10.8. Zend_Db_Table Relationships
10.8.1. Introduction
Tables have relationships to each other in a relational database. An entity in one table can be linked to one or more entities in another table by using referential integrity constraints defined in the database schema.
The Zend_Db_Table_Row
class has methods for querying related rows in other tables.
10.8.2. Defining Relationships
Define classes for each of your tables, extending the abstract class
Zend_Db_Table_Abstract
, as described in ???. Also
see ??? for a description of the
example database for which the following example code is designed.
Below are the PHP class definitions for these tables:
class Accounts extends Zend_Db_Table_Abstract { protected $_name = 'accounts'; protected $_dependentTables = array('Bugs'); } class Products extends Zend_Db_Table_Abstract { protected $_name = 'products'; protected $_dependentTables = array('BugsProducts'); } class Bugs extends Zend_Db_Table_Abstract { protected $_name = 'bugs'; protected $_dependentTables = array('BugsProducts'); protected $_referenceMap = array( 'Reporter' => array( 'columns' => 'reported_by', 'refTableClass' => 'Accounts', 'refColumns' => 'account_name' ), 'Engineer' => array( 'columns' => 'assigned_to', 'refTableClass' => 'Accounts', 'refColumns' => 'account_name' ), 'Verifier' => array( 'columns' => array('verified_by'), 'refTableClass' => 'Accounts', 'refColumns' => array('account_name') ) ); } class BugsProducts extends Zend_Db_Table_Abstract { protected $_name = 'bugs_products'; protected $_referenceMap = array( 'Bug' => array( 'columns' => array('bug_id'), 'refTableClass' => 'Bugs', 'refColumns' => array('bug_id') ), 'Product' => array( 'columns' => array('product_id'), 'refTableClass' => 'Products', 'refColumns' => array('product_id') ) ); }
If you use Zend_Db_Table
to emulate cascading UPDATE and DELETE operations, declare the
$_dependentTables
array in the class for the parent table. List the class
name for each dependent table. Use the class name, not the physical name of the SQL
table.
注意 | |
---|---|
Skip declaration of |
Declare the $_referenceMap
array in the class for each dependent table.
This is an associative array of reference "rules". A reference rule identifies which
table is the parent table in the relationship, and also lists which columns in the
dependent table reference which columns in the parent table.
The rule key is a string used as an index to the $_referenceMap
array.
This rule key is used to identify each reference relationship. Choose a descriptive
name for this rule key. It's best to use a string that can be part of a PHP method
name, as you will see later.
In the example PHP code above, the rule keys in the Bugs table class are:
'Reporter'
, 'Engineer'
, 'Verifier'
, and
'Product'
.
The value of each rule entry in the $_referenceMap
array is also an
associative array. The elements of this rule entry are described below:
-
columns => A string or an array of strings naming the foreign key column name(s) in the dependent table.
It's common for this to be a single column, but some tables have multi-column keys.
-
refTableClass => The class name of the parent table. Use the class name, not the physical name of the SQL table.
It's common for a dependent table to have only one reference to its parent table, but some tables have multiple references to the same parent table. In the example database, there is one reference from the
bugs
table to theproducts
table, but three references from thebugs
table to theaccounts
table. Put each reference in a separate entry in the$_referenceMap
array. -
refColumns => A string or an array of strings naming the primary key column name(s) in the parent table.
It's common for this to be a single column, but some tables have multi-column keys. If the reference uses a multi-column key, the order of columns in the
'columns'
entry must match the order of columns in the'refColumns'
entry.It is optional to specify this element. If you don't specify the
refColumns
, the column(s) reported as the primary key columns of the parent table are used by default. onDelete => The rule for an action to execute if a row is deleted in the parent table. See 第 10.8.6 节 “Cascading Write Operations” for more information.
onUpdate => The rule for an action to execute if values in primary key columns are updated in the parent table. See 第 10.8.6 节 “Cascading Write Operations” for more information.
10.8.3. Fetching a Dependent Rowset
If you have a Row object as the result of a query on a parent table, you can fetch rows from dependent tables that reference the current row. Use the method:
$row->findDependentRowset($table, [$rule]);
This method returns a Zend_Db_Table_Rowset_Abstract
object, containing a set of rows
from the dependent table $table
that refer to the row identified by the
$row
object.
The first argument $table
can be a string that specifies the dependent
table by its class name. You can also specify the dependent table by using an object of
that table class.
例 10.12. Fetching a Dependent Rowset
This example shows getting a Row object from the table Accounts
, and
finding the Bugs
reported by that account.
$accountsTable = new Accounts(); $accountsRowset = $accountsTable->find(1234); $user1234 = $accountsRowset->current(); $bugsReportedByUser = $user1234->findDependentRowset('Bugs');
The second argument $rule
is optional. It is a string that names the rule
key in the $_referenceMap
array of the dependent table class. If you don't
specify a rule, the first rule in the array that references the parent table is used.
If you need to use a rule other than the first, you need to specify the key.
In the example code above, the rule key is not specified, so the rule used by default
is the first one that matches the parent table. This is the rule
'Reporter'
.
例 10.13. Fetching a Dependent Rowset By a Specific Rule
This example shows getting a Row object from the table Accounts
, and
finding the Bugs
assigned to be fixed by the user of that account. The
rule key string that corresponds to this reference relationship in this example is
'Engineer'
.
$accountsTable = new Accounts(); $accountsRowset = $accountsTable->find(1234); $user1234 = $accountsRowset->current(); $bugsAssignedToUser = $user1234->findDependentRowset('Bugs', 'Engineer');
You can also add criteria, ordering and limits to your relationships using the parent row's select object.
例 10.14. Fetching a Dependent Rowset using a Zend_Db_Table_Select
This example shows getting a Row object from the table Accounts
,
and finding the Bugs
assigned to be fixed by the user of that
account, limited only to 3 rows and ordered by name.
$accountsTable = new Accounts(); $accountsRowset = $accountsTable->find(1234); $user1234 = $accountsRowset->current(); $select = $accountsTable->select()->order('name ASC') ->limit(3); $bugsAssignedToUser = $user1234->findDependentRowset('Bugs', 'Engineer', $select);
Alternatively, you can query rows from a dependent table using a special mechanism
called a "magic method". Zend_Db_Table_Row_Abstract
invokes the method:
findDependentRowset('<TableClass>', '<Rule>')
if you invoke a method on
the Row object matching either of the following patterns:
$row->find<TableClass>()
$row->find<TableClass>By<Rule>()
In the patterns above, <TableClass>
and <Rule>
are strings
that correspond to the class name of the dependent table, and the dependent table's
rule key that references the parent table.
注意 | |
---|---|
Some application frameworks, such as Ruby on Rails, use a mechanism called
"inflection" to allow the spelling of identifiers to change depending on usage. For
simplicity, |
例 10.15. Fetching Dependent Rowsets using the Magic Method
This example shows finding dependent Rowsets equivalent to those in the previous examples. In this case, the application uses the magic method invocation instead of specifying the table and rule as strings.
$accountsTable = new Accounts(); $accountsRowset = $accountsTable->find(1234); $user1234 = $accountsRowset->current(); // Use the default reference rule $bugsReportedBy = $user1234->findBugs(); // Specify the reference rule $bugsAssignedTo = $user1234->findBugsByEngineer();
10.8.4. Fetching a Parent Row
If you have a Row object as the result of a query on a dependent table, you can fetch the row in the parent to which the dependent row refers. Use the method:
$row->findParentRow($table, [$rule]);
There always should be exactly one row in the parent table referenced by a dependent row, therefore this method returns a Row object, not a Rowset object.
The first argument $table
can be a string that specifies the parent table
by its class name. You can also specify the parent table by using an object of that
table class.
例 10.16. Fetching the Parent Row
This example shows getting a Row object from the table Bugs
(for
example one of those bugs with status 'NEW'), and finding the row in the
Accounts
table for the user who reported the bug.
$bugsTable = new Bugs(); $bugsRowset = $bugsTable->fetchAll(array('bug_status = ?' => 'NEW')); $bug1 = $bugsRowset->current(); $reporter = $bug1->findParentRow('Accounts');
The second argument $rule
is optional. It is a string that names the rule
key in the $_referenceMap
array of the dependent table class. If you don't
specify a rule, the first rule in the array that references the parent table is used.
If you need to use a rule other than the first, you need to specify the key.
In the example above, the rule key is not specified, so the rule used by default is the
first one that matches the parent table. This is the rule 'Reporter'
.
例 10.17. Fetching a Parent Row By a Specific Rule
This example shows getting a Row object from the table Bugs
, and
finding the account for the engineer assigned to fix that bug. The rule key string
that corresponds to this reference relationship in this example is
'Engineer'
.
$bugsTable = new Bugs(); $bugsRowset = $bugsTable->fetchAll(array('bug_status = ?', 'NEW')); $bug1 = $bugsRowset->current(); $engineer = $bug1->findParentRow('Accounts', 'Engineer');
Alternatively, you can query rows from a parent table using a "magic method".
Zend_Db_Table_Row_Abstract
invokes the method:
findParentRow('<TableClass>', '<Rule>')
if you invoke a method
on the Row object matching either of the following patterns:
$row->findParent<TableClass>([Zend_Db_Table_Select $select])
$row->findParent<TableClass>By<Rule>([Zend_Db_Table_Select $select])
In the patterns above, <TableClass>
and <Rule>
are strings that correspond to the class name of the parent table, and the dependent
table's rule key that references the parent table.
注意 | |
---|---|
The table identity and the rule key named in the method call must match the spelling of the class and rule key exactly. |
例 10.18. Fetching the Parent Row using the Magic Method
This example shows finding parent Rows equivalent to those in the previous examples. In this case, the application uses the magic method invocation instead of specifying the table and rule as strings.
$bugsTable = new Bugs(); $bugsRowset = $bugsTable->fetchAll(array('bug_status = ?', 'NEW')); $bug1 = $bugsRowset->current(); // Use the default reference rule $reporter = $bug1->findParentAccounts(); // Specify the reference rule $engineer = $bug1->findParentAccountsByEngineer();
10.8.5. Fetching a Rowset via a Many-to-many Relationship
If you have a Row object as the result of a query on one table in a many-to-many relationship (for purposes of the example, call this the "origin" table), you can fetch corresponding rows in the other table (call this the "destination" table) via an intersection table. Use the method:
$row->findManyToManyRowset($table, $intersectionTable, [$rule1, [$rule2, [Zend_Db_Table_Select $select] ] ]);
This method returns a Zend_Db_Table_Rowset_Abstract
containing rows from the table
$table
, satisfying the many-to-many relationship. The current Row object
$row
from the origin table is used to find rows in the intersection table,
and that is joined to the destination table.
The first argument $table
can be a string that specifies the destination
table in the many-to-many relationship by its class name. You can also specify the
destination table by using an object of that table class.
The second argument $intersectionTable
can be a string that specifies the
intersection table between the two tables in the the many-to-many relationship by its
class name. You can also specify the intersection table by using an object of that
table class.
例 10.19. Fetching a Rowset with the Many-to-many Method
This example shows getting a Row object from from the origin table
Bugs
, and finding rows from the destination table
Products
, representing products related to that bug.
$bugsTable = new Bugs(); $bugsRowset = $bugsTable->find(1234); $bug1234 = $bugsRowset->current(); $productsRowset = $bug1234->findManyToManyRowset('Products', 'BugsProducts');
The third and fourth arguments $rule1
and $rule2
are
optional. These are strings that name the rule keys in the $_referenceMap
array of the intersection table.
The $rule1
key names the rule for the relationship from the intersection
table to the origin table. In this example, this is the relationship from
BugsProducts
to Bugs
.
The $rule2
key names the rule for the relationship from the intersection
table to the destination table. In this example, this is the relationship from
Bugs
to Products
.
Similarly to the methods for finding parent and dependent rows, if you don't specify a
rule, the method uses the first rule in the $_referenceMap
array that
matches the tables in the relationship. If you need to use a rule other than the first,
you need to specify the key.
In the example code above, the rule key is not specified, so the rules used by default
are the first ones that match. In this case, $rule1
is
'Reporter'
and $rule2
is 'Product'
.
例 10.20. Fetching a Rowset with the Many-to-many Method By a Specific Rule
This example shows geting a Row object from from the origin table
Bugs
, and finding rows from the destination table
Products
, representing products related to that bug.
$bugsTable = new Bugs(); $bugsRowset = $bugsTable->find(1234); $bug1234 = $bugsRowset->current(); $productsRowset = $bug1234->findManyToManyRowset('Products', 'BugsProducts', 'Bug');
Alternatively, you can query rows from the destination table in a many-to-many
relationship using a "magic method." Zend_Db_Table_Row_Abstract
invokes the method:
findManyToManyRowset('<TableClass>', '<IntersectionTableClass>',
'<Rule1>', '<Rule2>')
if you invoke a method matching any of the
following patterns:
$row->find<TableClass>Via<IntersectionTableClass> ([Zend_Db_Table_Select $select])
$row->find<TableClass>Via<IntersectionTableClass>By<Rule1> ([Zend_Db_Table_Select $select])
$row->find<TableClass>Via<IntersectionTableClass>By<Rule1>And<Rule2> ([Zend_Db_Table_Select $select])
In the patterns above, <TableClass>
and
<IntersectionTableClass>
are strings that correspond to the class
names of the destination table and the intersection table, respectively.
<Rule1>
and <Rule2>
are strings that correspond
to the rule keys in the intersection table that reference the origin table and the
destination table, respectively.
注意 | |
---|---|
The table identities and the rule keys named in the method call must match the spelling of the class and rule key exactly. |
例 10.21. Fetching Rowsets using the Magic Many-to-many Method
This example shows finding rows in the destination table of a many-to-many relationship representing products related to a given bug.
$bugsTable = new Bugs(); $bugsRowset = $bugsTable->find(1234); $bug1234 = $bugsRowset->current(); // Use the default reference rule $products = $bug1234->findProductsViaBugsProducts(); // Specify the reference rule $products = $bug1234->findProductsViaBugsProductsByBug();
10.8.6. Cascading Write Operations
Declare DRI in the database: | |
---|---|
Declaring cascading operations in
For example, if you use mysql's myisam storage engine, or sqlite, these solutions
do not support DRI. You may find it helpful to declare the cascading operations
with
If your RDBMS implements DRI and the
Most importantly, do not declare cascading operations both in the RDBMS and in your
|
You can declare cascading operations to execute against a dependent table when you
apply an UPDATE
or a DELETE
to a row in a parent table.
例 10.22. Example of a Cascading Delete
This example shows deleting a row in the Products
table, which is
configured to automatically delete dependent rows in the Bugs
table.
$productsTable = new Products(); $productsRowset = $productsTable->find(1234); $product1234 = $productsRowset->current(); $product1234->delete(); // Automatically cascades to Bugs table // and deletes dependent rows.
Similarly, if you use UPDATE
to change the value of a primary key in a
parent table, you may want the value in foreign keys of dependent tables to be updated
automatically to match the new value, so that such references are kept up to date.
It's usually not necessary to update the value of a primary key that was generated by a sequence or other mechanism. But if you use a natural key that may change value occasionally, it is more likely that you need to apply cascading updates to dependent tables.
To declare a cascading relationship in the Zend_Db_Table
, edit the rules in the
$_referenceMap
. Set the associative array keys 'onDelete'
and
'onUpdate'
to the string 'cascade' (or the constant
self::CASCADE
). Before a row is deleted from the parent table, or its
primary key values updated, any rows in the dependent table that refer to the parent's
row are deleted or updated first.
例 10.23. Example Declaration of Cascading Operations
In the example below, rows in the Bugs
table are automatically deleted
if the row in the Products
table to which they refer is deleted. The
'onDelete'
element of the reference map entry is set to
self::CASCADE
.
No cascading update is done in the example below if the primary key value in the
parent class is changed. The 'onUpdate'
element of the reference map
entry is self::RESTRICT
. You can get the same result by omitting
the 'onUpdate'
entry.
class BugsProducts extends Zend_Db_Table_Abstract { ... protected $_referenceMap = array( 'Product' => array( 'columns' => array('product_id'), 'refTableClass' => 'Products', 'refColumns' => array('product_id'), 'onDelete' => self::CASCADE, 'onUpdate' => self::RESTRICT ), ... ); }
10.8.6.1. Notes Regarding Cascading Operations
Cascading operations invoked by Zend_Db_Table
are not
atomic.
This means that if your database implements and enforces referential integrity
constraints, a cascading UPDATE
executed by a Zend_Db_Table
class
conflicts with the constraint, and results in a referential integrity violation.
You can use cascading UPDATE
in Zend_Db_Table
only if your database does not enforce that referential
integrity constraint.
Cascading DELETE
suffers less from the problem of referential
integrity violations. You can delete dependent rows as a non-atomic action before
deleting the parent row that they reference.
However, for both UPDATE
and DELETE
, changing the
database in a non-atomic way also creates the risk that another database user can
see the data in an inconsistent state. For example, if you delete a row and all its
dependent rows, there is a small chance that another database client program can
query the database after you have deleted the dependent rows, but before you delete
the parent row. That client program may see the parent row with no dependent rows,
and assume this is the intended state of the data. There is no way for that client
to know that its query read the database in the middle of a change.
The issue of non-atomic change can be mitigated by using transactions to isolate your change. But some RDBMS brands don't support transactions, or allow clients to read "dirty" changes that have not been committed yet.
Cascading operations in Zend_Db_Table
are invoked only by
Zend_Db_Table
.
Cascading deletes and updates defined in your Zend_Db_Table
classes are applied if
you execute the save()
or delete()
methods on the Row
class. However, if you update or delete data using another interface, such as a
query tool or another application, the cascading operations are not applied. Even
when using update()
and delete()
methods in the
Zend_Db_Adapter
class, cascading operations defined in your Zend_Db_Table
classes
are not executed.
No Cascading INSERT
.
There is no support for a cascading INSERT
. You must insert a row to a
parent table in one operation, and insert row(s) to a dependent table in a separate
operation.