Performance optimizations

[GromNaN]

• Use use function to enable compiler optimization of some native functions.
• Use preg_match('/(.).*\1/', $alphabet) to validate there is no duplicate char, faster than splitting into an array
• Avoid splitting strings, and access chars directly with $string[$char]
• Avoid functions with callback (array_reduce, array_filter), they are slower than foreach

Benchmark

PHPBench code

composer req --dev phpbench/phpbench

In phpbench.json

{
    "$schema": "./vendor/phpbench/phpbench/phpbench.schema.json",
    "runner.bootstrap": "vendor/autoload.php",
    "runner.file_pattern": "*Bench.php",
    "runner.path": "tests",
    "runner.iterations": 3
}

In tests/SqidsBench.php

<?php

namespace Sqids\Benchmark;

use PhpBench\Attributes\ParamProviders;
use PhpBench\Attributes\Revs;
use PhpBench\Attributes\Warmup;
use Sqids\Sqids;

#[Warmup(1)]
final class SqidsBench
{
    private const IDS = [
        'SvIzsqYMyQwI3GWgJAe17URxX8V924Co0DaTZLtFjHriEn5bPhcSkfmvOslpBu' => [0, 0],
        'n3qafPOLKdfHpuNw3M61r95svbeJGk7aAEgYn4WlSjXURmF8IDqZBy0CT2VxQc' => [0, 1],
        'tryFJbWcFMiYPg8sASm51uIV93GXTnvRzyfLleh06CpodJD42B7OraKtkQNxUZ' => [0, 2],
        'eg6ql0A3XmvPoCzMlB6DraNGcWSIy5VR8iYup2Qk4tjZFKe1hbwfgHdUTsnLqE' => [0, 3],
        'rSCFlp0rB2inEljaRdxKt7FkIbODSf8wYgTsZM1HL9JzN35cyoqueUvVWCm4hX' => [0, 4],
        'sR8xjC8WQkOwo74PnglH1YFdTI0eaf56RGVSitzbjuZ3shNUXBrqLxEJyAmKv2' => [0, 5],
        'uY2MYFqCLpgx5XQcjdtZK286AwWV7IBGEfuS9yTmbJvkzoUPeYRHr4iDs3naN0' => [0, 6],
        '74dID7X28VLQhBlnGmjZrec5wTA1fqpWtK4YkaoEIM9SRNiC3gUJH0OFvsPDdy' => [0, 7],
        '30WXpesPhgKiEI5RHTY7xbB1GnytJvXOl2p0AcUjdF6waZDo9Qk8VLzMuWrqCS' => [0, 8],
        'moxr3HqLAK0GsTND6jowfZz3SUx7cQ8aC54Pl1RbIvFXmEJuBMYVeW9yrdOtin' => [0, 9],
        'JSwXFaosANEOuLlYb3jHCBpeSzx7cPRrgf1dNTZqE4nDytU09isA5ahm6kKGvM' => [1_000_000, 2_000_000],
    ];

    #[Revs(1_000)]
    #[ParamProviders('provideSqids')]
    public static function benchEncodeDecode(array $params): void
    {
        foreach (self::IDS as $id => $numbers) {
            $params[0]->encode($numbers);
            $params[0]->decode($id);
        }
    }

    public static function provideSqids(): \Generator
    {
        yield 'default' => [
            new Sqids()
        ];
        yield 'custom blocklist' => [
            new Sqids(blocklist: [
                'JSwXFaosAN',
                'OCjV9JK64o',
                'rBHf',
                '79SM',
                '7tE6',
            ])
        ];
        yield 'no blocklist' => [
            new Sqids(blocklist: [])
        ];
    }
}

Before

    benchEncodeDecode # default.............I2 - Mo2.195ms (±0.47%)
    benchEncodeDecode # custom blocklist....I2 - Mo1.099ms (±0.84%)
    benchEncodeDecode # no blocklist........I2 - Mo929.423μs (±0.87%)

After

    benchEncodeDecode # default.............I2 - Mo994.363μs (±0.35%)
    benchEncodeDecode # custom blocklist....I2 - Mo681.353μs (±0.36%)
    benchEncodeDecode # no blocklist........I2 - Mo552.545μs (±0.21%)

[vinkla]

This looks good to me. I'll leave it up to @4kimov to merge this who has a deeper knowledge across our language implementations.

[GromNaN]

I can understand that it's a big PR and that you might have doubts about the merge. I've added comments to explain the changes. Given the performance gain, I think it's important to look into it.

[GromNaN]

Type is already validated in the arg type.

[GromNaN]

Instead of creating a new array, the exception is thrown directly when there is an invalid number.

[GromNaN]

This split operation is replaced by a more efficient regex.

[GromNaN]

Manipulation of individual characters of the string, instead of using an array of chars.

[GromNaN]

Reuse and modify the variable $num instead of creating a new one.

[GromNaN]

Appending to the end of the string is the same as using array_unshift on an array.

[4kimov]

Well, to say that these optimizations are welcome would be an understatement. Thank you for taking the time!

A few basic questions as I look at this:

1. Are there any breaking changes in this PR or Optimize performance of blocklist filtering and checking by using Regex #17?
2. With new regex expressions, is there any character that a user might use in the alphabet to mess up regex matching?

[GromNaN]

1. Are there any breaking changes in this PR or [Optimize performance of blocklist filtering and checking by using Regex #17](https://github.com/sqids/sqids-php/pull/17)?

After multiple reviews, I don't see any breaking change in this PR.

In #17, there is something negligible with the $blocklist property, and maybe if someone customized the blocklist without adding all the same leet variations of the words.

2. With new regex expressions, is there any character that a user might use in the alphabet to mess up regex matching?

Regexes operate on bytes; unicode characters are split into bytes and not considered as a single character. But this was already the case with str_split. Unicode is already not supported.

There is no other restriction in the characters accepted by the alphabet. The alphabet is not used as part of the regex, and if it was, I would have used preg_quote to escape special chars. I escaped the blocklist works for this purpose.