[bench] Add FileInfoMatcher::matchPattern realpath benchmark (#8703)

Author: TomasVotruba

bench/file_info_matcher_bench.php

@@ -0,0 +1,79 @@
+<?php
+
+declare(strict_types=1);
+
+/**
+ * Benchmark for FileInfoMatcher::matchPattern().
+ *
+ * A skip pattern that is not a literal prefix/suffix and does not fnmatch falls through to
+ * matchRealpath(), which calls realpath() on the pattern. PHP does not cache a *failed*
+ * realpath() lookup, so a wildcard pattern (realpath = false) pays a filesystem syscall for
+ * every single file checked. On a large run that is one failed realpath() per file per pattern.
+ *
+ * This bench drives the real matchPattern() over many real files with a set of such patterns,
+ * so every call reaches the realpath() lookup. The patterns never match, so matchPattern()
+ * returns null and does no other work, making the calls repeatable.
+ *
+ * Run on both branches to compare:
+ *   php bench/file_info_matcher_bench.php
+ *
+ * Reports CPU time (user+sys) and wall time; CPU is the reliable metric, wall is noisy
+ * under concurrent load.
+ */
+
+use Rector\Skipper\Matcher\FileInfoMatcher;
+
+require __DIR__ . '/../vendor/autoload.php';
+
+const REPEAT = 40; // outer repeats over the whole file set
+
+$fileInfoMatcher = new FileInfoMatcher();
+
+// many real files, so each pattern's realpath() lookup is repeated per file within a run
+$filePaths = array_values(array_filter(
+    glob(__DIR__ . '/../src/*/*.php') ?: [],
+    static fn (string $filePath): bool => is_file($filePath)
+));
+
+// wildcard patterns whose realpath() is always false (nonexistent dirs) - the uncached hot case
+$filePatterns = [
+    '*/__no_such_dir_a__/*',
+    '*/__no_such_dir_b__/*',
+    '*/__no_such_dir_c__/*',
+];
+
+$cpuSeconds = static function (): float {
+    $usage = getrusage();
+    return $usage['ru_utime.tv_sec'] + $usage['ru_utime.tv_usec'] / 1_000_000
+        + $usage['ru_stime.tv_sec'] + $usage['ru_stime.tv_usec'] / 1_000_000;
+};
+
+$run = static function () use ($fileInfoMatcher, $filePaths, $filePatterns): void {
+    foreach ($filePaths as $filePath) {
+        $fileInfoMatcher->matchPattern($filePath, $filePatterns);
+    }
+};
+
+$measure = static function () use ($run, $cpuSeconds): array {
+    // warm up autoloading once, excluded from timing
+    $run();
+
+    $cpuStart = $cpuSeconds();
+    $wallStart = hrtime(true);
+
+    for ($repeat = 0; $repeat < REPEAT; ++$repeat) {
+        $run();
+    }
+
+    return [
+        'cpu' => $cpuSeconds() - $cpuStart,
+        'wall' => (hrtime(true) - $wallStart) / 1_000_000_000,
+    ];
+};
+
+$result = $measure();
+
+printf('FileInfoMatcher::matchPattern benchmark%s', PHP_EOL);
+printf('  files=%d patterns=%d repeat=%d%s', count($filePaths), count($filePatterns), REPEAT, PHP_EOL);
+printf('%s', PHP_EOL);
+printf('  matchPattern  cpu=%.4fs  wall=%.4fs%s', $result['cpu'], $result['wall'], PHP_EOL);