

With Active Record, instead of writing raw SQL statements to perform database queries, you can call intuitive methods to achieve the same goals. For example, calling ActiveRecord::save() would perform an INSERT or UPDATE query, creating or updating a row in the associated table of the ActiveRecord class:

$customer = new Customer();
$customer->name = 'Qiang';
$customer->save();  // a new row is inserted into tbl_customer

Declaring ActiveRecord Classes

To declare an ActiveRecord class you need to extend \yii\db\ActiveRecord and implement the tableName method:

use yii\db\ActiveRecord;

class Customer extends ActiveRecord
{
	/**
	 * @return string the name of the table associated with this ActiveRecord class.
	 */
	public static function tableName()
	{
		return 'tbl_customer';
	}
}

The tableName method only has to return the name of the database table associated with the class.

Class instances are obtained in one of two ways:

Using the new operator to create a new, empty object
Using a method to fetch an existing record (or records) from the database

Connecting to the Database

ActiveRecord relies on a Connection to perform the underlying DB operations. By default, ActiveRecord assumes that there is an application component named db which provides the needed Connection instance. Usually this component is configured in application configuration file:

return [
	'components' => [
		'db' => [
			'class' => 'yii\db\Connection',
			'dsn' => 'mysql:host=localhost;dbname=testdb',
			'username' => 'demo',
			'password' => 'demo',
		],
	],
];

Please read the Database basics section to learn more on how to configure and use database connections.

Querying Data from the Database

There are two ActiveRecord methods for querying data from database:

ActiveRecord::find()
ActiveRecord::findBySql()

Both methods return an ActiveQuery instance, which extends Query, and thus supports the same set of flexible and powerful DB query methods. The following examples demonstrate some of the possibilities.

// to retrieve all *active* customers and order them by their ID:
$customers = Customer::find()
	->where(['status' => $active])
	->orderBy('id')
	->all();

// to return a single customer whose ID is 1:
$customer = Customer::find(1);

// the above code is equivalent to the following:
$customer = Customer::find()
	->where(['id' => 1])
	->one();

// to retrieve customers using a raw SQL statement:
$sql = 'SELECT * FROM tbl_customer';
$customers = Customer::findBySql($sql)->all();

// to return the number of *active* customers:
$count = Customer::find()
	->where(['status' => $active])
	->count();

// to return customers in terms of arrays rather than `Customer` objects:
$customers = Customer::find()
	->asArray()
	->all();
// each element of $customers is an array of name-value pairs

// to index the result by customer IDs:
$customers = Customer::find()->indexBy('id')->all();
// $customers array is indexed by customer IDs

Accessing Column Data

ActiveRecord maps each column of the corresponding database table row to an attribute in the ActiveRecord object. The attribute behaves like any regular object public property. The attribute's name will be the same as the corresponding column name, and is case-sensitive.

To read the value of a column, you can use the following syntax:

// "id" and "email" are the names of columns in the table associated with $customer ActiveRecord object
$id = $customer->id;
$email = $customer->email;

To change the value of a column, assign a new value to the associated property and save the object:

$customer->email = 'jane@example.com';
$customer->save();

Manipulating Data in the Database

ActiveRecord provides the following methods to insert, update and delete data in the database:

Note that ActiveRecord::updateAll(), ActiveRecord::updateAllCounters() and ActiveRecord::deleteAll() are static methods that apply to the whole database table. The other methods only apply to the row associated with the ActiveRecord object through which the method is being called.

// to insert a new customer record
$customer = new Customer;
$customer->name = 'James';
$customer->email = 'james@example.com';
$customer->save();  // equivalent to $customer->insert();

// to update an existing customer record
$customer = Customer::find($id);
$customer->email = 'james@example.com';
$customer->save();  // equivalent to $customer->update();

// to delete an existing customer record
$customer = Customer::find($id);
$customer->delete();

// to increment the age of ALL customers by 1
Customer::updateAllCounters(['age' => 1]);

Info: The save() method will either perform an INSERT or UPDATE SQL statement, depending on whether the ActiveRecord being saved is new or not by checking ActiveRecord::isNewRecord.

Data Input and Validation

ActiveRecord inherits data validation and data input features from \yii\base\Model. Data validation is called automatically when save() is performed. If data validation fails, the saving operation will be cancelled.

For more details refer to the Model section of this guide.

Querying Relational Data

You can use ActiveRecord to also query a table's relational data (i.e., selection of data from Table A can also pull in related data from Table B). Thanks to ActiveRecord, the relational data returned can be accessed like a property of the ActiveRecord object associated with the primary table.

For example, with an appropriate relation declaration, by accessing $customer->orders you may obtain an array of Order objects which represent the orders placed by the specified customer.

To declare a relation, define a getter method which returns an ActiveRelation object. For example,

class Customer extends \yii\db\ActiveRecord
{
	public function getOrders()
	{
		return $this->hasMany(Order::className(), ['customer_id' => 'id']);
	}
}

class Order extends \yii\db\ActiveRecord
{
	public function getCustomer()
	{
		return $this->hasOne(Customer::className(), ['id' => 'customer_id']);
	}
}

The methods ActiveRecord::hasMany() and ActiveRecord::hasOne() used in the above are used to model the many-one relationship and one-one relationship in a relational database. For example, a customer has many orders, and an order has one customer. Both methods take two parameters and return an ActiveRelation object:

$class: the name of the class of the related model(s). This should be a fully qualified class name.
$link: the association between columns from the two tables. This should be given as an array. The keys of the array are the names of the columns from the table associated with $class, while the values of the array are the names of the columns from the declaring class. It is a good practice to define relationships based on table foreign keys.

After declaring relations, getting relational data is as easy as accessing a component property that is defined by the corresponding getter method:

// get the orders of a customer
$customer = Customer::find(1);
$orders = $customer->orders;  // $orders is an array of Order objects

Behind the scene, the above code executes the following two SQL queries, one for each line of code:

SELECT * FROM tbl_customer WHERE id=1;
SELECT * FROM tbl_order WHERE customer_id=1;

Tip: If you access the expression $customer->orders again, will it perform the second SQL query again? Nope. The SQL query is only performed the first time when this expression is accessed. Any further accesses will only return the previously fetched results that are cached internally. If you want to re-query the relational data, simply unset the existing one first: unset($customer->orders);.

Sometimes, you may want to pass parameters to a relational query. For example, instead of returning all orders of a customer, you may want to return only big orders whose subtotal exceeds a specified amount. To do so, declare a bigOrders relation with the following getter method:

class Customer extends \yii\db\ActiveRecord
{
	public function getBigOrders($threshold = 100)
	{
		return $this->hasMany(Order::className(), ['customer_id' => 'id'])
			->where('subtotal > :threshold', [':threshold' => $threshold])
			->orderBy('id');
	}
}

Remember that hasMany() returns an ActiveRelation object which extends from ActiveQuery and thus supports the same set of querying methods as ActiveQuery.

With the above declaration, if you access $customer->bigOrders, it will only return the orders whose subtotal is greater than 100. To specify a different threshold value, use the following code:

$orders = $customer->getBigOrders(200)->all();

Relations with Pivot Table

Sometimes, two tables are related together via an intermediary table called pivot table. To declare such relations, we can customize the ActiveRelation object by calling its ActiveRelation::via() or ActiveRelation::viaTable() method.

For example, if table tbl_order and table tbl_item are related via pivot table tbl_order_item, we can declare the items relation in the Order class like the following:

class Order extends \yii\db\ActiveRecord
{
	public function getItems()
	{
		return $this->hasMany(Item::className(), ['id' => 'item_id'])
			->viaTable('tbl_order_item', ['order_id' => 'id']);
	}
}

ActiveRelation::via() method is similar to ActiveRelation::viaTable() except that the first parameter of ActiveRelation::via() takes a relation name declared in the ActiveRecord class instead of the pivot table name. For example, the above items relation can be equivalently declared as follows:

class Order extends \yii\db\ActiveRecord
{
	public function getOrderItems()
	{
		return $this->hasMany(OrderItem::className(), ['order_id' => 'id']);
	}

	public function getItems()
	{
		return $this->hasMany(Item::className(), ['id' => 'item_id'])
			->via('orderItems');
	}
}

Lazy and Eager Loading

As described earlier, when you access the related objects the first time, ActiveRecord will perform a DB query to retrieve the corresponding data and populate it into the related objects. No query will be performed if you access the same related objects again. We call this lazy loading. For example,

// SQL executed: SELECT * FROM tbl_customer WHERE id=1
$customer = Customer::find(1);
// SQL executed: SELECT * FROM tbl_order WHERE customer_id=1
$orders = $customer->orders;
// no SQL executed
$orders2 = $customer->orders;

Lazy loading is very convenient to use. However, it may suffer from a performance issue in the following scenario:

// SQL executed: SELECT * FROM tbl_customer LIMIT 100
$customers = Customer::find()->limit(100)->all();

foreach ($customers as $customer) {
	// SQL executed: SELECT * FROM tbl_order WHERE customer_id=...
	$orders = $customer->orders;
	// ...handle $orders...
}

How many SQL queries will be performed in the above code, assuming there are more than 100 customers in the database? 101! The first SQL query brings back 100 customers. Then for each customer, a SQL query is performed to bring back the orders of that customer.

To solve the above performance problem, you can use the so-called eager loading approach by calling ActiveQuery::with():

// SQL executed: SELECT * FROM tbl_customer LIMIT 100;
//               SELECT * FROM tbl_orders WHERE customer_id IN (1,2,...)
$customers = Customer::find()->limit(100)
	->with('orders')->all();

foreach ($customers as $customer) {
	// no SQL executed
	$orders = $customer->orders;
	// ...handle $orders...
}

As you can see, only two SQL queries are needed for the same task.

Sometimes, you may want to customize the relational queries on the fly. This can be done for both lazy loading and eager loading. For example,

$customer = Customer::find(1);
// lazy loading: SELECT * FROM tbl_order WHERE customer_id=1 AND subtotal>100
$orders = $customer->getOrders()->where('subtotal>100')->all();

// eager loading: SELECT * FROM tbl_customer LIMIT 100
//                SELECT * FROM tbl_order WHERE customer_id IN (1,2,...) AND subtotal>100
$customers = Customer::find()->limit(100)->with([
	'orders' => function($query) {
		$query->andWhere('subtotal>100');
	},
])->all();

Joining with Relations

When working with relational databases, a common task is to join multiple tables and apply various query conditions and parameters to the JOIN SQL statement. Instead of calling ActiveQuery::join() explicitly to build up the JOIN query, you may reuse the existing relation definitions and call ActiveQuery::joinWith() to achieve this goal. For example,

// find all orders and sort the orders by the customer id and the order id. also eager loading "customer"
$orders = Order::find()->joinWith('customer')->orderBy('tbl_customer.id, tbl_order.id')->all();
// find all orders that contain books, and eager loading "books"
$orders = Order::find()->innerJoinWith('books')->all();

In the above, the method ActiveQuery::innerJoinWith() is a shortcut to ActiveQuery::joinWith() with the join type set as INNER JOIN.

You may join with one or multiple relations; you may apply query conditions to the relations on-the-fly; and you may also join with sub-relations. For example,

// join with multiple relations
// find out the orders that contain books and are placed by customers who registered within the past 24 hours
$orders = Order::find()->innerJoinWith([
	'books',
	'customer' => function ($query) {
		$query->where('tbl_customer.create_time > ' . (time() - 24 * 3600));
	}
])->all();
// join with sub-relations: join with books and books' authors
$orders = Order::find()->joinWith('books.author')->all();

Behind the scene, Yii will first execute a JOIN SQL statement to bring back the primary models satisfying the conditions applied to the JOIN SQL. It will then execute a query for each relation and populate the corresponding related records.

The difference between ActiveQuery::joinWith() and ActiveQuery::with() is that the former joins the tables for the primary model class and the related model classes to retrieve the primary models, while the latter just queries against the table for the primary model class to retrieve the primary models.

Because of this difference, you may apply query conditions that are only available to a JOIN SQL statement. For example, you may filter the primary models by the conditions on the related models, like the example above. You may also sort the primary models using columns from the related tables.

When using ActiveQuery::joinWith(), you are responsible to disambiguate column names. In the above examples, we use tbl_item.id and tbl_order.id to disambiguate the id column references because both of the order table and the item table contain a column named id.

By default, when you join with a relation, the relation will also be eagerly loaded. You may change this behavior by passing the $eagerLoading parameter which specifies whether to eager load the specified relations.

And also by default, ActiveQuery::joinWith() uses LEFT JOIN to join the related tables. You may pass it with the $joinType parameter to customize the join type. As a shortcut to the INNER JOIN type, you may use ActiveQuery::innerJoinWith().

Below are some more examples,

// find all orders that contain books, but do not eager loading "books".
$orders = Order::find()->innerJoinWith('books', false)->all();
// which is equivalent to the above
$orders = Order::find()->joinWith('books', false, 'INNER JOIN')->all();

Sometimes when joining two tables, you may need to specify some extra condition in the ON part of the JOIN query. This can be done by calling the \yii\db\ActiveRelation::onCondition() method like the following:

class User extends ActiveRecord
{
	public function getBooks()
	{
		return $this->hasMany(Item::className(), ['owner_id' => 'id'])->onCondition(['category_id' => 1]);
	}
}

In the above, the hasMany() method returns an ActiveRelation instance, upon which onCondition() is called to specify that only items whose category_id is 1 should be returned.

When you perform query using ActiveQuery::joinWith(), the on-condition will be put in the ON part of the corresponding JOIN query. For example,

// SELECT tbl_user.* FROM tbl_user LEFT JOIN tbl_item ON tbl_item.owner_id=tbl_user.id AND category_id=1
// SELECT * FROM tbl_item WHERE owner_id IN (...) AND category_id=1
$users = User::model()->joinWith('books')->all();

Note that if you use eager loading via ActiveQuery::with() or lazy loading, the on-condition will be put in the WHERE part of the corresponding SQL statement, because there is no JOIN query involved. For example,

// SELECT * FROM tbl_user WHERE id=10
$user = User::model(10);
// SELECT * FROM tbl_item WHERE owner_id=10 AND category_id=1
$books = $user->books;

Working with Relationships

ActiveRecord provides the following two methods for establishing and breaking a relationship between two ActiveRecord objects:

For example, given a customer and a new order, we can use the following code to make the order owned by the customer:

$customer = Customer::find(1);
$order = new Order;
$order->subtotal = 100;
$customer->link('orders', $order);

The link() call above will set the customer_id of the order to be the primary key value of $customer and then call save() to save the order into database.

Life Cycles of an ActiveRecord Object

An ActiveRecord object undergoes different life cycles when it is used in different cases. Subclasses or ActiveRecord behaviors may "inject" custom code in these life cycles through method overriding and event handling mechanisms.

When instantiating a new ActiveRecord instance, we will have the following life cycles:

constructor
init(): will trigger an EVENT_INIT event

When getting an ActiveRecord instance through the find() method, we will have the following life cycles:

constructor
init(): will trigger an EVENT_INIT event
afterFind(): will trigger an EVENT_AFTER_FIND event

When calling save() to insert or update an ActiveRecord, we will have the following life cycles:

beforeValidate(): will trigger an EVENT_BEFORE_VALIDATE event
afterValidate(): will trigger an EVENT_AFTER_VALIDATE event
beforeSave(): will trigger an EVENT_BEFORE_INSERT or EVENT_BEFORE_UPDATE event
perform the actual data insertion or updating
afterSave(): will trigger an EVENT_AFTER_INSERT or EVENT_AFTER_UPDATE event

Finally when calling delete() to delete an ActiveRecord, we will have the following life cycles:

beforeDelete(): will trigger an EVENT_BEFORE_DELETE event
perform the actual data deletion
afterDelete(): will trigger an EVENT_AFTER_DELETE event

Scopes

A scope is a method that customizes a given ActiveQuery object. Scope methods are static and are defined in the ActiveRecord classes. They can be invoked through the ActiveQuery object that is created via find() or findBySql(). The following is an example:

class Comment extends \yii\db\ActiveRecord
{
	// ...

	/**
	 * @param ActiveQuery $query
	 */
	public static function active($query)
	{
		$query->andWhere('status = 1');
	}
}

$comments = Comment::find()->active()->all();

In the above, the active() method is defined in Comment while we are calling it through ActiveQuery returned by Comment::find().

You can also use scopes when defining relations. For example,

class Post extends \yii\db\ActiveRecord
{
	public function getComments()
	{
		return $this->hasMany(Comment::className(), ['post_id' => 'id'])->active();

	}
}

Or use the scopes on-the-fly when performing relational query:

$posts = Post::find()->with([
	'comments' => function($q) {
		$q->active();
	}
])->all();

Scopes can be parameterized. For example, we can define and use the following olderThan scope:

class Customer extends \yii\db\ActiveRecord
{
	// ...

	/**
	 * @param ActiveQuery $query
	 * @param integer $age
	 */
	public static function olderThan($query, $age = 30)
	{
		$query->andWhere('age > :age', [':age' => $age]);
	}
}

$customers = Customer::find()->olderThan(50)->all();

The parameters should follow after the $query parameter when defining the scope method, and they can take default values like shown above.

Transactional operations

When a few DB operations are related and are executed

TODO: FIXME: WIP, TBD, https://github.com/yiisoft/yii2/issues/226

, afterSave(), beforeDelete() and/or afterDelete() life cycle methods. Developer may come to the solution of overriding ActiveRecord save() method with database transaction wrapping or even using transaction in controller action, which is strictly speaking doesn't seems to be a good practice (recall skinny-controller fat-model fundamental rule).

Here these ways are (DO NOT use them unless you're sure what are you actually doing). Models:

class Feature extends \yii\db\ActiveRecord
{
	// ...

	public function getProduct()
	{
		return $this->hasOne(Product::className(), ['product_id' => 'id']);
	}
}

class Product extends \yii\db\ActiveRecord
{
	// ...

	public function getFeatures()
	{
		return $this->hasMany(Feature::className(), ['id' => 'product_id']);
	}
}

Overriding save() method:


class ProductController extends \yii\web\Controller
{
	public function actionCreate()
	{
		// FIXME: TODO: WIP, TBD
	}
}

Using transactions within controller layer:

class ProductController extends \yii\web\Controller
{
	public function actionCreate()
	{
		// FIXME: TODO: WIP, TBD
	}
}

Instead of using these fragile methods you should consider using atomic scenarios and operations feature.

class Feature extends \yii\db\ActiveRecord
{
	// ...

	public function getProduct()
	{
		return $this->hasOne(Product::className(), ['product_id' => 'id']);
	}

	public function scenarios()
	{
		return [
			'userCreates' => [
				'attributes' => ['name', 'value'],
				'atomic' => [self::OP_INSERT],
			],
		];
	}
}

class Product extends \yii\db\ActiveRecord
{
	// ...

	public function getFeatures()
	{
		return $this->hasMany(Feature::className(), ['id' => 'product_id']);
	}

	public function scenarios()
	{
		return [
			'userCreates' => [
				'attributes' => ['title', 'price'],
				'atomic' => [self::OP_INSERT],
			],
		];
	}

	public function afterValidate()
	{
		parent::afterValidate();
		// FIXME: TODO: WIP, TBD
	}

	public function afterSave($insert)
	{
		parent::afterSave($insert);
		if ($this->getScenario() === 'userCreates') {
			// FIXME: TODO: WIP, TBD
		}
	}
}

Controller is very thin and neat:

class ProductController extends \yii\web\Controller
{
	public function actionCreate()
	{
		// FIXME: TODO: WIP, TBD
	}
}

Optimistic Locks

TODO

Dirty Attributes