[php-src] Issue #20543: Pipe operators are twice slower than their equivalents
| From: | kamil-tekiela | Date: | Thu, 20 Nov 2025 19:25:24 +0000 |
| Subject: | [php-src] Issue #20543: Pipe operators are twice slower than their equivalents | ||
| Groups: | php.bugs | ||
| Request: | Send a blank email to php-bugs+get-251262@lists.php.net to get a copy of this message | ||
Issue: https://github.com/php/php-src/issues/20543
Author: kamil-tekiela
### Description
The following code:
```php
<?php
$arr = [
new Widget(tags: ['a', 'b', 'c']),
new Widget(tags: ['c', 'd', 'e']),
new Widget(tags: ['x', 'y', 'a']),
];
class Widget {
public function __construct(
public array $tags = [],
) {}
}
class Benchmark {
private $startTime;
private $endTime;
public function start() {
$this->startTime = hrtime(true);
}
public function end() {
$this->endTime = hrtime(true);
}
public function getElapsedTime() {
return ($this->endTime - $this->startTime)/1_000_000_000;
}
}
// Example usage
$benchmark = new Benchmark();
$trimClosure = trim(...);
$strtolowerClosure = strtolower(...);
// ====================================
$benchmark->start();
// Code to benchmark
for ($i = 0; $i < 1000000; $i++) {
$title = ' PHP 8.5 Released ';
$slug1 = $title
|> $trimClosure
|> $strtolowerClosure;
// string(15) "php-85-released"
}
$benchmark->end();
echo "Elapsed time: " . $benchmark->getElapsedTime() . " seconds\n";
$benchmark->start();
// Code to benchmark
for ($i = 0; $i < 1000000; $i++) {
$title = ' PHP 8.5 Released ';
$slug2 = $strtolowerClosure(
$trimClosure($title)
);
// string(15) "php-85-released"
}
$benchmark->end();
echo "Elapsed time: " . $benchmark->getElapsedTime() . " seconds\n";
var_dump($slug1 === $slug2);
```
Resulted in this output:
```
Elapsed time: 0.09693445 seconds
Elapsed time: 0.059030439 seconds
bool(true)
```
But I expected this output instead:
```
Elapsed time: 0.059030439 seconds
Elapsed time: 0.059030439 seconds
bool(true)
```
I thought that the slowdown came from preparing the closures in the PHP manual, but I am pretty sure
I have elemited all extra overhead from the benchmark above and the timings haven't changed. It
seems that the pipe operator adds a lot of overhead itself.
### PHP Version
```plain
PHP 8.5.0 as tested on https://3v4l.org/uu0Q5#v8.5.0
```
### Operating System
_No response_