Active Records are objects that wrap a row in a database table or view, encapsulate the database access and add domain logic on that data. The basics of an Active Record are business classes, e.g., a Products class, that match very closely the record structure of an underlying database table. Each Active Record will be responsible for saving and loading data to and from the database.
Active Record is a good choice for domain logic that isn't too complex, such as creates, reads, updates, and deletes. Derivations and validations based on a single record work well in this structure. Active Record has the primary advantage of simplicity. It's easy to build Active Records, and they are easy to understand.
However, as your business logic grows in complexity, you'll soon want to use your object's direct relationships, collections, inheritance, and so forth. These don't map easily onto Active Record, and adding them piecemeal gets very messy. Another argument against Active Record is the fact that it couples the object design to the database design. This makes it more difficult to refactor as a project goes forward.
The alternative is to use a Data Mapper that separates the roles of the business object and how these objects are stored. Prado provides a complimentary choice between Active Record and SqlMap Data Mapper. A SqlMap Data Mapper can be used to load Active Record objects, in turn; these Active Record objects can be used to update the database. The "relationship" between Active Records and SqlMap is illustrated in the following diagram. More details regarding the SqlMap Data Mapper can be found in the SqlMap Manual. alt="Active Records and SqlMap DataMapper" id="fig:diagram.png" class="figure"/>
The Active Record class has functionality to perform the following tasks.
Prado's implementation of Active Record does not maintain referential identity. Each object obtained using Active Record is a copy of the data in the database. For example, If you ask for a particular customer and get back a Customer object, the next time you ask for that customer you get back another instance of a Customer object. This implies that a strict comparison (i.e., using ===) will return false, while loose comparison (i.e., using ==) will return true if the object values are equal by loose comparison.
This is design implication related to the following question. "Do you think of the customer as an object, of which there's only one, or do you think of the objects you operate on as copies of the database?" Other O/R mappings will imply that there is only one Customer object with custID 100, and it literally is that customer. If you get the customer and change a field on it, then you have now changed that customer. "That constrasts with: you have changed this copy of the customer, but not that copy. And if two people update the customer on two copies of the object, whoever updates first, or maybe last, wins." [A. Hejlsberg 2003]
The Active Record implementation utilizes the Prado DAO classes for data access. The current Active Record implementation supports the following database.
Support for other databases can be provided when there are sufficient demands.
Let us
consider the following "users" table that contains two columns named "username" and "email",
where "username" is also the primary key.
Next we define our Active Record class that corresponds to the "users" table.
Each column of the "users" table must have corresponding property of the same name as the column name in the UserRecord class. Of course, you also define additional member variables or properties that does not exist in the table structure. The class constant TABLE is optional when the class name is the same as the table name in the database, otherwise TABLE must specify the table name that corresponds to your Active Record class.
Since TActiveRecord extends TComponent, setter and getter methods can be defined to allow control over how variables are set and returned. For example, adding a $level property to the UserRecord class:
More details regarding TComponent can be found in the Components documentation. Later we shall use the getter/setters to allow for lazy loading of relationship objects.
The static method finder() returns an UserRecord instance that can be used to load records from the database. The loading of records using the finder methods is discussed a little later. The TActiveRecord::finder() static method takes the name of an Active Record class as parameter.
A default database connection for Active Record can be set as follows. See Establishing Database Connection for further details regarding creation of database connection in general.
Alternatively, you can create a base class and override the getDbConnection() method to return a database connection. This is a simple way to permit multiple connections and multiple databases. The following code demonstrates defining the database connection in a base class (not need to set the DB connection anywhere else).
The default database connection can also be configured using a <module>
tag in the application.xml
or config.xml as follows.
A ConnectionID property can be specified with value corresponding
to another TDataSourceConfig module configuration's ID value. This allows
the same database connection to be used in other modules such as SqlMap.
The TActiveRecord class provides many convenient methods to find
records from the database. The simplest is finding one record by matching a primary key or a
composite key (primary keys that consists of multiple columns).
See the
Finds one record using only a primary key or a composite key.
Finds multiple records using a list of primary keys or composite keys. The following are equivalent for primary keys (primary key consisting of only one column/field).
Finds one single record that matches the criteria. The criteria can be a partial SQL string or a TActiveRecordCriteria object.
The TActiveRecordCriteria class has the following properties:
Same as find() but returns an array of objects.
Dynamic find method using parts of the method name as search criteria. Method names starting with findBy return 1 record only and method names starting with findAllBy return an array of records. The condition is taken as part of the method name after findBy or findAllBy. The following blocks of code are equivalent:
Finds records using full SQL where findBySql()
return an Active Record and findAllBySql()returns an array of record objects.
For each column returned, the corresponding Active Record class must define a member variable or
property for each corresponding column name.
Find the number of matchings records, accepts same parameters as the findAll() method.
Add a new record using TActiveRecord is very simple, just create a new Active Record object and call the save() method. E.g.
To update a record in the database, just change one or more properties of
the Active Record object that has been loaded from the database and then
call the save() method.
Active Record objects have a simple life-cycle illustrated in the following diagram.
alt="Active Records Life Cycle" id="fig:cycle.png" class="figure"/>We see that new TActiveRecord objects are created by either using one of the find*() methods or using creating a new instance by using PHP's new keyword. Objects created by a find*() method starts with clean state. New instance of TActiveRecord created other than by a find*() method starts with new state. Whenever you call the save() method on the TActiveRecord object, the object enters the clean state. Objects in the clean becomes dirty whenever one of more of its internal states are changed. Calling the delete() method on the object ends the object life-cycle, no further actions can be performed on the object.
To delete an existing record that is already loaded, just call the delete() method. You can also delete records in the database by primary keys without loading any records using the deleteByPk() method (and equivalently the deleteAllByPks() method). For example, to delete one or several records with tables using one or more primary keys.
For composite keys (determined automatically from the table definitions):
To delete by a criteria, use deleteAll($criteria) and deleteBy*() with similar syntax to findAll($criteria) and findAllBy*() as described above.
All Active Record objects contain the property DbConnection
that can be used to obtain a transaction object.
The TActiveRecord offers two events, OnCreateCommand and OnExecuteCommand.
The OnCreateCommand event is raised when a command is prepared and parameter binding is completed. The parameter object is TDataGatewayEventParameter of which the Command property can be inspected to obtain the SQL query to be executed.
The OnExecuteCommand event is raised when a command is executed and the result from the database was returned. The parameter object is TDataGatewayResultEventParameter of which the Result property contains the data return from the database. The data returned can be changed by setting the Result property.
Using the OnExecuteCommand we can attach an event handler to log the entire SQL query executed for a given TActiveRecord class or instance. For example, we define a base class and override either the getDbConnection() or the constructor.
The Prado Active Record implementation supports the foreign key mappings for database that supports foreign key constraints. For Active Record relationships to function the underlying database must support foreign key constraints (e.g. MySQL using InnoDB).
In the following sections we will consider the following table relationships between Teams, Players, Skills and Profiles.
class="figure" />The goal is to obtain object models that represent to some degree the entity relationships in the above figure.
class="figure" />There is a mismatch between relationships with objects and table relationships. First there's a difference in representation. Objects handle links by storing references that are held by the runtime memory-managed environment. Relational databases handle links by forming a key into another table. Second, objects can easily use collections to handle multiple references from a single field, while normalization forces all entity relation links to be single valued. This leads to reversals of the data structure between objects and tables. The approach taken in the Prado Active Record design is to use the table foreign key constraints to derive object relationships. This implies that the underlying database must support foreign key constraints.
The entity relationship between the Teams and Players table is what is known as an 1-M relationship. That is, one Team may contain 0 or more Players. In terms of object relationships, we say that a TeamRecord object has many PlayerRecord objects. (Notice the reversal of the direction of relationships between tables and objects.)
We model the Team object as the following Active Record classes.
The static $RELATIONS property of TeamRecord defines that the property $players has many PlayerRecords. Multiple relationships is permitted by defining each relationship with an entry in the $RELATIONS array where array key for the entry corresponds to the property name. In array(self::HAS_MANY, 'PlayerRecord'), the first element defines the relationship type, the valid types are self::HAS_MANY, self::HAS_ONE and self::BELONGS_TO. The second element is a string 'PlayerRecord' that corresponds to the class name of the PlayerRecord class.
The foreign key constraint of the Players table is used to determine the corresponding Teams table's corresponding key names. This is done automatically handled in Active Record by inspecting the Players and Teams table definitions.
The "has many" relationship is not fetched automatically when you use any of the Active Record finder methods. You will need to explicitly fetch the related objects as follows. In the code below, both lines are equivalent and the method names are case insensitive.
The method with_xxx() (where xxx is the relationship property name, in this case, players) fetches the corresponding PlayerRecords using a second query (not by using a join). The with_xxx() accepts the same arguments as other finder methods of TActiveRecord, e.g. with_players('age = ?', 35).
The "has many" relationship in the above section defines a collection of foreign objects. In particular, we have that a TeamRecord has many (zero or more) PlayerRecord objects. We can also add a back pointer by adding a property in the PlayerRecord class that links back to the TeamRecord object, effectively making the association bidirectional. We say that the $team property in PlayerRecord class belongs to a TeamRecord object. The following code defines the complete PlayerRecord class with 3 relationships.
The static $RELATIONS property of PlayerRecord defines that the property $team belongs to a TeamRecord. The $RELATIONS array also defines two other relationships that we shall examine in later sections below. In array(self::BELONGS_TO, 'TeamRecord'), the first element defines the relationship type, in this case self::BELONGS_TO and the second element is a string 'TeamRecord' that corresponds to the class name of the TeamRecord class. A player object with the corresponding team object may be fetched as follows.
The method with_xxx() (where xxx is the relationship property name, in this case, team) fetches the corresponding TeamRecords using a second query (not by using a join). The with_xxx() accepts the same arguments as other finder methods of TActiveRecord, e.g. with_team('location = ?', 'Madrid').
The "belongs to" relationship of ProfileRecord class is defined similarly.
In essence, there exists a "belongs to" relationship for objects corresponding to entities that has column which are foreign keys. In particular, we see that the Profiles table has a foreign key constraint on the column player_id that relates to the Players table's player_id column. Thus, the ProfileRecord object has a property ($player) that belongs to a PlayerRecord object. Similarly, the Players table has a foreign key constraint on the column team_name that relates to the Teams table's name column. Thus, the PlayerRecord object has a property ($team) that belongs to a TeamRecord object.
The entity relationship between Players and Profiles is one to one. That is, each PlayerRecord object has one ProfileRecord object (may be none or null). A has one relationship is nearly identical to a has many relationship with the exception that the related object is only one object (not a collection of objects).
A parent child relationship can be defined using a combination of has many and belongs to relationship that refers to the same class. The following example shows a parent children relationship between "categories" and a "parent category".
Objects can handle multivalued fields quite easily by using collections as field values. Relational databases don't have this feature and are constrained to single-valued fields only. When you're mapping a one-to-many association you can handle this using has many relationships, essentially using a foreign key for the single-valued end of the association. But a many-to-many association can't do this because there is no single-valued end to hold the foreign key.
The answer is the classic resolution that's been used by relational data people for decades: create an extra table (an association table) to record the relationship. The basic idea is using an association table to store the association. This table has only the foreign key IDs for the two tables that are linked together, it has one row for each pair of associated objects.
The association table has no corresponding in-memory object and its primary key is the compound of the two primary keys of the tables that are associated. In simple terms, to load data from the association table you perform two queries (in general, it may also be achieved using one query consisting of joins). Consider loading the SkillRecord collection for a list PlayerRecord objects. In this case, you do queries in two stages. The first stage queries the Players table to find all the rows of the players you want. The second stage finds the SkillRecord object for the related player ID for each row in the Player_Skills association table using an inner join.
The Prado Active Record design implements the two stage approach. For the Players-Skills M-N (many-to-many) entity relationship, we need to define a has many relationship in the PlayerRecord class and in addition define a has many relationship in the SkillRecord class as well. The following sample code defines the complete SkillRecord class with a many-to-many relationship with the PlayerRecord class. (See the PlayerRecord class definition above to the corresponding many-to-many relationship with the SkillRecord class.)
The static $RELATIONS property of SkillRecord defines that the property $players has many PlayerRecords via an association table 'Player_Skills'. In array(self::HAS_MANY, 'PlayerRecord', 'Player_Skills'), the first element defines the relationship type, in this case self::HAS_MANY, the second element is a string 'PlayerRecord' that corresponds to the class name of the PlayerRecord class, and the third element is the name of the association table name.
A list of player objects with the corresponding collection of skill objects may be fetched as follows.
The method with_xxx() (where xxx is the relationship property name, in this case, Skill) fetches the corresponding SkillRecords using a second query (not by using a join). The with_xxx() accepts the same arguments as other finder methods of TActiveRecord.
For self referenced association tables, that is, the association points to the same
table. For example, consider the items table with M-N related
item via the related_items association table. The syntax in the following
example is valid for a PostgreSQL database. For other database, consult their respective documentation for
defining the foreign key constraints.
The association table name in third element of the relationship array may contain the foreign table column names. The columns defined in the association table must also be defined in the record class (e.g. the $related_item_id property corresponds to the related_item_id column in the related_items table).
Related objects can be simply inserted/updated by first adding those related objects to the current source object (i.e. the object currently been worked on) and then call the save() method on the source object. The related object's references and the association reference (if required) will be added and/or updated. For example, to add two new players to the team (assuming that 'Team A' exists), we can simply do the following.
Since the TeamRecord class contains a has many relationship with the PlayerRecord, then saving a TeamRecord object will also update the corresponding foreign objects in $players array. That is, the objects in $players are inserted/updated in the database and the $team_name property of those objects will contain the foreign key value that corresponds to the $team object's primary key value.
To delete a particular foreign object (or any Active Record object), simply call the object's delete() method. You may setup the database table's foreign key constraints such that when deleting a particular data in the database it will delete the referenced data as well (it may also be achieved using database triggers). E.g. such as having a "ON DELETE CASCADE" constraint. Deleting foreign objects by either setting the property value to null or removing the object from an array will NOT remove the corresponding data in the database.
To remove associations for the many-to-many relationships via an association table, an Active Record that corresponds to the association table can be used. Then the association can be removed by calling the deleteByPk() method, for example:
Using the with_xxx() methods will load the relationship record on demand. Retrieving the related record using lazy loading (that is, only when those related objects are accessed) can be achieved by using a feature of the TComponent that provides accessor methods. In particular, we define a pair of getter and setter methods where the getter method will retrieve the relationship conditionally. The following example illustrates that the PlayerRecord can retrieve its $skills foreign objects conditionally.
We first need to change the $skills=array() declaration to a private property $_skills (notice the underscore) and set it to null instead. This allows us to define the skills property using getter/setter methods (see Components for details). The getSkills() getter method for the skills property will lazy load the corresponding skills foreign record when it is used as follows. Notice that we only do a lazy load when its $player_id is not null (that is, when the record is already fetched from the database or player id was already set).
The setSkills() ensures that the skills property will always be a TList. Using a TList allows us to set the elements of the skills property as if they were arrays. E.g. $player->skills[] = new SkillRecord(). If array was used, a PHP error will be thrown.
Since v3.1.1, Active Record starts to support column mapping. Column mapping allows developers to address columns in Active Record using a more consistent naming convention. In particular, using column mapping, one can access a column using whatever name he likes, rather than limited by the name defined in the database schema.
To use column mapping, declare a static array named COLUMN_MAPPING in the Active Record class. The keys of the array are column names (called physical column names) as defined in the database schema, while the values are corresponding property names (called logical column names) defined in the Active Record class. The property names can be either public class member variable names or component property names defined via getters/setters. If a physical column name happens to be the same as the logical column name, they do not need to be listed in COLUMN_MAPPING.
With the above column mapping, we can address first_name using $userRecord->firstName instead of $userRecord->first_name. This helps separation of logic and model.