repo_zf1/documentation/manual/es/module_specs/Zend_Db_Table-Relationships.xml

<?xml version="1.0" encoding="UTF-8"?>
<!-- EN-Revision: 17598 -->
<!-- Reviewed: no -->
<sect1 id="zend.db.table.relationships">

    <title>Zend_Db_Table Relationships</title>

    <sect2 id="zend.db.table.relationships.introduction">

        <title>Introduction</title>

       <para>
            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.
       </para>

       <para>
            The <classname>Zend_Db_Table_Row</classname> class has methods for querying related rows
            in other tables.
       </para>

    </sect2>

    <sect2 id="zend.db.table.relationships.defining">

        <title>Defining Relationships</title>

       <para>
            Define classes for each of your tables, extending the abstract class
            <classname>Zend_Db_Table_Abstract</classname>, as described in
            <xref linkend="zend.db.table.defining" />. Also
            see <xref linkend="zend.db.adapter.example-database" /> for a description of the
            example database for which the following example code is designed.
       </para>

       <para>
            Below are the <acronym>PHP</acronym> class definitions for these tables:
       </para>

        <programlisting language="php"><![CDATA[
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')
        )
    );

}
]]></programlisting>

       <para>
            If you use <classname>Zend_Db_Table</classname> to emulate cascading UPDATE and DELETE
            operations, declare the <varname>$_dependentTables</varname> 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 <acronym>SQL</acronym> table.
       </para>

        <note>

           <para>
                Skip declaration of <varname>$_dependentTables</varname> if you use referential
                integrity constraints in the <acronym>RDBMS</acronym> server to implement cascading
                operations. See <xref linkend="zend.db.table.relationships.cascading" /> for more
                information.
           </para>

        </note>

       <para>
            Declare the <varname>$_referenceMap</varname> 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.
       </para>

       <para>
            The rule key is a string used as an index to the <varname>$_referenceMap</varname>
            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
            <acronym>PHP</acronym> method name, as you will see later.
       </para>

       <para>
            In the example <acronym>PHP</acronym> code above, the rule keys in the Bugs table class
            are: <code>'Reporter'</code>, <code>'Engineer'</code>, <code>'Verifier'</code>, and
            <code>'Product'</code>.
       </para>

       <para>
            The value of each rule entry in the <varname>$_referenceMap</varname> array is also an
            associative array. The elements of this rule entry are described below:
       </para>

        <itemizedlist>
            <listitem>
               <para>
                    <emphasis>columns</emphasis> => A string or an array of strings
                    naming the foreign key column name(s) in the dependent table.
               </para>

               <para>
                    It's common for this to be a single column, but some tables have multi-column
                    keys.
               </para>
            </listitem>

            <listitem>
               <para>
                    <emphasis>refTableClass</emphasis> => The class name of the parent table. Use
                    the class name, not the physical name of the <acronym>SQL</acronym> table.
               </para>

               <para>
                    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 <code>bugs</code> table
                    to the <code>products</code> table, but three references from the
                    <code>bugs</code> table to the <code>accounts</code> table. Put each reference
                    in a separate entry in the <varname>$_referenceMap</varname> array.
               </para>
            </listitem>

            <listitem>
               <para>
                    <emphasis>refColumns</emphasis> => A string or an array of
                    strings naming the primary key column name(s) in the parent table.
               </para>

               <para>
                    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
                    <code>'columns'</code> entry must match the order of columns in the
                    <code>'refColumns'</code> entry.
               </para>

               <para>
                    It is optional to specify this element. If you don't specify the
                    <code>refColumns</code>, the column(s) reported as the primary key columns of
                    the parent table are used by default.
               </para>
            </listitem>

            <listitem>
               <para>
                    <emphasis>onDelete</emphasis> => The rule for an action to
                    execute if a row is deleted in the parent table. See
                    <xref linkend="zend.db.table.relationships.cascading" /> for more information.
               </para>
            </listitem>

            <listitem>
               <para>
                    <emphasis>onUpdate</emphasis> => The rule for an action to
                    execute if values in primary key columns are updated in the parent table. See
                    <xref linkend="zend.db.table.relationships.cascading" /> for more information.
               </para>
            </listitem>
        </itemizedlist>

    </sect2>

    <sect2 id="zend.db.table.relationships.fetching.dependent">

        <title>Fetching a Dependent Rowset</title>

       <para>
            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:
       </para>

        <programlisting language="php"><![CDATA[
$row->findDependentRowset($table, [$rule]);
]]></programlisting>

       <para>
            This method returns a <classname>Zend_Db_Table_Rowset_Abstract</classname> object,
            containing a set of rows from the dependent table <varname>$table</varname> that refer
            to the row identified by the <varname>$row</varname> object.
       </para>

       <para>
            The first argument <varname>$table</varname> 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.
       </para>

        <example id="zend.db.table.relationships.fetching.dependent.example">

            <title>Fetching a Dependent Rowset</title>

           <para>
                This example shows getting a Row object from the table <code>Accounts</code>, and
                finding the <code>Bugs</code> reported by that account.
           </para>

            <programlisting language="php"><![CDATA[
$accountsTable = new Accounts();
$accountsRowset = $accountsTable->find(1234);
$user1234 = $accountsRowset->current();

$bugsReportedByUser = $user1234->findDependentRowset('Bugs');
]]></programlisting>

        </example>

       <para>
            The second argument <varname>$rule</varname> is optional. It is a string that names the
            rule key in the <varname>$_referenceMap</varname> 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.
       </para>

       <para>
            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
            <code>'Reporter'</code>.
       </para>

        <example id="zend.db.table.relationships.fetching.dependent.example-by">

            <title>Fetching a Dependent Rowset By a Specific Rule</title>

           <para>
                This example shows getting a Row object from the table <code>Accounts</code>, and
                finding the <code>Bugs</code> assigned to be fixed by the user of that account. The
                rule key string that corresponds to this reference relationship in this example is
                <code>'Engineer'</code>.
           </para>

            <programlisting language="php"><![CDATA[
$accountsTable = new Accounts();
$accountsRowset = $accountsTable->find(1234);
$user1234 = $accountsRowset->current();

$bugsAssignedToUser = $user1234->findDependentRowset('Bugs', 'Engineer');
]]></programlisting>

        </example>

       <para>
            You can also add criteria, ordering and limits to your relationships using the parent
            row's select object.
       </para>

       <para>

            <example id="zend.db.table.relationships.fetching.dependent.example-by-select">

                <title>Fetching a Dependent Rowset using a Zend_Db_Table_Select</title>

               <para>
                    This example shows getting a Row object from the table <code>Accounts</code>,
                    and finding the <code>Bugs</code> assigned to be fixed by the user of that
                    account, limited only to 3 rows and ordered by name.
               </para>

                <programlisting language="php"><![CDATA[
$accountsTable = new Accounts();
$accountsRowset = $accountsTable->find(1234);
$user1234 = $accountsRowset->current();
$select = $accountsTable->select()->order('name ASC')
                                  ->limit(3);

$bugsAssignedToUser = $user1234->findDependentRowset('Bugs',
                                                     'Engineer',
                                                     $select);
]]></programlisting>

            </example>

            Alternatively, you can query rows from a dependent table using a special mechanism
            called a "magic method". <classname>Zend_Db_Table_Row_Abstract</classname> invokes the
            method: <methodname>findDependentRowset('&lt;TableClass&gt;',
                '&lt;Rule&gt;')</methodname> if you invoke a method on the Row object matching
            either of the following patterns:
       </para>

        <itemizedlist>
            <listitem>
               <para>
                    <code>$row->find&lt;TableClass&gt;()</code>
               </para>
            </listitem>

            <listitem>
               <para>
                    <code>$row->find&lt;TableClass&gt;By&lt;Rule&gt;()</code>
               </para>
            </listitem>
        </itemizedlist>

       <para>
            In the patterns above, <code>&lt;TableClass&gt;</code> and <code>&lt;Rule&gt;</code> are
            strings that correspond to the class name of the dependent table, and the dependent
            table's rule key that references the parent table.
       </para>

        <note>

           <para>
                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, <classname>Zend_Db_Table_Row</classname> does not provide any inflection
                mechanism. The table identity and the rule key named in the method call must match
                the spelling of the class and rule key exactly.
           </para>

        </note>

        <example id="zend.db.table.relationships.fetching.dependent.example-magic">

            <title>Fetching Dependent Rowsets using the Magic Method</title>

           <para>
                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.
           </para>

            <programlisting language="php"><![CDATA[
$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();
]]></programlisting>

        </example>

    </sect2>

    <sect2 id="zend.db.table.relationships.fetching.parent">

        <title>Fetching a Parent Row</title>

       <para>
            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:
       </para>

        <programlisting language="php"><![CDATA[
$row->findParentRow($table, [$rule]);
]]></programlisting>

       <para>
            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.
       </para>

       <para>
            The first argument <varname>$table</varname> 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.
       </para>

        <example id="zend.db.table.relationships.fetching.parent.example">

            <title>Fetching the Parent Row</title>

           <para>
                This example shows getting a Row object from the table <code>Bugs</code> (for
                example one of those bugs with status 'NEW'), and finding the row in the
                <code>Accounts</code> table for the user who reported the bug.
           </para>

            <programlisting language="php"><![CDATA[
$bugsTable = new Bugs();
$bugsRowset = $bugsTable->fetchAll(array('bug_status = ?' => 'NEW'));
$bug1 = $bugsRowset->current();

$reporter = $bug1->findParentRow('Accounts');
]]></programlisting>

        </example>

       <para>
            The second argument <varname>$rule</varname> is optional. It is a string that names the
            rule key in the <varname>$_referenceMap</varname> 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.
       </para>

       <para>
            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 <code>'Reporter'</code>.
       </para>

        <example id="zend.db.table.relationships.fetching.parent.example-by">

            <title>Fetching a Parent Row By a Specific Rule</title>

           <para>
                This example shows getting a Row object from the table <code>Bugs</code>, 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
                <code>'Engineer'</code>.
           </para>

            <programlisting language="php"><![CDATA[
$bugsTable = new Bugs();
$bugsRowset = $bugsTable->fetchAll(array('bug_status = ?', 'NEW'));
$bug1 = $bugsRowset->current();

$engineer = $bug1->findParentRow('Accounts', 'Engineer');
]]></programlisting>

        </example>

       <para>
            Alternatively, you can query rows from a parent table using a "magic method".
            <classname>Zend_Db_Table_Row_Abstract</classname> invokes the method:
            <methodname>findParentRow('&lt;TableClass&gt;', '&lt;Rule&gt;')</methodname> if you
            invoke a method on the Row object matching either of the following patterns:
       </para>

        <itemizedlist>
            <listitem>
               <para>
                    <code>$row->findParent&lt;TableClass&gt;([Zend_Db_Table_Select $select])</code>
               </para>
            </listitem>

            <listitem>
               <para>
                    <code>$row->findParent&lt;TableClass&gt;By&lt;Rule&gt;([Zend_Db_Table_Select
                    $select])</code>
               </para>
            </listitem>
        </itemizedlist>

       <para>
            In the patterns above, <code>&lt;TableClass&gt;</code> and <code>&lt;Rule&gt;</code>
            are strings that correspond to the class name of the parent table, and the dependent
            table's rule key that references the parent table.
       </para>

        <note>

           <para>
                The table identity and the rule key named in the method call must match the
                spelling of the class and rule key exactly.
           </para>

        </note>

        <example id="zend.db.table.relationships.fetching.parent.example-magic">

            <title>Fetching the Parent Row using the Magic Method</title>

           <para>
                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.
           </para>

            <programlisting language="php"><![CDATA[
$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();
]]></programlisting>

        </example>

    </sect2>

    <sect2 id="zend.db.table.relationships.fetching.many-to-many">

        <title>Fetching a Rowset via a Many-to-many Relationship</title>

       <para>
            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:
       </para>

        <programlisting language="php"><![CDATA[
$row->findManyToManyRowset($table,
                           $intersectionTable,
                           [$rule1,
                               [$rule2,
                                   [Zend_Db_Table_Select $select]
                               ]
                           ]);
]]></programlisting>

       <para>
            This method returns a <classname>Zend_Db_Table_Rowset_Abstract</classname> containing
            rows from the table <varname>$table</varname>, satisfying the many-to-many relationship.
            The current Row object <varname>$row</varname> from the origin table is used to find
            rows in the intersection table, and that is joined to the destination table.
       </para>

       <para>
            The first argument <varname>$table</varname> 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.
       </para>

       <para>
            The second argument <varname>$intersectionTable</varname> can be a string that specifies
            the intersection table between the two tables in the many-to-many relationship by
            its class name. You can also specify the intersection table by using an object of that
            table class.
       </para>

        <example id="zend.db.table.relationships.fetching.many-to-many.example">

            <title>Fetching a Rowset with the Many-to-many Method</title>

           <para>
                This example shows getting a Row object from the origin table
                <code>Bugs</code>, and finding rows from the destination table
                <code>Products</code>, representing products related to that bug.
           </para>

            <programlisting language="php"><![CDATA[
$bugsTable = new Bugs();
$bugsRowset = $bugsTable->find(1234);
$bug1234 = $bugsRowset->current();

$productsRowset = $bug1234->findManyToManyRowset('Products',
                                                 'BugsProducts');
]]></programlisting>

        </example>

       <para>
            The third and fourth arguments <varname>$rule1</varname> and <varname>$rule2</varname>
            are optional. These are strings that name the rule keys in the
            <varname>$_referenceMap</varname> array of the intersection table.
       </para>

       <para>
            The <varname>$rule1</varname> key names the rule for the relationship from the
            intersection table to the origin table. In this example, this is the relationship from
            <code>BugsProducts</code> to <code>Bugs</code>.
       </para>

       <para>
            The <varname>$rule2</varname> key names the rule for the relationship from the
            intersection table to the destination table. In this example, this is the relationship
            from <code>Bugs</code> to <code>Products</code>.
       </para>

       <para>
            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 <varname>$_referenceMap</varname> 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.
       </para>

       <para>
            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, <varname>$rule1</varname> is
            <code>'Reporter'</code> and <varname>$rule2</varname> is <code>'Product'</code>.
       </para>

        <example id="zend.db.table.relationships.fetching.many-to-many.example-by">

            <title>Fetching a Rowset with the Many-to-many Method By a Specific Rule</title>

           <para>
                This example shows geting a Row object from the origin table
                <code>Bugs</code>, and finding rows from the destination table
                <code>Products</code>, representing products related to that bug.
           </para>

            <programlisting language="php"><![CDATA[
$bugsTable = new Bugs();
$bugsRowset = $bugsTable->find(1234);
$bug1234 = $bugsRowset->current();

$productsRowset = $bug1234->findManyToManyRowset('Products',
                                                 'BugsProducts',
                                                 'Bug');
]]></programlisting>

        </example>

       <para>
            Alternatively, you can query rows from the destination table in a many-to-many
            relationship using a "magic method." <classname>Zend_Db_Table_Row_Abstract</classname>
            invokes the method: <code>findManyToManyRowset('&lt;TableClass&gt;',
            '&lt;IntersectionTableClass&gt;', '&lt;Rule1&gt;', '&lt;Rule2&gt;')</code> if you invoke
            a method matching any of the following patterns:
       </para>

        <itemizedlist>
            <listitem>
               <para>
                    <code>$row->find&lt;TableClass&gt;Via&lt;IntersectionTableClass&gt;
                    ([Zend_Db_Table_Select $select])</code>
               </para>
            </listitem>

            <listitem>
               <para>
                    <code>$row->find&lt;TableClass&gt;Via&lt;IntersectionTableClass&gt;By&lt;Rule1&gt;
                        ([Zend_Db_Table_Select $select])</code>
               </para>
            </listitem>

            <listitem>
               <para>
                    <code>$row->find&lt;TableClass&gt;Via&lt;IntersectionTableClass&gt;By&lt;Rule1&gt;And&lt;Rule2&gt;
                        ([Zend_Db_Table_Select $select])</code>
               </para>
            </listitem>
        </itemizedlist>

       <para>
            In the patterns above, <code>&lt;TableClass&gt;</code> and
            <code>&lt;IntersectionTableClass&gt;</code> are strings that correspond to the class
            names of the destination table and the intersection table, respectively.
            <code>&lt;Rule1&gt;</code> and <code>&lt;Rule2&gt;</code> are strings that correspond
            to the rule keys in the intersection table that reference the origin table and the
            destination table, respectively.
       </para>

        <note>

           <para>
                The table identities and the rule keys named in the method call must match the
                spelling of the class and rule key exactly.
           </para>

        </note>

        <example id="zend.db.table.relationships.fetching.many-to-many.example-magic">

            <title>Fetching Rowsets using the Magic Many-to-many Method</title>

           <para>
                This example shows finding rows in the destination table of a many-to-many
                relationship representing products related to a given bug.
           </para>

            <programlisting language="php"><![CDATA[
$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();
]]></programlisting>

        </example>

    </sect2>

    <sect2 id="zend.db.table.relationships.cascading">

        <title>Cascading Write Operations</title>

        <note>

            <title>Declare DRI in the database:</title>

           <para>
                Declaring cascading operations in <classname>Zend_Db_Table</classname> is intended
                <emphasis>only</emphasis> for <acronym>RDBMS</acronym> brands that do not support
                declarative referential integrity (DRI).
           </para>

           <para>
                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 <classname>Zend_Db_Table</classname>.
           </para>

           <para>
                If your <acronym>RDBMS</acronym> implements DRI and the <code>ON DELETE</code> and
                <code>ON UPDATE</code> clauses, you should declare these clauses in your database
                schema, instead of using the cascading feature in
                <classname>Zend_Db_Table</classname>. Declaring cascading DRI rules in the
                <acronym>RDBMS</acronym> is better for database performance, consistency, and
                integrity.
           </para>

           <para>
                Most importantly, do not declare cascading operations both in the
                <acronym>RDBMS</acronym> and in your <classname>Zend_Db_Table</classname> class.
           </para>

        </note>

       <para>
            You can declare cascading operations to execute against a dependent table when you
            apply an <constant>UPDATE</constant> or a <constant>DELETE</constant> to a row in a
            parent table.
       </para>

        <example id="zend.db.table.relationships.cascading.example-delete">

            <title>Example of a Cascading Delete</title>

           <para>
                This example shows deleting a row in the <code>Products</code> table, which is
                configured to automatically delete dependent rows in the <code>Bugs</code> table.
           </para>

            <programlisting language="php"><![CDATA[
$productsTable = new Products();
$productsRowset = $productsTable->find(1234);
$product1234 = $productsRowset->current();

$product1234->delete();
// Automatically cascades to Bugs table
// and deletes dependent rows.
]]></programlisting>

        </example>

       <para>
            Similarly, if you use <constant>UPDATE</constant> 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.
       </para>

       <para>
            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 <emphasis>natural key</emphasis> that may
            change value occasionally, it is more likely that you need to apply cascading updates
            to dependent tables.
       </para>

       <para>
            To declare a cascading relationship in the <classname>Zend_Db_Table</classname>, edit
            the rules in the <varname>$_referenceMap</varname>. Set the associative array keys
            <code>'onDelete'</code> and <code>'onUpdate'</code> to the string 'cascade' (or the
            constant <constant>self::CASCADE</constant>). 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.
       </para>

        <example id="zend.db.table.relationships.cascading.example-declaration">

            <title>Example Declaration of Cascading Operations</title>

           <para>
                In the example below, rows in the <code>Bugs</code> table are automatically deleted
                if the row in the <code>Products</code> table to which they refer is deleted. The
                <code>'onDelete'</code> element of the reference map entry is set to
                <constant>self::CASCADE</constant>.
           </para>

           <para>
                No cascading update is done in the example below if the primary key value in the
                parent class is changed. The <code>'onUpdate'</code> element of the reference map
                entry is <constant>self::RESTRICT</constant>. You can get the same result by
                omitting the <code>'onUpdate'</code> entry.
           </para>

            <programlisting language="php"><![CDATA[
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
        ),
        ...
    );
}
]]></programlisting>

        </example>

        <sect3 id="zend.db.table.relationships.cascading.notes">

            <title>Notes Regarding Cascading Operations</title>

           <para>
                <emphasis>Cascading operations invoked by <classname>Zend_Db_Table</classname> are
                not atomic.</emphasis>
           </para>

           <para>
                This means that if your database implements and enforces referential integrity
                constraints, a cascading <constant>UPDATE</constant> executed by a
                <classname>Zend_Db_Table</classname> class conflicts with the constraint, and
                results in a referential integrity violation. You can use cascading
                <constant>UPDATE</constant> in <classname>Zend_Db_Table</classname>
                <emphasis>only</emphasis> if your database does not enforce that referential
                integrity constraint.
           </para>

           <para>
                Cascading <constant>DELETE</constant> 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.
           </para>

           <para>
                However, for both <constant>UPDATE</constant> and <constant>DELETE</constant>,
                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.
           </para>

           <para>
                The issue of non-atomic change can be mitigated by using transactions to isolate
                your change. But some <acronym>RDBMS</acronym> brands don't support transactions, or
                allow clients to read "dirty" changes that have not been committed yet.
           </para>

           <para>
                <emphasis>Cascading operations in <classname>Zend_Db_Table</classname> are invoked
                only by <classname>Zend_Db_Table</classname>.</emphasis>
           </para>

           <para>
                Cascading deletes and updates defined in your <classname>Zend_Db_Table</classname>
                classes are applied if you execute the <methodname>save()</methodname> or
                <methodname>delete()</methodname> 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
                <methodname>update()</methodname> and <methodname>delete()</methodname> methods
                in the <classname>Zend_Db_Adapter</classname> class, cascading operations defined in
                your <classname>Zend_Db_Table</classname> classes are not executed.
           </para>

           <para>
                <emphasis>No Cascading <constant>INSERT</constant>.</emphasis>
           </para>

           <para>
                There is no support for a cascading <constant>INSERT</constant>. You must insert a
                row to a parent table in one operation, and insert row(s) to a dependent table in a
                separate operation.
           </para>

        </sect3>

    </sect2>

</sect1>
<!--
vim:se ts=4 sw=4 et:
-->