PHP 论坛采集程序 模拟登陆，抓取页面 实现代码

代码如下:

1. <?php
   
2. // 吴燕军
   
3. // 2009-06-27
   
4. // 采集程序php
   
5. set_time_limit(0);
   
6. //cookie保存目录
   
7. $cookie_jar = '/tmp/cookie.tmp';
   
8. /*函数------------------------------------------------------------------------------------------------------------*/
   
9. //模拟请求数据
   
10. function request($url,$postfields,$cookie_jar,$referer){
    
11. $ch = curl_init();
    
12. $options = array(CURLOPT_URL => $url,
    
13. CURLOPT_HEADER => 0,
    
14. CURLOPT_NOBODY => 0,
    
15. CURLOPT_PORT => 80,
    
16. CURLOPT_POST => 1,
    
17. CURLOPT_POSTFIELDS => $postfields,
    
18. CURLOPT_RETURNTRANSFER => 1,
    
19. CURLOPT_FOLLOWLOCATION => 1,
    
20. CURLOPT_COOKIEJAR => $cookie_jar,
    
21. CURLOPT_COOKIEFILE => $cookie_jar,
    
22. CURLOPT_REFERER => $referer
    
23. );
    
24. curl_setopt_array($ch, $options);
    
25. $code = curl_exec($ch);
    
26. curl_close($ch);
    
27. return $code;
    
28. }
    
29. //获取帖子列表
    
30. function getThreadsList($code){
    
31. preg_match_all('/ <!--[.|\r|\n]*? <a href="viewthread.php\?tid=(\d+)/',$code,$threads);
    
32. return $threads[1];
    
33. }
    
34. //判断该帖子是否存在
    
35. function isExits($code){
    
36. 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);
    
37. return isset($error[0])?false:true;
    
38. }
    
39. //获取帖子标题
    
40. function getTitle($code){
    
41. preg_match('/ <h1>[^ <\/h1>]*/',$code,$title_tmp);
    
42. $title = $title_tmp[0];
    
43. return $title;
    
44. }
    
45. //获取帖子作者:
    
46. function getAuthor($code){
    
47. preg_match('/ <a href="space.php\?uid=\d+" target="_blank" id="userinfo\d+" onmouseover="showMenu\(this\.id\)">.+/',$code,$author_tmp);
    
48. $author = strip_tags($author_tmp[0]);
    
49. return $author;
    
50. }
    
51. //获取楼主发表的内容
    
52. function getContents($code){
    
53. preg_match('/ <div id="postmessage_\d+" class="t_msgfont">(.|\r|\n)*? <\/div>/',$code,$contents_tmp);
    
54. $contents = preg_replace('/images\//','http://bbs.war3.cn/images/',$contents_tmp[0]);
    
55. return $contents;
    
56. }
    
57. //打印帖子标题
    
58. function printTitle($title){
    
59. echo " <strong> <h2>帖子标题: </h2> </strong>",strip_tags($title)," <br/> <br/>";
    
60. }
    
61. //输出帖子作者
    
62. function printAuthor($author){
    
63. echo " <strong> <h2>帖子作者: </h2> </strong>",strip_tags($author)," <br/> <br/>";
    
64. }
    
65. //打印帖子内容
    
66. function printContents($contents){
    
67. echo " <strong> <h2>作者发表的内容: </h2>",$contents," </strong> <br/>";
    
68. }
    
69. //错误
    
70. function printError(){
    
71. echo " <i>该帖子不存在! </i>";
    
72. }
    
73. /*函数列表end---------------------------------------------------------------------------------------------------*/
    
74. /*登录论坛 begin*/
    
75. $url = 'http://bbs.war3.cn/logging.php?action=login';
    
76. $postfields='loginfield=username&username=1nject10n& password=xxxxxx&questionid=0&cookietime=315360000& referer=http://bbs.war3.cn/&loginsubmit=提交';
    
77. request($url,$postfields,$cookie_jar,'');
    
78. unset($postfields,$url);
    
79. /*登录论坛 end*/
    
80. /*获取帖子列表(位于第一页的帖子) begin*/
    
81. $url = 'http://bbs.war3.cn/forumdisplay.php?fid=57';
    
82. $code = request($url,'',$cookie_jar,'');
    
83. $threadsList = getThreadsList($code);
    
84. /*获取帖子列表 end*/
    
85. //帖子序列
    
86. $rows = 0;
    
87. /*循环抓取所有帖子源代码 begin*/
    
88. foreach($threadsList as $list){
    
89. $url = "http://bbs.war3.cn/viewthread.php?tid=$list";
    
90. if(isExits($code)){
    
91. $code = request($url,'',$cookie_jar,'');
    
92. $color = $rows%2==0?'#00CCFF':'#FFFF33';
    
93. echo " <div style='background-color:$color'>";
    
94. echo " <h1>第",($rows+1),"贴: </h1> <br/>";
    
95. $author = getAuthor($code);
    
96. printAuthor($author);
    
97. $title = getTitle($code);
    
98. printTitle($title);
    
99. $contents = getContents($code);
    
100. printContents($contents);
     
101. echo " </div>";
     
102. $rows++;
     
103. }
     
104. else
     
105. printError();
     
106. echo "----------------------------------------------------------------------------------------- <br/> <br/>";
     
107. }
     
108. /*抓取源代码 end*/
     
109. ?>
     

复制代码