PHP version 7.4.5 is running on this server.
Please note that these are micro benchmarks. Micro benchmarks are stupid. I created this comparison to learn something about PHP and how the PHP compiler works. This can not be used to compare PHP versions or servers.
| Method | Undefined | Null | False | Empty string | String "0" | String "1" | Long string | Summary | Index |
|---|---|---|---|---|---|---|---|---|---|
if ( ! $var ) | 11 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 11 ms | 3395 |
if (empty( $var) ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 259 |
if ( $var == "" ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 21 ms | 22 ms | 6756 |
if ( "" == $var ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 10 ms | 11 ms | 3254 |
if ( $var === "" ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 101 |
if ( "" === $var ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 102 |
if ( strcmp( $var, "" ) == 0 ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 234 |
if ( strcmp( "", $var ) == 0 ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 234 |
if ( strlen( $var ) == 0 ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 127 |
if ( ! strlen( $var ) ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 100 |
My conclusion:
In most cases, use empty() because it does not trigger a warning when used with undefined variables.
Note that empty( "0" ) returns true.
Use strlen() if you want to detect "0".
Try to avoid == at all because it may cause strange behaviour
(e.g. "9a" == 9 returns true).
Prefer === over == and !== over != if possible
because it does compare the variable types in addition to the contents.
| Method | Equal | First character not equal | Last character not equal | Summary | Index |
|---|---|---|---|---|---|
$a == $b | >0 ms | >0 ms | >0 ms | >0 ms | 100 |
!strcmp( $a, $b ) | >0 ms | >0 ms | >0 ms | 1 ms | 190 |
strcmp( $a, $b ) == 0 | >0 ms | >0 ms | >0 ms | 1 ms | 187 |
strcmp( $a, $b ) === 0 | >0 ms | >0 ms | >0 ms | 1 ms | 191 |
strcasecmp( $a, $b ) === 0 | >0 ms | >0 ms | 1 ms | 2 ms | 443 |
My conclusion: Use what fits your needs.
| Method | Not found | Found at the start | Found in the middle | Found at the end | Summary | Index |
|---|---|---|---|---|---|---|
strstr( $haystack, $needle ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 13892 |
strpos( $haystack, $needle ) !== FALSE | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 11333 |
strstr( $haystack, $needle ) !== FALSE | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 14542 |
stristr( $haystack, $needle ) | >0 ms | >0 ms | >0 ms | >0 ms | 2 ms | 52192 |
preg_match( "/$needle/", $haystack ) | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 22825 |
preg_match( "/$needle/i", $haystack ) | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 23183 |
preg_match( "/$needle/S", $haystack ) | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 23208 |
Function ereg is deprecated since 5.3 | 0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 100 |
My conclusion:
It does not matter if you use strstr() or strpos().
Use the preg…() functions only if you need the power of regular expressions.
Never use the ereg…() functions.
| Method | Not found | Found at the start | Found in the middle | Found at the end | Summary | Index |
|---|---|---|---|---|---|---|
strncmp( $haystack, $needle, strlen( $needle ) ) === 0 | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 135 |
strncmp( $haystack, $needle, 4 ) === 0 | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 100 |
strncasecmp( $haystack, $needle, strlen( $needle ) ) === 0 | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 101 |
strpos( $haystack, $needle ) === 0 | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 119 |
substr( $haystack, 0, strlen( $needle ) ) === $needle | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 128 |
strcmp( substr( $haystack, 0, strlen( $needle ) ), $needle ) === 0 | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 200 |
preg_match( "/^" . preg_quote( $needle, "/" ) . "/", $haystack ) | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 444 |
My conclusion:
strpos() is very fast and can be used in almost all cases.
strncmp() is good if you are looking for a constant length needle.
| Method | Not found | Found at the start | Found in the middle | Found at the end | Summary | Index |
|---|---|---|---|---|---|---|
substr( $haystack, strlen( $haystack ) - strlen( $needle) ) === $needle | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 118 |
substr( $haystack, -strlen( $needle) ) === $needle | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 100 |
strcmp( substr( $haystack, - strlen( $needle) ), $needle) === 0 | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 151 |
preg_match( "/" . preg_quote( $needle, "/" ) . "$/", $haystack ) | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 379 |
My conclusion:
Using substr() with a negative position is a good trick.
| Method | Not found | Found at the start | Found in the middle | Found at the end | Summary | Index |
|---|---|---|---|---|---|---|
str_replace( $search, $replace, $subject ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 25038 |
preg_replace( "/$search/", $replace, $subject ) | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 43888 |
preg_replace( "/$search/S", $replace, $subject ) | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 47100 |
Function ereg_replace is deprecated since 5.3 | >0 ms | 0 ms | 0 ms | >0 ms | >0 ms | 100 |
My conclusion:
Never use the ereg…() functions.
| Method | Not found | Found at start | Found at end | Found at both sides | Summary | Index |
|---|---|---|---|---|---|---|
trim( $string, "," ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 100 |
preg_replace( '/^,*|,*$/', "", $string ) | >0 ms | >0 ms | >0 ms | >0 ms | 1 ms | 5067 |
preg_replace( '/^,*|,*$/m', "", $string ) | 2 ms | 2 ms | 2 ms | 2 ms | 6 ms | 23639 |
preg_replace( '/^,+|,+$/', "", $string ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 329 |
preg_replace( '/^,+|,+$/m', "", $string ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 318 |
preg_replace( '/^,+/', "", preg_replace( '/,+$/', "", … ) ) | >0 ms | >0 ms | >0 ms | >0 ms | >0 ms | 538 |
My conclusion:
Always benchmark your regular expressions!
In this case, with .* you also replace nothing with nothing which takes time
because there is a lot of “nothing” in every string.
| Method | Empty string | Single occurrence | Multiple occurrences | Summary | Index |
|---|---|---|---|---|---|
explode( ",", $string ) | >0 ms | >0 ms | 1 ms | 1 ms | 43175 |
Function split is deprecated since 5.3 | >0 ms | >0 ms | 0 ms | >0 ms | 100 |
preg_split( "/,/", $string ) | >0 ms | >0 ms | 1 ms | 2 ms | 79262 |
preg_match_all( '/[^,]+/', $string, $matches ) | >0 ms | >0 ms | 2 ms | 3 ms | 129738 |
My conclusion:
Don't use split(). It's deprecated in PHP 5.3 and will be removed in PHP 6.
| Method | Summary | Index |
|---|---|---|
for ( $i = 0; $i < count( $array ; $i++ ) | >0 ms | 226 |
for ( $i = 0, $count = count( $array ); $i < $count; $i++ ) | >0 ms | 100 |
for ( $i = count( $array ) - 1; $i >= 0; $i-- ) | >0 ms | 119 |
for ( $i = count( $array ) - 1; $i >= 0; --$i ) | >0 ms | 113 |
$i = count( $array ); while ( $i-- ) | >0 ms | 154 |
My conclusion:
count() is horribly slow. Always precalculate it, if possible.
| Method | Summary | Index |
|---|---|---|
$array[0] | 2 ms | 109 |
$array['key'] | 2 ms | 100 |
My conclusion: I like associative arrays.
| Method | Summary | Index |
|---|---|---|
implode( " ", $array ) | >0 ms | 101 |
"$array[0] $array[1] $array[2]" | >0 ms | 101 |
$array[0] . " " . $array[1] . " " . $array[2] | >0 ms | 100 |
sprintf( "%s %s %s", $array[0], $array[1], $array[2] ) | 1 ms | 1650 |
vsprintf( "%s %s %s", $array ) | 1 ms | 1918 |
My conclusion: String concatenation is a cheap operation in PHP. Don't waste your time benchmarking this.
| Method | Summary | Index |
|---|---|---|
'contains no dollar signs' | >0 ms | 127 |
"contains no dollar signs" | >0 ms | 168 |
'$variables $are $not $replaced' | >0 ms | 100 |
"\$variables \$are \$not \$replaced" | >0 ms | 100 |
"$variables $are $replaced" | >0 ms | 100 |
$variables . ' ' . $are . ' ' . $replaced | >0 ms | 100 |
$variables . " " . $are . " " . $replaced | >0 ms | 100 |
My conclusion: It does not matter if you use single or double quotes at all. The inclusion of variables has a measurable effect, but that's independent from the quotes.