<?xml version="1.0" encoding="utf-8"?>
<!-- $Revision$ -->
 <chapter xml:id="language.references" xmlns="http://docbook.org/ns/docbook">
  <title>References Explained</title>
  
  <sect1 xml:id="language.references.whatare">
   <title>What References Are</title>
   <simpara>
    References in PHP are a means to access the same variable content
    by different names. They are not like C pointers; for instance,
    you cannot perform pointer arithmetic using them, they are not
    actual memory addresses, and so on. See
    <xref linkend="language.references.arent" /> for more
    information. Instead, they are
    <link linkend="features.gc.refcounting-basics">symbol table</link>
    aliases. Note that in PHP, variable name and variable content are different, so the same
    content can have different names. The closest analogy is with
    Unix filenames and files - variable names are directory entries,
    while variable content is the file itself. References can be
    likened to hardlinking in Unix filesystem.
   </simpara>
  </sect1>

  <sect1 xml:id="language.references.whatdo">
   <title>What References Do</title>
   <para>
    There are three basic operations performed using references:
    <link linkend="language.references.whatdo.assign">assigning by
    reference</link>, <link linkend="language.references.whatdo.pass">passing
    by reference</link>,
    and <link linkend="language.references.whatdo.return">returning by
    reference</link>. This section will give an introduction to these
    operations, with links for further reading.
   </para>
   <sect2 xml:id="language.references.whatdo.assign">
    <title>Assign By Reference</title>
    <para>
     In the first of these, PHP references allow you to make two
     variables refer to the same content. Meaning, when you do:
     <informalexample>
      <programlisting role="php">
<![CDATA[
<?php

$a =& $b;

?>
]]>
      </programlisting>
     </informalexample>
     it means that <varname>$a</varname> and <varname>$b</varname>
     point to the same content.
     <note>
      <para>
       <varname>$a</varname> and <varname>$b</varname> are completely
       equal here. <varname>$a</varname> is not pointing to
       <varname>$b</varname> or vice versa.
       <varname>$a</varname> and <varname>$b</varname> are pointing to the
       same place.
      </para>
     </note>
    </para>
    <note>
     <para>
      If you assign, pass, or return an undefined variable by reference, 
      it will get created.
      <example>
       <title>Using references with undefined variables</title>
       <programlisting role="php">
<![CDATA[
<?php

function foo(&$var) {}

foo($a); // $a is "created" and assigned to null

$b = array();
foo($b['b']);
var_dump(array_key_exists('b', $b)); // bool(true)

$c = new stdClass();
foo($c->d);
var_dump(property_exists($c, 'd')); // bool(true)

?>
]]>
       </programlisting>
      </example>
     </para>
    </note>
    <para>
     The same syntax can be used with functions that return
     references:
     <informalexample>
      <programlisting role="php">
<![CDATA[
<?php

$foo =& find_var($bar);

?>
]]>
      </programlisting>
     </informalexample>
    </para>
    <para>
     Using the same syntax with a function that does <emphasis>not</emphasis>
     return by reference will give an error, as will using it with the result
     of the <link linkend="language.oop5.basic.new">new</link> operator.
     Although objects are passed around as pointers, these are not the same as references,
     as explained under <link linkend="language.oop5.references">Objects and references</link>.
    </para>
    <warning>
     <para>
      If you assign a reference to a variable declared <literal>global</literal>
      inside a function, the reference will be visible only inside the function.
      You can avoid this by using the <varname>$GLOBALS</varname> array.
      <example>
       <title>Referencing global variables inside functions</title>
       <programlisting role="php">
<![CDATA[
<?php

$var1 = "Example variable";
$var2 = "";

function global_references($use_globals)
{
    global $var1, $var2;

    if (!$use_globals) {
        $var2 =& $var1; // visible only inside the function
    } else {
        $GLOBALS["var2"] =& $var1; // visible also in global context
    }
}

global_references(false);
echo "var2 is set to '$var2'\n"; // var2 is set to ''

global_references(true);
echo "var2 is set to '$var2'\n"; // var2 is set to 'Example variable'

?>
]]>
       </programlisting>
      </example>
      Think about <literal>global $var;</literal> as a shortcut to <literal>$var
        =&amp; $GLOBALS['var'];</literal>. Thus assigning another reference
      to <literal>$var</literal> only changes the local variable's reference.
     </para>
    </warning>
    <note>
     <para>
      If you assign a value to a variable with references in a 
      &foreach; statement, the references are modified too.
      <example>
       <title>References and foreach statement</title>
       <programlisting role="php">
<![CDATA[
<?php

$ref = 0;
$row =& $ref;

foreach (array(1, 2, 3) as $row) {
    // Do something
}

echo $ref; // 3 - last element of the iterated array

?>
]]>
       </programlisting>
      </example>
     </para>
    </note>
    <para>
     While not being strictly an assignment by reference, expressions created
     with the language construct
     <link linkend="function.array"><literal>array()</literal></link> can also
     behave as such by prefixing <literal>&amp;</literal> to the array element
     to add. Example:
     <informalexample>
      <programlisting role="php">
<![CDATA[
<?php

$a = 1;
$b = array(2, 3);

$arr = array(&$a, &$b[0], &$b[1]);
$arr[0]++;
$arr[1]++;
$arr[2]++;
/* $a == 2, $b == array(3, 4); */

?>
]]>
      </programlisting>
     </informalexample>
    </para>
    <para>
     Note, however, that references inside arrays are potentially dangerous.
     Doing a normal (not by reference) assignment with a reference on the
     right side does not turn the left side into a reference, but references
     inside arrays are preserved in these normal assignments. This also applies
     to function calls where the array is passed by value. Example:
     <informalexample>
      <programlisting role="php">
<![CDATA[
<?php

/* Assignment of scalar variables */
$a = 1;
$b =& $a;
$c = $b;
$c = 7; // $c is not a reference; no change to $a or $b

/* Assignment of array variables */
$arr = array(1);
$a =& $arr[0]; // $a and $arr[0] are in the same reference set
$arr2 = $arr; // Not an assignment-by-reference!
$arr2[0]++;
/* $a == 2, $arr == array(2) */
/* The contents of $arr are changed even though it's not a reference! */

?>
]]>
      </programlisting>
     </informalexample>
     In other words, the reference behavior of arrays is defined in an
     element-by-element basis; the reference behavior of individual elements
     is dissociated from the reference status of the array container.
    </para>
   </sect2>
   <sect2 xml:id="language.references.whatdo.pass">
    <title>Pass By Reference</title>
    <para>
     The second thing references do is to pass variables by
     reference. This is done by making a local variable in a function
     and a variable in the calling scope referencing the same
     content. Example:
     <informalexample>
      <programlisting role="php">
<![CDATA[
<?php

function foo(&$var)
{
    $var++;
}

$a=5;
foo($a);

?>
]]>
      </programlisting>
     </informalexample>
     will make <varname>$a</varname> to be 6. This happens because in
     the function <varname>foo</varname> the variable
     <varname>$var</varname> refers to the same content as
     <varname>$a</varname>. For more information on this, read
     the <link linkend="language.references.pass">passing by
       reference</link> section.
    </para>
   </sect2>
   <sect2 xml:id="language.references.whatdo.return">
    <title>Return By Reference</title>
    <para>
     The third thing references can do is <link
     linkend="language.references.return">return by reference</link>.
    </para>
   </sect2>
  </sect1>

  <sect1 xml:id="language.references.arent">
   <title>What References Are Not</title>
   <para>
    As said before, references are not pointers. That means, the
    following construct won't do what you expect:
    <informalexample>
     <programlisting role="php">
<![CDATA[
<?php

function foo(&$var)
{
    $var =& $GLOBALS["baz"];
}

foo($bar);

?>
]]>
     </programlisting>
    </informalexample>
   </para>
   <simpara>
    What happens is that <varname>$var</varname> in
    <varname>foo</varname> will be bound with
    <varname>$bar</varname> in the caller, but then
    re-bound with <varname>$GLOBALS["baz"]</varname>. There's no way
    to bind <varname>$bar</varname> in the calling scope to something else
    using the reference mechanism, since <varname>$bar</varname> is not
    available in the function <varname>foo</varname> (it is represented by
    <varname>$var</varname>, but <varname>$var</varname> has only
    variable contents and not name-to-value binding in the calling
    <link linkend="features.gc.refcounting-basics">symbol table</link>).
    You can use <link linkend="language.references.return">returning
    references</link> to reference variables selected by the function.
   </simpara>
  </sect1>

  <sect1 xml:id="language.references.pass">
   <title>Passing by Reference</title>
   <para>
   You can pass a variable by reference to a function so the function
   can modify the variable. The syntax is as follows:
    <informalexample>
     <programlisting role="php">
<![CDATA[
<?php

function foo(&$var)
{
    $var++;
}

$a=5;

foo($a);
// $a is 6 here

?>
]]>
     </programlisting>
    </informalexample>
    <note>
     <simpara>
      There is no reference sign on a function call - only on
      function definitions. Function definitions alone are enough to
      correctly pass the argument by reference.
     </simpara>
    </note>
  </para>
  <para>
  The following things can be passed by reference:
   <itemizedlist>
    <listitem>
     <simpara>
      Variables, i.e. <literal>foo($a)</literal>
     </simpara>
    </listitem>
    <listitem>
     <para>
      References returned from functions, i.e.:
    <informalexample>
     <programlisting role="php">
<![CDATA[
<?php

function foo(&$var)
{
    $var++;
}

function &bar()
{
    $a = 5;
    return $a;
}

foo(bar());

?>
]]>
     </programlisting>
    </informalexample>
    See more about <link
    linkend="language.references.return">returning by reference</link>. 
     </para>
    </listitem>
  </itemizedlist>
  </para>
  <para>
  No other expressions should be passed by reference, as the
  result is undefined. For example, the following examples of passing
  by reference are invalid:
    <informalexample>
     <programlisting role="php">
<![CDATA[
<?php

function foo(&$var)
{
    $var++;
}

function bar() // Note the missing &
{
    $a = 5;
    return $a;
}

foo(bar()); // Produces a notice

foo($a = 5); // Expression, not variable
foo(5); // Produces fatal error

class Foobar {}

foo(new Foobar()) // Produces a notice as of PHP 7.0.7
                  // Notice: Only variables should be passed by reference

?>
]]>
     </programlisting>
    </informalexample>
   </para>
  </sect1>

  <sect1 xml:id="language.references.return">
   <title>Returning References</title>
   <para>
    Returning by reference is useful when you want to use a function
    to find to which variable a reference should be bound. Do
    <emphasis>not</emphasis> use return-by-reference to increase performance.
    The engine will automatically optimize this on its own. Only return
    references when you have a valid technical reason to do so. To
    return references, use this syntax:
    <informalexample>
     <programlisting role="php">
<![CDATA[
<?php

class Foo
{
    public $value = 42;

    public function &getValue()
    {
        return $this->value;
    }
}

$obj = new Foo();
$myValue = &$obj->getValue(); // $myValue is a reference to $obj->value, which is 42
$obj->value = 2;
echo $myValue;                // Prints the new value of $obj->value, i.e. 2

?> 
]]>
     </programlisting>
    </informalexample>
    In this example, the property of the object returned by the
    <varname>getValue</varname> function would be set, not the
    copy, as it would be without using reference syntax.
   </para>
   <note>
    <simpara>
     Unlike parameter passing, here you have to use
     <literal>&amp;</literal> in both places - to indicate that you
     want to return by reference, not a copy, and to indicate that
     reference binding, rather than usual assignment, should be done
     for <varname>$myValue</varname>.
    </simpara>
   </note>
   <note>
    <simpara>
     If you try to return a reference from a function with the syntax:
     <literal>return ($this->value);</literal> this will <emphasis>not</emphasis>
     work as you are attempting to return the result of an
     <emphasis>expression</emphasis>, and not a variable, by reference. You can
     only return variables by reference from a function - nothing else.
    </simpara>
   </note>
   <para>
   To use the returned reference, you must use reference assignment:
    <informalexample>
     <programlisting role="php">
<![CDATA[
<?php

function &collector()
{
    static $collection = array();
    return $collection;
}

$collection = &collector();
// Now the $collection is a referenced variable that references the static array inside the function

$collection[] = 'foo';

print_r(collector());
// Array
// (
//    [0] => foo
// )

?> 
]]>
     </programlisting>
    </informalexample>
   <note>
    <simpara>
     If the assignment is done without the <literal>&amp;</literal> symbol, e.g. <code>$collection = collector();</code>,
     the <varname>$collection</varname> variable will receive a copy of the value, not the reference returned by the function.
    </simpara>
   </note>
   To pass the returned reference to another function expecting a reference
   you can use this syntax:
    <informalexample>
     <programlisting role="php">
<![CDATA[
<?php

function &collector()
{
    static $collection = array();
    return $collection;
}

array_push(collector(), 'foo');

?> 
]]>
     </programlisting>
    </informalexample>
   </para>
   <note>
    <simpara>
      Note that <literal>array_push(&amp;collector(), 'foo');</literal> will
      <emphasis>not</emphasis> work, it results in a fatal error.
    </simpara>
   </note>
  </sect1>

  <sect1 xml:id="language.references.unset">
   <title>Unsetting References</title>
   <para>
    When you unset the reference, you just break the binding between
    variable name and variable content. This does not mean that
    variable content will be destroyed. For example:
    <informalexample>
     <programlisting role="php">
<![CDATA[
<?php

$a = 1;
$b =& $a;
unset($a); 

?>
]]>
     </programlisting>
    </informalexample>
    won't unset <varname>$b</varname>, just <varname>$a</varname>. 
   </para>
   <simpara>
    Again, it might be useful to think about this as analogous to the Unix
    <command>unlink</command> call.
   </simpara>
  </sect1>

  <sect1 xml:id="language.references.spot">
   <title>Spotting References</title>
   <simpara>
    Many syntax constructs in PHP are implemented via referencing
    mechanisms, so everything mentioned herein about reference binding also
    applies to these constructs. Some constructs, like passing and
    returning by reference, are mentioned above. Other constructs that
    use references are:
   </simpara>

   <sect2 xml:id="references.global">
    <title>global References</title>
    <para>
     When you declare a variable as <command>global $var</command> you
     are in fact creating reference to a global variable. That means,
     this is the same as:
     <informalexample>
      <programlisting role="php">
<![CDATA[
<?php

$var =& $GLOBALS["var"];

?>
]]>
      </programlisting>
     </informalexample>
    </para>
    <simpara>
     This also means that unsetting <varname>$var</varname>
     won't unset the global variable.
    </simpara>
   </sect2>
  </sect1>

 </chapter>
 
<!-- Keep this comment at the end of the file
Local variables:
mode: sgml
sgml-omittag:t
sgml-shorttag:t
sgml-minimize-attributes:nil
sgml-always-quote-attributes:t
sgml-indent-step:1
sgml-indent-data:t
indent-tabs-mode:nil
sgml-parent-document:nil
sgml-default-dtd-file:"~/.phpdoc/manual.ced"
sgml-exposed-tags:nil
sgml-local-catalogs:nil
sgml-local-ecat-files:nil
End:
vim600: syn=xml fen fdm=syntax fdl=2 si
vim: et tw=78 syn=sgml
vi: ts=1 sw=1
-->