php 论坛采集程序 模拟登陆，抓取页面 实现代码

<?php
// 吴燕军
// 2009-06-27
// 采集程序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);
$contents = preg_replace('/images\//','http://bbs.war3.cn/images/',$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://bbs.war3.cn/logging.php?action=login';
$postfields='loginfield=username&username=1nject10n& password=xxxxxx&questionid=0&cookietime=315360000& referer=http://bbs.war3.cn/&loginsubmit=提交';
request($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://bbs.war3.cn/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){
$url = "http://bbs.war3.cn/viewthread.php?tid=$list";
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*/
?>