PHP DOM XPath Test

bench.php

<?php
$xml = new DOMDocument();


$xml->loadXML('
<a xmlns="http://www.aaa.bbb.cc/xx">
    <b/>
    <b xmlns:x="http://www.aaa.bbb.cc/xxyy">
        <c xml:id="c" aa="a" bb="bb" xmlns:c="http://www.aaa.bbb.cc/xx">
            <c:d/>
        </c>
    </b>
    <b/>
    <b/>
</a>
');


$e = $xml->getElementById("c");

$n = 10000;

$t = microtime(1);
echo "Iterate over attributes: ";
for ($i = 0; $i<$n; $i++){
    foreach($e->attributes as $att){
        $att->nodeName;
    }
}
$t0 = microtime(1)-$t;
echo round($t0,4);
echo "\n\n";

$t = microtime(1);
echo "Iterate get each attribute: ";
for ($i = 0; $i<$n; $i++){
    foreach($e->attributes as $att){
        $e->getAttribute($att->nodeName);
    }
}
$t3 = microtime(1)-$t;
echo round($t3,4)."s\n\t";
echo round($t3/$t0)."x slower than iteration over attributes\n\n";

$t = microtime(1);
echo "Reuse DOMXPath instance: ";
$xp = new DOMXPath($xml);
for ($i = 0; $i<$n; $i++){
    foreach($xp->query('namespace::*', $e, false) as $att){
        $att->nodeName;
    }
}
$t2 = microtime(1)-$t;
echo round($t2,4)."s\n\t";
echo round($t2/$t0)."x slower than iteration over attributes\n\n";


$t = microtime(1);
echo "New DOMXPath instance: ";
for ($i = 0; $i<$n; $i++){
    $xp = new DOMXPath($xml);
    foreach($xp->query('namespace::*', $e, false) as $att){
        $att->nodeName;
    }
}
$t1 = microtime(1)-$t;
echo round($t1,4)."s\n\t";
echo round($t1/$t0)."x slower than iteration over attributes\n\n";

results.txt

Iterate over attributes: 0.0179

Iterate get each attribute: 0.043s
	2x slower than iteration over attributes

Reuse DOMXPath instance: 0.0802s
	4x slower than iteration over attributes

New DOMXPath instance: 0.1584s
	9x slower than iteration over attributes

I would be curious whether the performance ratio stayed the same for a larger DOM. If some of the XPath overhead comes from compiling the XPath, then this might even out with an average or large DOM.