Which method is preferred strstr or strpos?

Here are some other answers (+benchmarks) I got to my question, which is almost the same (I didn't realize yours when asking).

---

In the meantime I also made my own benchmark test, which I ran 1000000 times for each relevant functions (strstr(), strpos(), stristr() and stripos()).
Here's the code:

<?php

function getmicrotime() {
    list($usec, $sec) = explode(" ", microtime());
    return ((float) $usec + (float) $sec);
}

$mystring = 'blahblahblah';  
$findme = 'bla';  

echo 'strstr & strpos TEST:<pre>';
$time_start = getmicrotime();
for($i=0; $i<1000000; $i++) strstr($mystring, $findme);
$time_needed_strstr = getmicrotime() - $time_start;
echo 'strstr():            ',
    round( $time_needed_strstr , 8 ). PHP_EOL;

$time_start = getmicrotime();
for($i=0; $i<1000000; $i++) stristr($mystring, $findme);
$time_needed_stristr = getmicrotime() - $time_start;
echo 'stristr():           ',
    round( $time_needed_stristr , 8 ) . PHP_EOL;

$time_start = getmicrotime();
for($i=0; $i<1000000; $i++) strpos($mystring, $findme) !== false;
$time_needed_strpos = getmicrotime() - $time_start;
echo 'strpos() !== false:  ',
    round( $time_needed_strpos , 8 ) . PHP_EOL;

$time_start = getmicrotime();
for($i=0; $i<1000000; $i++) stripos($mystring, $findme) !== false;
$time_needed_stripos = getmicrotime() - $time_start;
echo 'stripos() !== false: ',
    round( $time_needed_stripos , 8 ) . PHP_EOL;

echo PHP_EOL;

echo 'time_needed_stristr - time_needed_strstr: ',
     round( $time_needed_stristr - $time_needed_strstr , 8) . PHP_EOL;
echo 'time_needed_stripos - time_needed_strpos: ',
     round( $time_needed_stripos - $time_needed_strpos , 8) . PHP_EOL;

echo PHP_EOL;

echo 'time_needed_strstr  - time_needed_strpos:  ',
     round( $time_needed_strstr - $time_needed_strpos , 8) . PHP_EOL;
echo 'time_needed_stristr - time_needed_stripos: ',
     round( $time_needed_stristr - $time_needed_stripos , 8) . PHP_EOL;

echo '</pre>';

?>

---

And here is the first output, which shows that strpos() is the winner:

strstr & strpos TEST:
strstr():            2.39144707
stristr():           3.65685797
strpos() !== false:  2.39055395
stripos() !== false: 3.54681897

time_needed_stristr - time_needed_strstr: 1.2654109
time_needed_stripos - time_needed_strpos: 1.15626502

time_needed_strstr  - time_needed_strpos:  0.00089312
time_needed_stristr - time_needed_stripos: 0.110039 

The next one is similar to the first output (strpos() is the winner again):

strstr & strpos TEST:
strstr():            2.39969015
stristr():           3.60772395
strpos() !== false:  2.38610101
stripos() !== false: 3.34951186

time_needed_stristr - time_needed_strstr: 1.2080338
time_needed_stripos - time_needed_strpos: 0.96341085

time_needed_strstr  - time_needed_strpos:  0.01358914
time_needed_stristr - time_needed_stripos: 0.25821209

Below is another one, which is more interesting, because in this case, strstr() is the winner:

strstr & strpos TEST:
strstr():            2.35499191
stristr():           3.60589004
strpos() !== false:  2.37646604
stripos() !== false: 3.51773095

time_needed_stristr - time_needed_strstr: 1.25089812
time_needed_stripos - time_needed_strpos: 1.14126492

time_needed_strstr  - time_needed_strpos:  -0.02147412
time_needed_stristr - time_needed_stripos: 0.08815908

This means it can really depend on "environmental circumstances", which are sometimes hard to influence, and can change the result of "micro optimization tasks" like this, in case you are just checking whether a string exists in another one or not.

BUT I think in most cases, strpos() is the winner in comparison to strstr().

I hope this test was useful for someone.

From the PHP online manual:

If you only want to determine if a particular needle occurs within haystack, use the faster and less memory intensive function strpos() instead.

Many developers use strpos for micro optimization purposes.

Using strstr also only works if the resulting string cannot be interpreted as false in boolean context.

Overcome by events: PHP8 introduced str_contains, the "right tool for the job" (with shims available for older setups). Which does exactly what everyone has been glamoring for, but without the interpreter-level comparison and syntactic overhead.