串行抓取 10 個頁面,每個 500ms,總耗時 5 秒。並行抓取僅需最慢的那一個 (約 500ms) + 少量開銷。**PHP 單執行緒也能並行**,關鍵在於 非阻塞 I/O + I/O 多路復用 (stream_select)。
stream_socket_client) 並發送 HTTP 請求。stream_select() 監聽讀取就緒事件。fread() 讀取資料,累積到對應緩衝區。<?php
/**
* 並行 HTTP GET
* @param string[] $urls 目標 URL 陣列
* @param float $timeout 總超時秒數
* @return array url => responseBody (含 header+body) 或 false
*/
function httpMultiGet(array $urls, float $timeout = 30.0): array {
$results = [];
$sockets = [];
// 1. 建立所有非阻塞連線並發送請求
foreach ($urls as $key => $url) {
$parts = parse_url($url);
$host = $parts['host'] ?? '';
$path = ($parts['path'] ?? '/') . ($parts['query'] ? '?' . $parts['query'] : '');
$port = $parts['port'] ?? 80;
$flags = STREAM_CLIENT_CONNECT | STREAM_CLIENT_ASYNC_CONNECT;
$socket = @stream_socket_client("tcp://$host:$port", $errno, $errstr, 5, $flags);
if (!$socket) {
$results[$key] = false;
continue;
}
// 設為非阻塞
stream_set_blocking($socket, false);
// 發送 HTTP/1.0 請求 (簡化,無需處理 Chunked)
fwrite($socket, "GET $path HTTP/1.0\r\nHost: $host\r\nConnection: close\r\n\r\n");
$sockets[$key] = $socket;
$results[$key] = '';
}
$startTime = microtime(true);
$read = $write = $except = null;
// 2. 事件循環
while (!empty($sockets)) {
// 檢查總超時
if (microtime(true) - $startTime > $timeout) {
foreach ($sockets as $k => $s) { fclose($s); $results[$k] = false; }
break;
}
$read = $sockets;
$n = @stream_select($read, $write, $except, 0, 200000); // 0.2 秒
if ($n === false) { /* select error */ break; }
if ($n === 0) continue; // timeout, continue loop
// 3. 處理可讀 sockets
foreach ($read as $socket) {
$key = array_search($socket, $sockets, true);
if ($key === false) continue;
$chunk = @fread($socket, 8192);
if ($chunk === '' || $chunk === false) { // EOF 或錯誤
fclose($socket);
unset($sockets[$key]);
} else {
$results[$key] .= $chunk;
}
}
}
// 4. 分離 Header/Body
foreach ($results as $key => &$resp) {
if ($resp !== false) {
[$header, $body] = explode("\r\n\r\n", $resp, 2) + ['', ''];
$resp = $body;
}
}
return $results;
}
// ===== 使用範例 =====
$urls = [
'https://www.qq.com',
'https://www.sina.com.cn',
'https://www.sohu.com',
'https://www.163.com',
];
echo "並行抓取開始...\n";
$t1 = microtime(true);
$results = httpMultiGet($urls, 10.0);
echo "並行耗時: " . (microtime(true) - $t1) . " 秒\n";
echo "\n串行抓取對比...\n";
$t2 = microtime(true);
foreach ($urls as $u) @file_get_contents($u);
echo "串行耗時: " . (microtime(true) - $t2) . " 秒\n";
print_r(array_map(fn($v) => $v ? strlen($v) : 'failed', $results));
並行耗時: 1.23 秒 串行耗時: 4.87 秒 加速比: ~4x
stream_socket_client("ssl://$host:443", ...) 需處理 TLS 握手非阻塞 (較複雜,建議用 cURL multi)。curl_multi_* 族函數是生產級並行 HTTP 客戶端的標準實作,支援 HTTPS、重導向、Cookie、壓縮等完整特性。生產環境強烈建議直接使用 Guzzle + curl_multi 或 Swoole 協程。