2017-09-19 02:33:38 +08:00
|
|
|
<?php
|
|
|
|
|
namespace QL;
|
|
|
|
|
use phpQuery;
|
2017-09-19 17:48:48 +08:00
|
|
|
use QL\Dom\Dom;
|
2017-09-19 02:33:38 +08:00
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* QueryList
|
|
|
|
|
*
|
|
|
|
|
* 一个基于phpQuery的通用列表采集类
|
|
|
|
|
*
|
|
|
|
|
* @author Jaeger
|
|
|
|
|
* @email JaegerCode@gmail.com
|
|
|
|
|
* @link https://github.com/jae-jae/QueryList
|
|
|
|
|
* @version 4.0.0
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
class QueryList
|
|
|
|
|
{
|
2017-09-19 19:06:16 +08:00
|
|
|
protected $html;
|
|
|
|
|
protected $document;
|
|
|
|
|
protected $rules;
|
|
|
|
|
protected $range = null;
|
2017-09-21 01:44:03 +08:00
|
|
|
protected $encoder;
|
2017-09-19 02:33:38 +08:00
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* QueryList constructor.
|
|
|
|
|
*/
|
|
|
|
|
public function __construct()
|
|
|
|
|
{
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* @return mixed
|
|
|
|
|
*/
|
|
|
|
|
public function getHtml()
|
|
|
|
|
{
|
|
|
|
|
return $this->html;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2017-09-19 17:48:48 +08:00
|
|
|
* @param $html
|
|
|
|
|
* @return $this
|
2017-09-19 02:33:38 +08:00
|
|
|
*/
|
|
|
|
|
public function setHtml($html)
|
|
|
|
|
{
|
|
|
|
|
$this->html = $html;
|
|
|
|
|
$this->document = phpQuery::newDocumentHTML($this->html);
|
|
|
|
|
return $this;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
public function find($selector)
|
|
|
|
|
{
|
2017-09-19 17:48:48 +08:00
|
|
|
return (new Dom($this->document))->find($selector);
|
2017-09-19 02:33:38 +08:00
|
|
|
}
|
|
|
|
|
|
2017-09-19 19:06:16 +08:00
|
|
|
public function rules(array $rules)
|
|
|
|
|
{
|
|
|
|
|
$this->rules = $rules;
|
|
|
|
|
return $this;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
public function range($range)
|
|
|
|
|
{
|
|
|
|
|
$this->range = $range;
|
|
|
|
|
return $this;
|
|
|
|
|
}
|
|
|
|
|
|
2017-09-20 01:12:54 +08:00
|
|
|
public function removeHead()
|
2017-09-19 19:06:16 +08:00
|
|
|
{
|
2017-09-20 01:12:54 +08:00
|
|
|
$html = preg_replace('/<head.+?>.+<\/head>/is','<head></head>',$this->html);
|
|
|
|
|
$this->setHtml($html);
|
2017-09-19 19:06:16 +08:00
|
|
|
return $this;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
public function query($callback = null)
|
|
|
|
|
{
|
2017-09-20 01:12:54 +08:00
|
|
|
$data = $this->getList();
|
2017-09-19 19:06:16 +08:00
|
|
|
return is_null($callback)?$data:$data->map($callback);
|
|
|
|
|
}
|
|
|
|
|
|
2017-09-20 01:12:54 +08:00
|
|
|
protected function getList()
|
2017-09-19 19:06:16 +08:00
|
|
|
{
|
|
|
|
|
$data = [];
|
|
|
|
|
$document = $this->document;
|
|
|
|
|
if (!empty($this->range)) {
|
|
|
|
|
$robj = pq($document)->find($this->range);
|
|
|
|
|
$i = 0;
|
|
|
|
|
foreach ($robj as $item) {
|
2017-09-21 01:44:03 +08:00
|
|
|
foreach ($this->rules as $key => $reg_value){
|
|
|
|
|
$tags = $reg_value[2] ?? '';
|
2017-09-19 19:06:16 +08:00
|
|
|
$iobj = pq($item)->find($reg_value[0]);
|
|
|
|
|
switch ($reg_value[1]) {
|
|
|
|
|
case 'text':
|
2017-09-20 01:12:54 +08:00
|
|
|
$data[$i][$key] = $this->allowTags(pq($iobj)->html(),$tags);
|
2017-09-19 19:06:16 +08:00
|
|
|
break;
|
|
|
|
|
case 'html':
|
2017-09-20 01:12:54 +08:00
|
|
|
$data[$i][$key] = $this->stripTags(pq($iobj)->html(),$tags);
|
2017-09-19 19:06:16 +08:00
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
$data[$i][$key] = pq($iobj)->attr($reg_value[1]);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if(isset($reg_value[3])){
|
|
|
|
|
$data[$i][$key] = call_user_func($reg_value[3],$data[$i][$key],$key);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
$i++;
|
|
|
|
|
}
|
|
|
|
|
} else {
|
2017-09-21 01:44:03 +08:00
|
|
|
foreach ($this->rules as $key => $reg_value){
|
|
|
|
|
$tags = $reg_value[2] ?? '';
|
2017-09-19 19:06:16 +08:00
|
|
|
$lobj = pq($document)->find($reg_value[0]);
|
|
|
|
|
$i = 0;
|
|
|
|
|
foreach ($lobj as $item) {
|
|
|
|
|
switch ($reg_value[1]) {
|
|
|
|
|
case 'text':
|
2017-09-20 01:12:54 +08:00
|
|
|
$data[$i][$key] = $this->allowTags(pq($item)->html(),$tags);
|
2017-09-19 19:06:16 +08:00
|
|
|
break;
|
|
|
|
|
case 'html':
|
2017-09-20 01:12:54 +08:00
|
|
|
$data[$i][$key] = $this->stripTags(pq($item)->html(),$tags);
|
2017-09-19 19:06:16 +08:00
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
$data[$i][$key] = pq($item)->attr($reg_value[1]);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if(isset($reg_value[3])){
|
|
|
|
|
$data[$i][$key] = call_user_func($reg_value[3],$data[$i][$key],$key);
|
|
|
|
|
}
|
2017-09-19 02:33:38 +08:00
|
|
|
|
2017-09-19 19:06:16 +08:00
|
|
|
$i++;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
phpQuery::$documents = array();
|
|
|
|
|
return collect($data);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* 去除特定的html标签
|
|
|
|
|
* @param string $html
|
|
|
|
|
* @param string $tags_str 多个标签名之间用空格隔开
|
|
|
|
|
* @return string
|
|
|
|
|
*/
|
2017-09-20 01:12:54 +08:00
|
|
|
protected function stripTags($html,$tags_str)
|
2017-09-19 19:06:16 +08:00
|
|
|
{
|
2017-09-20 01:12:54 +08:00
|
|
|
$tagsArr = $this->tag($tags_str);
|
|
|
|
|
$html = $this->removeTags($html,$tagsArr[1]);
|
2017-09-19 19:06:16 +08:00
|
|
|
$p = array();
|
|
|
|
|
foreach ($tagsArr[0] as $tag) {
|
|
|
|
|
$p[]="/(<(?:\/".$tag."|".$tag.")[^>]*>)/i";
|
|
|
|
|
}
|
|
|
|
|
$html = preg_replace($p,"",trim($html));
|
|
|
|
|
return $html;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* 保留特定的html标签
|
|
|
|
|
* @param string $html
|
|
|
|
|
* @param string $tags_str 多个标签名之间用空格隔开
|
|
|
|
|
* @return string
|
|
|
|
|
*/
|
2017-09-20 01:12:54 +08:00
|
|
|
protected function allowTags($html,$tags_str)
|
2017-09-19 19:06:16 +08:00
|
|
|
{
|
2017-09-20 01:12:54 +08:00
|
|
|
$tagsArr = $this->tag($tags_str);
|
|
|
|
|
$html = $this->removeTags($html,$tagsArr[1]);
|
2017-09-19 19:06:16 +08:00
|
|
|
$allow = '';
|
|
|
|
|
foreach ($tagsArr[0] as $tag) {
|
|
|
|
|
$allow .= "<$tag> ";
|
|
|
|
|
}
|
|
|
|
|
return strip_tags(trim($html),$allow);
|
|
|
|
|
}
|
|
|
|
|
|
2017-09-20 01:12:54 +08:00
|
|
|
protected function tag($tags_str)
|
2017-09-19 19:06:16 +08:00
|
|
|
{
|
|
|
|
|
$tagArr = preg_split("/\s+/",$tags_str,-1,PREG_SPLIT_NO_EMPTY);
|
|
|
|
|
$tags = array(array(),array());
|
|
|
|
|
foreach($tagArr as $tag)
|
|
|
|
|
{
|
|
|
|
|
if(preg_match('/-(.+)/', $tag,$arr))
|
|
|
|
|
{
|
|
|
|
|
array_push($tags[1], $arr[1]);
|
|
|
|
|
}else{
|
|
|
|
|
array_push($tags[0], $tag);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return $tags;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* 移除特定的html标签
|
|
|
|
|
* @param string $html
|
|
|
|
|
* @param array $tags 标签数组
|
|
|
|
|
* @return string
|
|
|
|
|
*/
|
2017-09-20 01:12:54 +08:00
|
|
|
protected function removeTags($html,$tags)
|
2017-09-19 19:06:16 +08:00
|
|
|
{
|
|
|
|
|
$tag_str = '';
|
|
|
|
|
if(count($tags))
|
|
|
|
|
{
|
|
|
|
|
foreach ($tags as $tag) {
|
|
|
|
|
$tag_str .= $tag_str?','.$tag:$tag;
|
|
|
|
|
}
|
|
|
|
|
phpQuery::$defaultCharset = $this->inputEncoding?$this->inputEncoding:$this->htmlEncoding;
|
|
|
|
|
$doc = phpQuery::newDocumentHTML($html);
|
|
|
|
|
pq($doc)->find($tag_str)->remove();
|
|
|
|
|
$html = pq($doc)->htmlOuter();
|
|
|
|
|
$doc->unloadDocument();
|
|
|
|
|
}
|
|
|
|
|
return $html;
|
|
|
|
|
}
|
2017-09-19 02:33:38 +08:00
|
|
|
}
|
