Par un détour à Yoda

I tested the rapidity of isset vs array_key_exists. I always thougt that array_key_exists would be the winner ...

Here is my script :

┌─(yoda@ev5)(14:25:29)
└─(~/var/www/test)-> cat isset_array-key-exists.php 
#!/usr/bin/php
<?php
 
define('N', "\n");
 
error_reporting(E_ALL | E_NOTICE);
 
$nbLoop = 10000;
 
$testArray = array('foo' => 'bar', 'bar' => 'quux');
 
$time = array();
 
$desc = array(
    0 => 'array_key_exists on existant key',
    1 => 'array_key_exists on non existant key',
    2 => 'isset on existant key',
    3 => 'isset on non existant key',
);
 
 
/**
 * begin with array_key_exists, existant key
 */
$time[0]['start'] = microtime(true);
 
for($i = 0; $i < $nbLoop; $i++) {
    array_key_exists('foo', $testArray);
}
 
$time[0]['stop'] = microtime(true);
 
 
/**
 * begin with array_key_exists, non existant key
 */
$time[1]['start'] = microtime(true);
 
for($i = 0; $i < $nbLoop; $i++) {
    array_key_exists('foz', $testArray);
}
 
$time[1]['stop'] = microtime(true);
 
 
 
/**
 * begin with isset, existant key
 */
$time[2]['start'] = microtime(true);
 
for($i = 0; $i < $nbLoop; $i++) {
    isset($testArray['foo']);
}
 
$time[2]['stop'] = microtime(true);
 
 
/**
 * begin with isset, non existant key
 */
$time[3]['start'] = microtime(true);
 
for($i = 0; $i < $nbLoop; $i++) {
    isset($testArray['foz']);
}
 
$time[3]['stop'] = microtime(true);
 
 
$differences = array();
echo 'results: '.N;
foreach($time as $k => $t) {
    echo 'test '.$k.' "'.sprintf('%40s', $desc[$k]).'" : '.($difference[$k] = $t['stop'] - $t['start']).N;
}

And here is the results :

┌─(yoda@ev5)(14:25:32)
└─(~/var/www/test)-> php isset_array-key-exists.php 
results: 
test 0 "        array_key_exists on existant key" : 0.00795984268188
test 1 "    array_key_exists on non existant key" : 0.00665903091431
test 2 "                   isset on existant key" : 0.00737690925598
test 3 "               isset on non existant key" : 0.00131511688232

And the winner is : isset !