Merge pull request #1 from ENM1989/feature/optimize-array-combine

Optimize array_combine with a fast path and fix memory leak

Author: ENM1989

benchmark.php

@@ -0,0 +1,26 @@
+<?php
+
+function benchmark_strtoupper(string $str) {
+    $start = microtime(true);
+    for ($i = 0; $i < 100000; $i++) {
+        strtoupper($str);
+    }
+    $end = microtime(true);
+    return $end - $start;
+}
+
+$short_ascii = 'abcdefghijklmnopqrstuvwxyz';
+$long_ascii = str_repeat($short_ascii, 1000);
+$short_mixed = 'aBcDeFgHiJkLmNoPqRsTuVwXyZ';
+$long_mixed = str_repeat($short_mixed, 1000);
+$short_upper = 'ABCDEFGHIJKLMNOPQRSTUVWXYZ';
+$long_upper = str_repeat($short_upper, 1000);
+
+echo "Benchmarking current strtoupper implementation...\n";
+
+echo "Short ASCII string: " . benchmark_strtoupper($short_ascii) . "s\n";
+echo "Long ASCII string: " . benchmark_strtoupper($long_ascii) . "s\n";
+echo "Short mixed-case string: " . benchmark_strtoupper($short_mixed) . "s\n";
+echo "Long mixed-case string: " . benchmark_strtoupper($long_mixed) . "s\n";
+echo "Short uppercase string: " . benchmark_strtoupper($short_upper) . "s\n";
+echo "Long uppercase string: " . benchmark_strtoupper($long_upper) . "s\n";

benchmark_array_combine.php

@@ -0,0 +1,38 @@
+<?php
+
+function benchmark($name, $func) {
+    $start = microtime(true);
+    // Run the function multiple times to get a more stable reading
+    for ($i = 0; $i < 10; $i++) {
+        $func();
+    }
+    $end = microtime(true);
+    printf("%-40s: %8.4f seconds\n", $name, ($end - $start));
+}
+
+const LARGE_SIZE = 1000000;
+
+// --- Scenario 1: Packed Indexed Arrays (Target for optimization) ---
+$indexed_keys = range(0, LARGE_SIZE - 1);
+$indexed_values = range(0, LARGE_SIZE - 1);
+
+benchmark("Packed Indexed Arrays", function() use ($indexed_keys, $indexed_values) {
+    $a = array_combine($indexed_keys, $indexed_values);
+});
+
+
+// --- Scenario 2: Associative Arrays (Check for regressions) ---
+$assoc_keys = [];
+$assoc_values = [];
+for ($i = 0; $i < LARGE_SIZE; $i++) {
+    $assoc_keys[] = "key_" . $i;
+    $assoc_values[] = $i;
+}
+
+benchmark("Associative Arrays", function() use ($assoc_keys, $assoc_values) {
+    $a = array_combine($assoc_keys, $assoc_values);
+});
+
+echo "\n";
+
+?>

benchmark_grapheme_str_split.php

@@ -0,0 +1,48 @@
+<?php
+
+function benchmark($name, $func) {
+    // A relatively low iteration count because grapheme operations can be intensive.
+    $iterations = 500;
+    $start = microtime(true);
+    for ($i = 0; $i < $iterations; $i++) {
+        $func();
+    }
+    $end = microtime(true);
+    echo str_pad($name, 45) . ": " . number_format($end - $start, 6) . "s\n";
+}
+
+echo "Benchmarking current grapheme_str_split implementation...\n";
+
+// A long string (~50KB) mixing ASCII, multi-byte, and complex graphemes.
+$long_string = str_repeat("Hello world! Это тест. The quick brown 👨‍👩‍👧‍👦 fox. Á, B́, Ć.", 200);
+
+// ===== TEST CASES =====
+
+// Case 1: Simple ASCII string
+benchmark("Simple ASCII string", function() {
+    grapheme_str_split("abcdefghijklmnopqrstuvwxyz");
+});
+
+// Case 2: Multi-byte UTF-8 string (Cyrillic)
+benchmark("Multi-byte UTF-8 string (Cyrillic)", function() {
+    grapheme_str_split("абвгдеёжзийклмнопрстуфхцчшщъыьэюя");
+});
+
+// Case 3: Complex Graphemes (Combining Marks)
+// 'e' with 3 combining marks is one grapheme
+benchmark("Complex Graphemes (Combining Marks)", function() {
+    grapheme_str_split("é̄̃");
+});
+
+// Case 4: Complex Graphemes (Emoji)
+// Family emoji and woman with skin tone modifier are single graphemes
+benchmark("Complex Graphemes (Emoji)", function() {
+    grapheme_str_split("👨‍👩‍👧‍👦👩🏽‍💻");
+});
+
+// Case 5: Long mixed string
+benchmark("Long mixed string", function() use ($long_string) {
+    grapheme_str_split($long_string);
+});
+
+?>

benchmark_range.php

@@ -0,0 +1,44 @@
+<?php
+
+function benchmark($name, $func) {
+    $start = microtime(true);
+    $func();
+    $end = microtime(true);
+    printf("%-40s: %8.4f seconds\n", $name, $end - $start);
+}
+
+const LARGE_INT = 10000000;
+const LARGE_FLOAT = 10000000.0;
+
+// --- Integer Benchmarks ---
+benchmark("Integer Range (1 to 10,000,000)", function() {
+    $a = range(1, LARGE_INT);
+});
+
+benchmark("Integer Range (10,000,000 to 1)", function() {
+    $a = range(LARGE_INT, 1);
+});
+
+benchmark("Integer Range with Step (1 to 10,000,000, step 2)", function() {
+    $a = range(1, LARGE_INT, 2);
+});
+
+
+// --- Float Benchmarks ---
+benchmark("Float Range (1.0 to 10,000,000.0)", function() {
+    $a = range(1.0, LARGE_FLOAT);
+});
+
+benchmark("Float Range with Step (10,000,000.0 to 1.0, step 2.5)", function() {
+    $a = range(LARGE_FLOAT, 1.0, 2.5);
+});
+
+
+// --- Character Benchmark ---
+benchmark("Character Range ('a' to 'z')", function() {
+    $a = range('a', 'z');
+});
+
+echo "\n";
+
+?>

benchmark_str_decrement.php

@@ -0,0 +1,51 @@
+<?php
+
+function benchmark($name, $func) {
+    $start = microtime(true);
+    // Using a smaller iteration count because string operations can be slow
+    for ($i = 0; $i < 20000; $i++) {
+        $func();
+    }
+    $end = microtime(true);
+    echo "$name: " . ($end - $start) . "s\n";
+}
+
+echo "Benchmarking current str_decrement implementation...\n";
+
+// Case 1: Simple numeric string, no borrow needed.
+$s1 = "123456789123456789";
+benchmark("Simple numeric string", function() use ($s1) {
+    str_decrement($s1);
+});
+
+// Case 2: Long numeric string with full borrow. This is the key case for my optimization.
+$s2 = "1" . str_repeat("0", 50);
+benchmark("Long numeric string with full borrow", function() use ($s2) {
+    str_decrement($s2);
+});
+
+// Case 3: Simple alphanumeric string, no borrow needed.
+$s3 = "abcdefg9";
+benchmark("Simple alphanumeric string", function() use ($s3) {
+    str_decrement($s3);
+});
+
+// Case 4: Long alphanumeric string with numeric borrow.
+$s4 = "b" . str_repeat("0", 50);
+benchmark("Long alphanumeric with numeric borrow", function() use ($s4) {
+    str_decrement($s4);
+});
+
+// Case 5: Long alphanumeric string with letter borrow.
+$s5 = "z" . str_repeat("a", 50);
+benchmark("Long alphanumeric with letter borrow", function() use ($s5) {
+    str_decrement($s5);
+});
+
+// Case 6: A long string that does not trigger the leading-zero optimization.
+$s6 = "2" . str_repeat("0", 50);
+benchmark("Long numeric string without leading-zero result", function() use ($s6) {
+    str_decrement($s6);
+});
+
+?>

benchmark_str_ends_with.php

@@ -0,0 +1,49 @@
+<?php
+
+function benchmark($name, $func) {
+    $start = microtime(true);
+    // This function is very fast, so a high iteration count is needed.
+    for ($i = 0; $i < 500000; $i++) {
+        $func();
+    }
+    $end = microtime(true);
+    echo str_pad($name, 40) . ": " . number_format($end - $start, 6) . "s\n";
+}
+
+echo "Benchmarking current str_ends_with implementation...\n";
+
+$long_haystack = "This is a very long string that is used for benchmarking purposes to see how the function performs with a significant amount of data to process.";
+$long_needle_match = "a significant amount of data to process.";
+$long_needle_no_match = "a significant amount of data to process!";
+
+// Case 1: Haystack shorter than needle (fast path)
+benchmark("Haystack shorter than needle", function() {
+    str_ends_with("short", "this is much longer");
+});
+
+// Case 2: Matching short needle
+benchmark("Matching short needle", function() use ($long_haystack) {
+    str_ends_with($long_haystack, "process.");
+});
+
+// Case 3: Non-matching short needle
+benchmark("Non-matching short needle", function() use ($long_haystack) {
+    str_ends_with($long_haystack, "process!");
+});
+
+// Case 4: Matching long needle
+benchmark("Matching long needle", function() use ($long_haystack, $long_needle_match) {
+    str_ends_with($long_haystack, $long_needle_match);
+});
+
+// Case 5: Non-matching long needle
+benchmark("Non-matching long needle", function() use ($long_haystack, $long_needle_no_match) {
+    str_ends_with($long_haystack, $long_needle_no_match);
+});
+
+// Case 6: Matching empty needle (edge case)
+benchmark("Matching empty needle", function() use ($long_haystack) {
+    str_ends_with($long_haystack, "");
+});
+
+?>

benchmark_str_increment.php

@@ -0,0 +1,50 @@
+<?php
+
+function benchmark($name, $func) {
+    $start = microtime(true);
+    for ($i = 0; $i < 20000; $i++) {
+        $func();
+    }
+    $end = microtime(true);
+    echo "$name: " . ($end - $start) . "s\n";
+}
+
+echo "Benchmarking current str_increment implementation...\n";
+
+// Case 1: Simple numeric string, no carry needed.
+$s1 = "123456788";
+benchmark("Simple numeric string", function() use ($s1) {
+    str_increment($s1);
+});
+
+// Case 2: Long numeric string with full carry. This is the key case for my optimization.
+$s2 = str_repeat("9", 50);
+benchmark("Long numeric string with full carry", function() use ($s2) {
+    str_increment($s2);
+});
+
+// Case 3: Simple alphanumeric string, no carry needed.
+$s3 = "abcde8";
+benchmark("Simple alphanumeric string", function() use ($s3) {
+    str_increment($s3);
+});
+
+// Case 4: Long alphanumeric string with numeric carry.
+$s4 = "a" . str_repeat("9", 50);
+benchmark("Long alphanumeric with numeric carry", function() use ($s4) {
+    str_increment($s4);
+});
+
+// Case 5: Long alphanumeric string with full letter and numeric carry.
+$s5 = "z" . str_repeat("9", 50);
+benchmark("Long alphanumeric with full carry", function() use ($s5) {
+    str_increment($s5);
+});
+
+// Case 6: A long string that does not trigger reallocation.
+$s6 = "8" . str_repeat("9", 50);
+benchmark("Long numeric string without reallocation", function() use ($s6) {
+    str_increment($s6);
+});
+
+?>

benchmark_str_ireplace.php

@@ -0,0 +1,47 @@
+<?php
+
+function benchmark($name, $func) {
+    $start = microtime(true);
+    // A lower iteration count because str_ireplace can be very slow with arrays on long strings.
+    for ($i = 0; $i < 500; $i++) {
+        $func();
+    }
+    $end = microtime(true);
+    echo str_pad($name, 50) . ": " . number_format($end - $start, 6) . "s\n";
+}
+
+echo "Benchmarking current str_ireplace implementation...\n";
+
+// A long string (~22KB) with a variety of characters to search for.
+$long_haystack = str_repeat("The quick brown fox jumps over the lazy dog. ", 500);
+
+// An array of search terms that will all be found.
+$search_array = ['the', 'quick', 'brown', 'fox', 'jumps', 'over', 'lazy', 'dog'];
+$replace_array = ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H'];
+
+// An array of search terms that will not be found.
+$search_array_no_match = ['xylophone', 'yak', 'zebra', 'walrus', 'vulture', 'unicorn', 'tiger', 'snake'];
+
+// ===== TEST CASES =====
+
+// Case 1: Simple string search (no arrays) for baseline comparison.
+benchmark("Simple string search", function() use ($long_haystack) {
+    str_ireplace('fox', 'cat', $long_haystack);
+});
+
+// Case 2: Array search, single replace. THIS IS THE PRIMARY OPTIMIZATION TARGET.
+benchmark("Array search, single replacement", function() use ($long_haystack, $search_array) {
+    str_ireplace($search_array, 'REPLACED', $long_haystack);
+});
+
+// Case 3: Array search, array replace. THIS IS THE PRIMARY OPTIMIZATION TARGET.
+benchmark("Array search, array replacement", function() use ($long_haystack, $search_array, $replace_array) {
+    str_ireplace($search_array, $replace_array, $long_haystack);
+});
+
+// Case 4: Array search with no matches. To ensure no performance regression.
+benchmark("Array search, no matches", function() use ($long_haystack, $search_array_no_match) {
+    str_ireplace($search_array_no_match, 'REPLACED', $long_haystack);
+});
+
+?>

benchmark_str_pad.php

@@ -0,0 +1,27 @@
+<?php
+
+function benchmark_str_pad(string $input, int $pad_length, string $pad_string, int $pad_type) {
+    $start = microtime(true);
+    for ($i = 0; $i < 100000; $i++) {
+        str_pad($input, $pad_length, $pad_string, $pad_type);
+    }
+    $end = microtime(true);
+    return $end - $start;
+}
+
+$short_string = 'Hello';
+$long_string = str_repeat('Hello', 100);
+
+echo "Benchmarking current str_pad implementation...\n";
+
+echo "Short string, right pad: " . benchmark_str_pad($short_string, 100, ' ', STR_PAD_RIGHT) . "s\n";
+echo "Short string, left pad: " . benchmark_str_pad($short_string, 100, ' ', STR_PAD_LEFT) . "s\n";
+echo "Short string, both pad: " . benchmark_str_pad($short_string, 100, ' ', STR_PAD_BOTH) . "s\n";
+
+echo "Long string, right pad: " . benchmark_str_pad($long_string, 1000, ' ', STR_PAD_RIGHT) . "s\n";
+echo "Long string, left pad: " . benchmark_str_pad($long_string, 1000, ' ', STR_PAD_LEFT) . "s\n";
+echo "Long string, both pad: " . benchmark_str_pad($long_string, 1000, ' ', STR_PAD_BOTH) . "s\n";
+
+echo "Short string, multi-char pad, right: " . benchmark_str_pad($short_string, 100, '-=', STR_PAD_RIGHT) . "s\n";
+echo "Short string, multi-char pad, left: " . benchmark_str_pad($short_string, 100, '-=', STR_PAD_LEFT) . "s\n";
+echo "Short string, multi-char pad, both: " . benchmark_str_pad($short_string, 100, '-=', STR_PAD_BOTH) . "s\n";