Question

The code below illustrates the strange behaviour of PHP references:

<?php

function this_works()
{
    $root = array('name'=>'root', 'children'=>array());
    $level_1 = array('name'=>'level_1', 'children'=>array());
    $item1 = array('name'=>'level_2_1', 'children'=>array());
    $item2 = array('name'=>'level_2_2', 'children'=>array());

    $croot = &$root;

    $croot['children'][] = &$level_1;

    $croot = &$level_1;

    $croot['children'][] = &$item1;
    $croot['children'][] = &$item2;

    $croot = &$root;

    print_r($croot);
}    

function this_fails()
{    
    $root = array('name'=>'root', 'children'=>array());
    $level_1 = array('name'=>'level_1', 'children'=>array());
    $item1 = array('name'=>'level_2_1', 'children'=>array());
    $item2 = array('name'=>'level_2_2', 'children'=>array());
    $croot = &$root;

    $stack = array();

    $croot['children'][] = &$level_1;
    $crootref = &$croot;

    array_push($stack, $crootref);

    $croot = &$level_1;

    $croot['children'][] = &$item1;
    $croot['children'][] = &$item2;

    # this works, assignment below - doesn't... WHY?
    #$x = array_pop($stack);
    #var_dump($x);

    $croot = array_pop($stack);

    print_r($croot);
}    

this_works();
echo "------------------\n";
this_fails();

?>

First function provides expected results,
while the second fails and claims about recursion loop:

Array
(
    [name] => root
    [children] => Array
        (
            [0] => Array
                (
                    [name] => level_1
                    [children] => Array
                        (
                            [0] => Array
                                (
                                    [name] => level_2_1
                                    [children] => Array
                                    (
                                    )

                                )

                            [1] => Array
                                (
                                    [name] => level_2_2
                                    [children] => Array
                                        (
                                        )

                                )

                           )

                )

        )

)
------------------
Array
(
    [name] => root
    [children] => Array
        (
            [0] => Array
                (
                    [name] => root
                    [children] => Array
 *RECURSION*
                )

    )

)

What is strange, is that if in the second function, intermediate
variable will be used to get value from stack, results are OK again.
I don’t understand what is going on. How do I get root element as
a child of itself many times due to one assinment?

Originally, I needed to build the tree from XML (using sax parser)
and intented to have ‘current root’ that points to tree node at
the current level and push/pop it to/from stack and add child elements to it,
but, surprisingly, I failed to implement this scheme due to issues demonstrated
by two functions above.

So, what is wrong with such approach?

Answer 1

One picture worths 1000 words.
It took some time to understand
what’s exactly is going on.

I had to use Xdebug to dump internal
data properly and see refcounts
and effects of copy on write.

The issue with the code in 1st post
is that assignment

$croot = array_pop($stack);

when croot is ‘level_1’ is done by copy,
i.e. elements of croot (same as level_1)
are populated with data from stack and
croot is not the same as original root after this operation.

The image will explain better.