PHP array comparison algorithm

Question

While trying to simulate a bit of PHP behaviour I stumbled across this:

    $a=array(0 => 1, 'test' => 2);
    $b=array('test' => 3, 0 => 1);
    var_dump($a==$b, $a>$b, $b>$a);

According to the output from var_dump $b is bigger than $a. In the PHP manual there is a Transcription of standard array comparison which states that the values of the arrays are compared one by one and if a key from the first array is missing in the second array, the arrays are uncomparable. So far so good. But if I try this (change in the first element of $a only):

    $a=array(0 => 2, 'test' => 2);
    $b=array('test' => 3, 0 => 1);
    var_dump($a==$b, $a>$b, $b>$a);

All three comparison results are false. This looks like "uncomparable" to me (because the > result is the same as the < result, while the arrays are not ==either, which makes no sense) but this does not fit the transcription from the PHP manual. Both keys are present in both arrays and I would expect $a to be bigger this time because the content of key 0 is bigger in $a (2 vs. 1).

I've tried to dig into the PHP source code and found zend_hash_compare() in zend_hash.c, but the code there seems to work as the manual describes.

What's going on here?

Answer 1

EDIT: As Joachim has shown, it deals with the order called. To steal his words: "$a>$b loops over b and finds 'test' first. 'test' is greater in $b so $b is greater and it returns false. $b>$a loops over a and finds '0' first. '0' is greater in $a so $a is greater and it returns false."

-- Original Post --

I'm not 100% sure I'm right on this; I haven't seen this before, and have only briefly looked into it (major kudos, by the way, on an excellent question!). Anyway, it would appear that either PHP documentation is wrong, or this is a bug (in which case you might want to submit it), and here is why:

in zend_hash_compare() in zend_hash.c, it seems as though there is some confusion over what ordered is (I'm looking at line 1514 and 1552-1561, which is my best guess is where the problem is, without doing lots of testing).

Here's what I mean; try this:

$a=array(0 => 2, 'test' => 2);
$b=array(0 => 1, 'test' => 3);
var_dump($a==$b, $a>$b, $b>$a);

Note I merely switched the order of indexes, and $a>$b returns true. Also see this:

$x=array(0 => 2, 'test' => 2);
$y = $x;
$y[0] = 1; $y['test'] = 3;
var_dump($x==$y, $x>$y, $y>$x);

Note here, as well, $x>$y returns true. In other words, PHP is not just matching array keys! It cares about the order of those keys in the arrays! You can prevent this situation by coming up with a "base" array and "copying" it into new variables (in my x/y example) before modifying, or you can create an object, if you so desire.

To say all that differently, and much more briefly, it would appear that PHP is not just looking at key values, but at both key values AND key order.

Again, I emphasize I don't know if this expected behavior (it seems like something they ought to have noted in the PHP manual if it was), or a bug/error/etc (which seems much more likely to me). But either way, I'm finding that it is compared first by number of keys (lines 1496-1501 in zend_hash.c), and then by both key value and key order.