curl抓取
<?php
header("content-type:text/html;charset=utf8");
set_time_limit(0);
//=================================工具函数=====================
function real_url($current_url, $base_url='') {
$data = parse_url($current_url);
if ( ! isset($data['host'])) {
$current_url = $base_url . $base_url;
}
return $current_url;
}
// 源文件下载地址 : http://curlmulti.com/index/download/CurlMulti
if ( ! is_file('CurlMulti.php')) {
$phpQuery = file_get_contents('http://curlmulti.com/index/download/CurlMulti');
file_put_contents("./CurlMulti.php", $phpQuery);
}
require 'CurlMulti.php';
// 源文件下载地址 ： http://curlmulti.com/index/download/phpQuery
if ( ! is_file('phpQuery.php')) {
$phpQuery = file_get_contents('http://curlmulti.com/index/download/phpQuery');
file_put_contents("./phpQuery.php", $phpQuery);
}
require 'phpQuery.php';
class myDebug {
static $start;
static $end;
static $times;
static function microtime_float(){
list ($usec, $sec) = explode(" ", microtime());
return ((float) $usec + (float) $sec);
}
public function set_start() {
self::$start = self::microtime_float();
}
public function set_end() {
self::$end = self::microtime_float();
}
public function report() {
return self::$end - self::$start;
}
}
class myCurl {
public $curl;
public $article_list;
protected $cacheDir;
protected $pageCount;
protected $articleCount;
protected $request;
public function __construct(request $request){
$this->_init_request($request);
$this->_init_curl();
}
protected function _init_var() {
$this->pageCount = 0;
}
protected function _init_request(request $request) {
$this->request = $request;
// $this->request->cache_path = __DIR__ . '/sjm_cache/';
// $this->request->fetch_item_query = '#J_posts_list .subject .title a';
// $this->request->fetch_page_current = '.J_page_wrap .pages strong';
// $this->request->base_url = 'http://bbs.sijiaomao.com/index.php?m=bbs&c=thread&fid=10&page=%d';
}
protected function _init_curl() {
$this->curl = new CurlMulti();
$this->cacheDir = $this->request->cache_path . 'cache';
if (! is_dir($this->cacheDir)) {
mkdir($this->cacheDir, 777, true);
}
$this->cacheDataDir = $this->request->cache_path . 'data';
if (! is_dir($this->cacheDataDir)) {
mkdir($this->cacheDataDir, 777, true);
}
$this->curl->cache = array(
'dir' => $this->cacheDir,
'on' => true,
'expire' => 3600 * 24
);
$this->curl->maxThread = 10;
$this->curl->opt[CURLOPT_CONNECTTIMEOUT] = 10;
}
public function fetch_list(){
$this->_add_fetch_list_url();
$this->curl->start();
$this->_save_article_list();
}
public function fetch_article() {
foreach ($this->article_list as $k => $v) {
$this->curl->add(array(
'url' => $v['href']
), array($this, '_success_article'));
}
$this->curl->start();
}
public function display() {
printf(
"\n共抓取%d个页面\n文章列表%d篇\n相关文章%d篇\n文章目录存放在%s\n",
$this->pageCount + $this->articleCount,
$this->pageCount,
count($this->article_list),
$this->cacheDataDir . '/list.php'
);
}
public function fetch() {
return sprintf(
"\n共抓取%d个页面\n文章列表%d篇\n相关文章%d篇\n文章目录存放在%s\n",
$this->pageCount + $this->articleCount,
$this->pageCount,
count($this->article_list),
$this->cacheDataDir . '/list.php'
);
}
public function _add_fetch_list_url($page = 1){
$this->curl->add(
array(
'url' => sprintf($this->request->base_url, $page),
'args' => array('page' => $page)
),
array($this, '_success_list')
);
}
protected function _save_article_list() {
$res = file_put_contents(
$this->cacheDataDir . '/list.php',
sprintf("<?php\n return\t%s;",
var_export($this->article_list, true))
);
// 相关性排序整理
/*uasort($this->article_list, function ($a, $b){
preg_match_all('#([a-zA-Z]+)#is', $a['title'], $match);
$a_title = strtoupper(implode("", $match[0]));
preg_match_all('#([a-zA-Z]+)#is', $b['title'], $match);
$b_title = strtoupper(implode("", $match[0]));
return $a_title > $b_title;
});*/
$res = file_put_contents(
$this->cacheDataDir . '/list.txt',
array_map(function($a_list){
$str = sprintf(
"标题：%s\t超链接:%s \n",
str_replace(" ", "", $a_list['title']),
$a_list['href']
);
return $str;
}, $this->article_list)
);
return $res;
}
public function _success_article($r, $param){
++$this->articleCount;
}
public function _success_list($r, $param){
++$this->pageCount;
$html = phpQuery::newDocumentHTML($r['content']);
$list = $html[$this->request->fetch_item_query];
foreach ($list as $v) {
$v = pq($v);
$item = array(
"title" => $v->attr('title') ? $v->attr('title') : $v->text(),
"href" => real_url($v->attr('href'), $this->request->base_url)
);
$this->article_list[md5($item['href'])] = $item;
}
$page_current = $html[$this->request->fetch_page_current];
if ($page_current->next()->text()) {
$page = ++ $param['page'];
$this->_add_fetch_list_url($page);
}
phpQuery::unloadDocuments();
}
}
class request{
/*url*/
public $base_url;
/*缓存文件路径*/
public $cache_path;
/*获取元素的CSS选择器*/
public $fetch_item_query;
/*分页当前页面元素的CSS选择器*/
public $fetch_page_current;
static $instance;
static public function getInstance() {
if (empty(self::$instance)) {
self::$instance = new self;
}
return self::$instance;
}
private function __construct() {
$this->_init_base();
}
function _init_base() {
$this->cache_path = __DIR__ . '/'. trim($_POST['cache_path'], '/') .'/';
$this->fetch_item_query = $_POST['fetch_item_query'];
$this->fetch_page_current = $_POST['fetch_page_current'];
$this->base_url = $_POST['url'];
}
function request() {
if (strstr($_POST['url'], '?')) {
$url = sprintf("%s&auth=%s", $_POST['url'], $auth);
} else {
$url = sprintf("%s?auth=%s", $_POST['url'], $auth);
}
$param = array();
if (isset($_POST['param'])) {
foreach($_POST['param'] as $k => $item) {
if (!empty($item['method']) && !empty($item['name'])) {
$param[$item['method']][$item['name']] = $item['value'];
}
}
}
if (isset($param['get']) && !empty($param['get'])) {
foreach ($param['get'] as $name => $value) {
$url = sprintf("%s&%s=%s", $url, $name, $value);
}
}
$post_data = null;
if (isset($param['post']) && !empty($param['post'])) {
$post_data = $param['post'];
}
}
}
?>
<?php
if (isset($_POST['submit'])) {
$request = request::getInstance();
$myCurl = new myCurl($request);
myDebug::set_start();
$myCurl->fetch_list();
//$myCurl->fetch_article();
myDebug::set_end();
} else {
$_POST['url'] = 'http://www.oschina.net/code/tag/php?show=time&lang=&catalog=&p=%d';
$_POST['cache_path'] = 'oschina';
$_POST['fetch_item_query'] = '.code_list ul li .code_title > a';
$_POST['fetch_page_current'] = '.pager li.current';
}
?>
<html lang="zh-CN">
<head>
<meta charset="utf-8">
<title>页面爬虫</title>
<link href="http://cdn.bootcss.com/bootstrap/3.2.0/css/bootstrap.min.css" rel="stylesheet">
<link href="http://cdn.bootcss.com/font-awesome/4.1.0/css/font-awesome.min.css" rel="stylesheet">
<link href="http://static.bootcss.com/www/assets/css/site.min.css?v3" rel="stylesheet">
<link href="http://static.bootcss.com/www/assets/ico/favicon.png" rel="shortcut icon">
<script src="http://cdn.bootcss.com/jquery/1.11.1/jquery.min.js"></script>
</head>
<body>
<div class="container">
<div class="row row-offcanvas row-offcanvas-right">
<div class="col-xs-12 col-sm-12">
<div class="row" >
<div class="col-xs-1 col-lg-4">
<h1>页面爬虫</h1>
<div class="thumbnail">
<form class="form-signin" action="" method="post">
<b>请填URL</b>:
<input value="<?php echo isset($_POST['url'])?$_POST['url']:'';?>" class="form-control" placeholder="填写完整地址，以http://开头" type="text" name="url" required><br>
<b>请填缓存文件路径</b>:
<input value="<?php echo isset($_POST['url'])?$_POST['cache_path']:'';?>" class="form-control" placeholder="填写缓存文件路径" type="text" name="cache_path" required><br>
<b>请填获取元素的CSS选择器</b>:
<input value="<?php echo isset($_POST['url'])?$_POST['fetch_item_query']:'';?>" class="form-control" placeholder="填写获取元素的CSS选择器" type="text" name="fetch_item_query" required><br>
<b>请填分页当前页面元素的CSS选择器</b>:
<input value="<?php echo isset($_POST['url'])?$_POST['fetch_page_current']:'';?>" class="form-control" placeholder="填写分页当前页面元素的CSS选择器" type="text" name="fetch_page_current" required><br>
<?php if (isset($_POST['param']) && !empty($_POST['param'])) :?>
<?php foreach ($_POST['param'] as $k => $item) :?>
<?php if (!empty($item['method']) && !empty($item['name'])) :?>
<div class="thumbnail">
<b>参数name</b>:
<input value="<?php echo $item['name'];?>" placeholder="请填写" type="text" name="param[<?php echo $k;?>][name]"><br>
<b>参数value</b>:
<input value="<?php echo $item['value'];?>" placeholder="请填写" type="text" name="param[<?php echo $k;?>][value]"><br>
<b>请求方式</b>:
<label><input <?php if($item['method']=='get'):?>checked<?php endif;?> value="get" type="radio" name="param[<?php echo $k;?>][method]">get</label>
<label><input <?php if($item['method']=='post'):?>checked<?php endif;?> value="post" type="radio" name="param[<?php echo $k;?>][method]">post</label><br />
<a href="#" onclick="del_param(this)">删除</a>
</div>
<?php endif;?>
<?php endforeach;?>
<?php endif;?>
<input type="button" name="add_param" id="add_param" value="添加参数" class="btn btn-lg btn-primary btn-block"><br />
<input type="submit" name="submit" value="下载" class="btn btn-lg btn-primary btn-block"><br />
</form>
</div>
</div>
<div class="col-xs-1 col-lg-8">
<?php
if (isset($_POST['submit'])) {
echo "<pre>";
echo "请求时间:";
var_dump(myDebug::report());
echo "<br />请求url:";
isset($request->base_url) && var_dump($request->base_url);
echo "<br />请求参数:";
isset($param) && var_dump($param);
echo "<hr />结果：";
var_dump($myCurl->fetch());
echo "</pre>";
}
?>
</div>
</div>
</div>
</div>
<hr />
</div>
<div class="blog-masthead">
<div class="container">
<nav class="blog-nav">
<p class="blog-nav-item">© Company 2014</p>
</nav>
</div>
</div>
</body>
</html>
<script>
$("#add_param").click(function(){
var input_len = $("form input").size();
input_len++;
$(this).before('\
<div class="thumbnail">\
<b>参数name</b>:\
<input value="" placeholder="请填写" type="text" name="param['+ input_len +'][name]"><br>\
<b>参数value</b>:\
<input value="" placeholder="请填写" type="text" name="param['+ input_len +'][value]"><br>\
<b>请求方式</b>:\
<label><input checked value="get" type="radio" name="param['+ input_len +'][method]">get</label>\
<label><input value="post" type="radio" name="param['+ input_len +'][method]">post</label><br />\
<a href="#" onclick="del_param(this)">删除</a>\
</div>\
');
});
function del_param(obj) {
$(obj).parent().remove();
}
</script>
