使用php的curl采集diszuz论坛
写了个diszuz采集论坛的小程序。里面包括了模拟登录，获取页面源代码，正则匹配结果等部分，希望对大家有用。

<?php
set_time_limit(0);
//cookie保存目录
$cookie_jar = '/tmp/cookie.tmp'
/*函数------------------------------------------------------------------------------------------------------------*/
//模拟请求数据
function request($url,$postfields,$cookie_jar,$referer){
$ch = curl_init();
$options = array(CURLOPT_URL => $url,
;;;;;;CURLOPT_HEADER => 0,
;;;;;;CURLOPT_NOBODY => 0,
;;;;;;CURLOPT_PORT => 80,
;;;;;;CURLOPT_POST => 1,
;;;;;;CURLOPT_POSTFIELDS => $postfields,
;;;;;;CURLOPT_RETURNTRANSFER => 1,
;;;;;;CURLOPT_FOLLOWLOCATION => 1,
;;;;;;CURLOPT_COOKIEJAR => $cookie_jar,
;;;;;;CURLOPT_COOKIEFILE => $cookie_jar,
;;;;;;CURLOPT_REFERER => $referer
);
curl_setopt_array($ch, $options);
$code = curl_exec($ch);
curl_close($ch);
return $code
}
//获取帖子列表
function getThreadsList($code){
preg_match_all('/ <!--[.|\r|\n]*? <a href=\"viewthread.php\?tid=(\d+)/',$code,$threads);
return $threads[1];
}
//判断该帖子是否存在
function isExits($code){
preg_match('/ <p>指定的主题不存在或已被删除或正在被审核，请返回。 <\/p>/',$code,$error);
return isset($error[0])?false:true
}
//获取帖子标题
function getTitle($code){
preg_match('/ <h1>[^ <\/h1>]*/',$code,$title_tmp);
$title = $title_tmp[0];
return $title
}
//获取帖子作者:
function getAuthor($code){
preg_match('/ <a href=\"space.php\?uid=\d+\" target=\"_blank\" id=\"userinfo\d+\" onmouseover=\"showMenu\(this\.id\)\">.+/',$code,$author_tmp);
$author = strip_tags($author_tmp[0]);
return $author
}
//获取楼主发表的内容
function getContents($code){
preg_match('/ <div id=\"postmessage_\d+\" class=\"t_msgfont\">(.|\r|\n)*? <\/div>/',$code,$contents_tmp);
$contents = preg_replace('/images\//','http://www.eb163.com/images/',$contents_tmp[0]);
return $contents
}
//打印帖子标题
function printTitle($title){
echo " <strong> <h2>帖子标题: </h2> </strong>",strip_tags($title)," <br/> <br/>"
}
//输出帖子作者
function printAuthor($author){
echo " <strong> <h2>帖子作者: </h2> </strong>",strip_tags($author)," <br/> <br/>"
}
//打印帖子内容
function printContents($contents){
echo " <strong> <h2>作者发表的内容: </h2>",$contents," </strong> <br/>"
}
//错误
function printError(){
echo " <i>该帖子不存在! </i>"
}
/*函数列表end---------------------------------------------------------------------------------------------------*/

/*登录论坛 begin*/
$url = 'http://www.eb163.com/logging.php?action=login'
$postfields='loginfield=username&username=1nject10n&password=xxxxxx&questionid=0&cookietime=315360000&referer=http://www.eb163.com/&loginsubmit=提交'
request($url,$postfields,$cookie_jar,'');
unset($postfields,$url);
/*登录论坛 end*/

/*获取帖子列表(位于第一页的帖子) begin*/
$url = 'http://www.eb163.com/forumdisplay.php?fid=57'
$code = request($url,'',$cookie_jar,'');
$threadsList = getThreadsList($code);
/*获取帖子列表 end*/
//帖子序列
$rows = 0
/*循环抓取所有帖子源代码 begin*/
foreach($threadsList as $list){
$url = "[url=http://www.eb163.com/viewthread.php?tid=$list]http://www.eb163.com/viewthread.php?tid=$list[/url]"
if(isExits($code)){
$code = request($url,'',$cookie_jar,'');
$color = $rows%2==0?'#00CCFF':'#FFFF33'
echo " <div style='background-color:$color'>"
echo " <h1>第",($rows+1),"贴: </h1> <br/>"
$author = getAuthor($code);
printAuthor($author);
$title = getTitle($code);
printTitle($title);
$contents = getContents($code);
printContents($contents);
echo " </div>"
$rows++;
}
else
printError();
echo "----------------------------------------------------------------------------------------- <br/> <br/>"
}
/*抓取源代码 end*/
?>

标签：