微信公众号文章内容采集

<?php
// 设置脚本执行不超时
set_time_limit ( 0 );
class Gather {

private $url;
private $path;
public function __construct($url, $path) {

$this->url = $url;
$this->path = $path;
}
public function fetch() {
return $this->transform ( $this->url, $this->path );
}
private function createPic($url, $path, $name) {

$img = file_get_contents ( $url );
$info = getimagesize ( $url );
$type = str_replace ( 'image/', '', $info ['mime'] );
$fileName = $path . DIRECTORY_SEPARATOR . $name . ".$type";
file_put_contents ( $fileName, $img );
return $fileName;

}
private function transform($url, $path) {
if (! file_exists ( $path ))
mkdir ( $path );
$content = file_get_contents ( $url );
preg_match ( '/<title>(.*)<\/title>/i', $content, $result );
$data ['title'] = $result [1]; // 文章标题
preg_match ( '/var\s+msg_cdn_url\s*=\s*"([^\s]*)"/', $content, $result );
preg_match ( '/var\s+msg_desc\s*=\s*"([^\s]*)"/', $content, $result );
$data ['description'] = $result [1]; // 公众号文章摘要

// 获取微信主体内容
preg_match ( '/<div\s+class="rich_media_content\s*"\s+id="js_content">(.*?)<\/div>/is', $content, $result );

// 获取微信主体中的防盗链图片
preg_match_all ( '/data-src="[a-zA-z]+:\/\/[^\s]*[mmbiz|mmbiz_jpg|mmbiz_gif]\/[^\s]*\/\d*\?([^\s]*=[^\s]*)*"|data-src="[a-zA-z]+:\/\/[^\s]*[mmbiz|mmbiz_jpg|mmbiz_gif]\/[^\s]*\/\d+"|background-image\s*:\s*url\s*\(\s*[a-zA-z]+:\/\/[^\s]*mmbiz\/[^\s]*\/\d+|background-image\s*:\s*url\s*\(\s*[a-zA-z]+:\/\/[^\s]*mmbiz\/[^\s]*\/\d+\?[^\s]*=[^\s]*/is', $result [1], $result2 );
// 判断微信主体中是否包含防盗链图片
if (! empty ( $result2 [0] )) {

foreach ( $result2 [0] as $value ) {

// 取出防盗链地址中的data-src值后的http://url主体
//preg_match ( '/[a-zA-z]+:\/\/[^\s]*\/[mmbiz|mmbiz_jpg]\/([^\s\/]*)\/\d+\?[^\s"]*|[a-zA-z]+:\/\/[^\s]*[mmbiz|mmbiz_jpg]\/([^\s\/]*)\/\d+/', $value, $temp );
preg_match ( '/[a-zA-z]+:\/\/[^\s]*[mmbiz|mmbiz_jpg|mmbiz_gif]\/([^\s\/]*)\/\d*\?([^\s]*=[^\s]*)*[^"]|[a-zA-z]+:\/\/[^\s]*[mmbiz|mmbiz_jpg|mmbiz_gif]\/([^\s\/]*)\/\d+/', $value, $temp );
$temp = array_filter ($temp);
$temp = array_values($temp);
$urlList [] = $temp [0];
$nameList [] = $temp [1];

}
$path = realpath($path);
foreach ( $urlList as $value ) {

$name = array_shift ( $nameList );

$fileName = $this->createPic ( $value, $path, $name );    // 把图片保存到本地
$result [1] = str_replace ( $value, $fileName, $result [1] );

}

}
// 更新所有data-src的地址
$result [1] = str_replace ( "data-src", "src", $result [1] );
// 返回处理后的微信主体内容。
$data ['content'] = trim($result [1]);
return $data;
}
}