str_iter のベンチマーク

01-README.md

str_iter のベンチマーク

https://github.com/masakielastic/php-ext-striter

str_iter の実装には PCRE2 JIT が使われています。

php -d extension=modules/striter.so  benchmark.php

[ASCII]
grapheme_*                    :   0.003695 sec, mem: 0 bytes
IntlBreakIterator             :   0.161382 sec, mem: 0 bytes
preg_match_all によるユーザー関数:   0.003848 sec, mem: 0 bytes
preg_match によるユーザー関数:   0.006448 sec, mem: 0 bytes
str_iter拡張                :   0.012894 sec, mem: 0 bytes

[日本語]
grapheme_*                    :   0.120388 sec, mem: 0 bytes
IntlBreakIterator             :   0.156671 sec, mem: 0 bytes
preg_match_all によるユーザー関数:   0.009735 sec, mem: 0 bytes
preg_match によるユーザー関数:   0.020445 sec, mem: 0 bytes
str_iter拡張                :   0.115297 sec, mem: 0 bytes

[絵文字]
grapheme_*                    :   0.064137 sec, mem: 0 bytes
IntlBreakIterator             :   0.153531 sec, mem: 0 bytes
preg_match_all によるユーザー関数:   0.003629 sec, mem: 0 bytes
preg_match によるユーザー関数:   0.004293 sec, mem: 0 bytes
str_iter拡張                :   0.008510 sec, mem: 0 bytes

[混在]
grapheme_*                    :   0.091253 sec, mem: 0 bytes
IntlBreakIterator             :   0.162230 sec, mem: 0 bytes
preg_match_all によるユーザー関数:   0.007549 sec, mem: 0 bytes
preg_match によるユーザー関数:   0.016544 sec, mem: 0 bytes
str_iter拡張                :   0.063441 sec, mem: 0 bytes

02-benchmark.php

<?php

function foreach_grapheme($str) {
    $len = grapheme_strlen($str);
    $chars = [];
    for ($i = 0; $i < $len; $i++) {
        $char = grapheme_substr($str, $i, 1);
        $chars[] = $char; // ダミー操作
    }
}

function foreach_intl($str) {
    $bi = IntlBreakIterator::createCharacterInstance('UTF-8');
    $bi->setText($str);
    $chars = [];
    $prev = 0;
    foreach ($bi as $pos) {
        if ($pos > $prev) {
            $char = substr($str, $prev, $pos - $prev);
            $chars[] = $char; // ダミー操作（例: 配列にpush）
            $prev = $pos;
        }
    }
}

function foreach_pcre($str) {
    preg_match_all('/\X/u', $str, $m);
    $chars = [];
    foreach ($m[0] as $char) {
        $chars[] = $char; // ダミー操作
    }
}

function foreach_pregmatch($str) {
    $offset = 0;
    $chars = [];
    $len = strlen($str);
    while ($offset < $len) {
        if (preg_match('/\X/u', $str, $m, 0, $offset)) {
            $chars[] = $m[0]; // ダミー操作
            $offset += strlen($m[0]);
        } else {
            break;
        }
    }
}

function foreach_striter($str) {
    $chars = [];
    foreach (str_iter($str, 'grapheme') as $char) {
        $chars[] = $char; // ダミー操作
    }
}

function bench($fn, $label, $repeat = 10000) {
    $start = microtime(true);
    $mem_start = memory_get_usage();
    for ($i = 0; $i < $repeat; $i++) {
        $fn();
    }
    $mem_end = memory_get_usage();
    $elapsed = microtime(true) - $start;
    printf("%-30s: %10.6f sec, mem: %d bytes\n", $label, $elapsed, $mem_end - $mem_start);
}

// サンプル入力
$inputs = [
    "ASCII"      => "abcde",
    "日本語"      => "あいうえお漢字カタカナひらがな",
    "絵文字"      => "😀🍣🍺🏳️‍🌈👨‍👩‍👧‍👦🇯🇵",
    "混在"        => "abcあいう🍺1２３👩‍🔬",
];

foreach ($inputs as $label => $text) {
    echo "[$label]\n";
    bench(fn() => foreach_grapheme($text), "grapheme_*");
    bench(fn() => foreach_intl($text), "IntlBreakIterator");
    bench(fn() => foreach_pcre($text), "preg_match_all によるユーザー関数");
    bench(fn() => foreach_pregmatch($text), "preg_match によるユーザー関数");
    bench(fn() => foreach_striter($text), "str_iter拡張");
    echo "\n";
}