php实现的百度知道与问问的采集程序

<?php
session_start();
header("content-type:text/html;charset=gbk");
require("stole_config.php");
require("conn.php");
require("keyword.php");
$searchStr=$_GET["searchStr"];
$ss=explode(" ",$searchStr);//拆分搜索关键字
$word="";//关键字设为空
foreach($ss as $key=>$t)
{
if($key>0)
{
$word .="+";
}
$word .=urlencode($t);
}
$jl=intval($_GET['jl']);
if(isset($_GET['page']))
{
$page=intval($_GET['page']);
}else{
$page=1;
}
$rs=intval($_GET['rs']);
if($rs>=10)
{
$rs=0;
$page++;
}
if($page>76)
{
echo "采集完毕 ${jl}";
exit();
}
if(!empty($searchStr))//如果搜索
{
//获取问题页面
$content=@file_get_contents("http://zhidao.baidu.com/q?ct=17&lm=0&tn=ikaslist&pn=".(($page-1)*10)."&rn=10&word=".$word);
//获取问题列表
preg_match_all("/<a href="/question/(.*).html/iUs",$content,$uid);
$uid=$uid[1];//获取详细页文章
$uid=$uid[$rs];
//判断数据是否存在
$suid="bd{$uid}";
$sct=mysql_query("select count(*) from {$table_prefix}c_article where suid='$suid' ");
$sct=mysql_fetch_array($sct);
$sct=$sct[0];
if($sct==0)
{
$content=@file_get_contents("http://zhidao.baidu.com/question/".$uid.".html") ;
$arr=explode('<cq>',$content);
$art_title=$arr[1];
$arr=explode('</cq>',$art_title);
$art_title=$arr[0];//获取标题结束
//判断内容是否符合
$word_arr=explode(",",$cj_word);
$word_allow=false;//初始化是否允许采集
$word_count=count($word_arr);//关键字总数
for($i=0;$i<$word_count;$i++)
{
if(substr_count($art_title,$word_arr[$i])>0)
{
$word_allow=1;
$i=$word_count;
}
}
if($word_allow)//如果满足条件
{
$arr=explode('<cd><pre>',$content);
$contentQuestion=$arr[1];
$arr=explode('</pre></cd>',$contentQuestion);
$contentQuestion=$arr[0];
echo "开始采集内容<br>";
echo "$art_title<br>";
@preg_match_all('/(<ca>|<cn>)<pre>(.*)</pre>(</ca>|</cn>)/iUs',$content,$answerArr);
$answerArr=$answerArr[2];
if($arr_order==1)//随机排序
{
shuffle($answerArr);
}
if($arr_order==2)//倒序
{
$answerArr=krsort($answerArr);//倒序
}
foreach($answerArr as $t)
{
$answerTemp=str_replace('<ca><pre>','',$t);
$answerTemp=str_replace('</pre></ca>','',$answerTemp);
$answerTemp=str_replace('<cn><pre>','',$answerTemp);
$answerTemp=str_replace('</pre></cn>','',$answerTemp);
if(strlen($answerTemp)>$min_t1)
{
$art_content .=$answerTemp."<br>";
}
}
//去除链接
$s1="/(<a .*>)(.*)</a>/iUs";
$art_content=preg_replace($s1,${2},trim($art_content));
$art_content=str_replace("nr","<br>",$art_content);
if(strlen($art_content)>$min_t2)
{
$title_ct=mysql_query("select count(*) from {$table_prefix}c_article where art_title ='$art_title' ");//查看标题是否重复
$title_ct=@mysql_fetch_array($title_ct);
$title_ct=$title_ct[0];
if($title_ct>0)
{
$art_title .="{$same_title}{$title_ct}";
}
$art_time=date("Y-m-d");
$art_content=strtr($art_content,$keyword);
$sql="insert into {$table_prefix}c_article(art_title,art_content,art_time,art_author,suid) values('$art_title','$art_content','$art_time','$art_author','$suid')";//插入采集表
mysql_query($sql);
if(empty($t_catx_id))//如果无分类
{
$sql2="insert into {$t_table}({$t_art_title},{$t_art_content},{$t_art_time},{$t_artx_author}) values('$art_title','$art_content','$art_time','$art_author')";
}else
{
$sql2="insert into {$t_table}({$t_art_title},{$t_art_content},{$t_art_time},{$t_artx_author},{$t_catx_id}) values('$art_title','$art_content','$art_time','$art_author','$cat_id')";
}
mysql_query($sql2);//插入文章表
$jl++;
//数据库处理完毕
}else
{
echo "内容长度不够";
}
//获取文章内容结束
}else
{
echo "主题不符合要求";
}
}else
{
echo "已经存在";
}$rs++;
file_put_contents("bd.txt","采集{$searchStr}到第{$page}第{$rs}条");
echo "<script>location.href='baidu.php?searchStr=".urlencode($searchStr)."&page=".$page."&rs=".$rs."&jl=".$jl." ';</script>";
exit();
}
?>
<link href="style.css" rel="stylesheet" type="text/css" />
<table width="700" border="0" align="center" cellspacing="1" bgcolor="#CCCCCC">
<tr>
<td height="50" align="center" bgcolor="#00CC00"><h1><a href="http://www.jianlila.com">荐礼啦</a>知道问问采集插件</h1></td>
</tr>
</table>
<table width="700" border="0" align="center" cellspacing="1" bgcolor="#CCCCCC" style="margin-top:6px; margin-bottom:6px;">
<tr>
<td height="30" align="center" bgcolor="#FFFFFF"><a href="cj_config.php">采集设置</a> <a href="uninstall.php" onclick="return confirm('您确定要卸载采集插件吗');">卸载采集</a> <a href="cj_view.php">查看采集记录</a> <a href="cj_help.php">采集帮助</a> <a href="baidu.php" target="_blank">知道采集</a>  <a href="wenwen.php" target="_blank">问问采集</a></td>
</tr>
</table>
<table width="537" height="45" align="center" style="margin-top:30px;"><tr><td height="39">
<form id="form1" name="form1" method="get" action="baidu.php">
<div id="search">
<input name="searchStr" type="text" id="searchStr" value="<?php echo $searchStr; ?>" size="60" />
<input type="submit" name="searchBtn" id="searchBtn" value="知道偷偷" style="height:25px; line-height:25px;" />
</div>
</form>
</td></tr></table>

<?php
session_start();
header("content-type:text/html;charset=utf-8");
require("stole_config.php");
require("conn.php");
require("keyword.php");
if(!empty($_POST['ask']))
{
$ask=urlencode(trim($_POST['ask']));//获取表单提交的问题
$sp="S".$ask;
}else
{
$sp=urlencode($_GET['sp']);
}
if(empty($_GET['jl']))
{
$_GET['jl']=1;
}
$jl=$_GET['jl'];
$pg=intval($_GET['pg']);//获取页数
$rs=intval($_GET['rs']);//获得 记录的参数
if($rs>9)
{
$rs=0;
$pg++;
}
if($pg>51)
{
echo "采集完毕！ 总共采集 ".urldecode($sp)." ".$jl."条记录";
exit();
}
if($sp)//有设定答案才开始
{
$str=@file_get_contents("http://wenwen.soso.com/z/Search.e?sp={$sp}&pg={$pg}");
@preg_match("/<ol class="result_list">(.*)</ol>/iUs",$str,$asklist);//获取问答列表
//echo $asklist[1];
$url="/<a target="_blank" href="/z/(q.*.htm)/iUs";
@preg_match_all($url,$asklist[1],$urllist);//获取 所有的问题
$t=$urllist[1][$rs];
$uid=$t;
$suid="ww{$uid}";
$sct=mysql_query("select count(*) from {$table_prefix}c_article where suid='$suid' ");
$sct=mysql_fetch_array($sct);
$sct=$sct[0];
if($sct==0)
{
$html=@file_get_contents("http://wenwen.soso.com/z/${t}");
$html=str_replace("<pre>","",str_replace("</pre>","",$html));
$html=str_replace("<br/><br/><br/>","<br/><br/>",$html);
//echo $html;
@preg_match("/<div class="question_main">.*<h3>(.*)</h3>/iUs",$html,$ask_title);
$art_title=$ask_title[1];
@preg_match("/<div class="answer_con">(.*)</div>/iUs",$html,$answer);
$j=count($answer)-1;
$art_content="";//商品详细
for($i=$j;$i>=1;$i--)
{
if(strlen($answer[$i])>$min_t1)
{
$art_content .= $answer[$i];
}
}
$art_content=trim($art_content);
$s1="/(<a .*>)(.*)</a>/iUs";
$art_content=preg_replace($s1,${2},trim($art_content));
$word_arr=explode(",",iconv("gbk","utf-8",$cj_word));
$word_allow=false;//初始化是否允许采集
$word_count=count($word_arr);//总数
for($i=0;$i<$word_count;$i++)
{
if(substr_count($art_title,$word_arr[$i])>0)
{
$word_allow=1;
$i=$word_count;
}
}
if($word_allow)//如果合法
{ //开始处理数据库
if(strlen($art_content)>$min_t2)
{
echo "<font color=red>添加中............................</font><br>";
echo $art_title."<br>";
$art_title=iconv('utf-8','gbk', $art_title);
$title_ct=mysql_query("select count(*) from {$table_prefix}c_article where art_title ='$art_title' ");//查看标题是否重复
$title_ct=@mysql_fetch_array($title_ct);
$title_ct=$title_ct[0];
if($title_ct>0)
{
$art_title .="{$same_title}{$title_ct}";
}
$art_content=iconv('utf-8','gbk',str_replace("rn","<br>",$art_content));
$art_content=strtr($art_content,$keyword);
$art_time=date("Y-m-d");
$sql="insert into {$table_prefix}c_article(art_title,art_content,art_time,art_author,suid) values('$art_title','$art_content','$art_time','$art_author','$suid')";//插入采集表
mysql_query($sql);
if(empty($t_catx_id))//如果无分类
{
$sql2="insert into {$t_table}({$t_art_title},{$t_art_content},{$t_art_time},{$t_artx_author}) values('$art_title','$art_content','$art_time','$art_author')";
}else
{
$sql2="insert into {$t_table}({$t_art_title},{$t_art_content},{$t_art_time},{$t_artx_author},{$t_catx_id}) values('$art_title','$art_content','$art_time','$art_author','$cat_id')";
}
mysql_query($sql2);//插入文章表
$jl++;//如果存放数据库中 则记录加1
//处理数据库结束
}else
{
echo "长度不够";
}
}else
{
echo "主题不符合要求";
}
}else
{
echo "已经存在";
}
$rs++;
//记录下本次采集 的状况
$f_tt= urldecode($sp)."--页数".$pg." 记录数 ".$jl ;
file_put_contents("ss.txt",$f_tt);
echo "<script>location.href='wenwen.php?jl=".$jl."&sp=".$sp."&pg=".$pg."&rs=".$rs." ';</script>";
exit();
}
?>
<link href="style.css" rel="stylesheet" type="text/css" />
<table width="700" border="0" align="center" cellspacing="1" bgcolor="#CCCCCC">
<tr>
<td height="50" align="center" bgcolor="#00CC00"><h1><a href="http://www.jianlila.com">荐礼啦</a>知道问问采集插件</h1></td>
</tr>
</table>
<table width="700" border="0" align="center" cellspacing="1" bgcolor="#CCCCCC" style="margin-top:6px; margin-bottom:6px;">
<tr>
<td height="30" align="center" bgcolor="#FFFFFF"><a href="cj_config.php">采集设置</a> <a href="uninstall.php" onclick="return confirm('您确定要卸载采集插件吗');">卸载采集</a> <a href="cj_view.php">查看采集记录</a> <a href="cj_help.php">采集帮助</a> <a href="baidu.php" target="_blank">知道采集</a>  <a href="wenwen.php" target="_blank">问问采集</a></td>
</tr>
</table>
<form action="wenwen.php" method="post">
<table width="628" height="49" border="0" align="center">
<tr>
<td width="413" align="right"><input name="ask" type="text" id="ask" size="50"></td>
<td width="205"><input type="submit" name="button" id="button" value="问问采集" style=" padding-left:15px; padding-right:15px; height:25px; line-height:25px;"></td>
</tr>
</table>
</form>