用python写的一个wordpress的采集程序

#! /usr/bin/python
 import os,urllib2,re,time,MySQLdb,sys
 reTitle          = re.compile('<font[^>]*>(.*"\\n---------------- BLOG OF YAO"
 bianhao=2859
 for i in range(1,1500):
     Str2=""
     ltime=time.localtime(shijian)
     timeStr=time.strftime("%Y%m%d",ltime)
     url="http://www.jokeswarehouse.com/cgi-bin/viewjoke2.cgi" %timeStr
     print url
     a=urllib2.urlopen(url).read()
     Title=reTitle.findall(a)
     print "=========================================================================================================="
     for titles in map(None,Title):
         titles=MySQLdb.escape_string(titles)
         print titles
     Neiron=re.findall(reNeiron,a)
     for i in map(None,Neiron):
         x=re.findall(retiqu,i)
         for str in x:
             str=MySQLdb.escape_string(str)
             Str2 += str+"\\n"
     shijian += 86400
     bianhao += 1
     try:
         conn=MySQLdb.connect("XXXX.XXXX.XXXX.XXXX","user","passwd","dbname",charset="utf8", init_command="set names utf8")
     except MySQLdb.OperationalError,message:
         print "like error"
     cursor=conn.cursor()
     sql="INSERT INTO wp_posts (post_author,post_date,post_date_gmt,post_content,post_content_filtered,post_title,post_excerpt,post_status,post_type,comment_status,ping_status,post_password,post_name,to_ping,pinged,post_modified,post_modified_gmt,post_parent,menu_order,guid) VALUES (\'1\',\'2011-06-01 22:12:25\',\'2011-05-09 04:12:25\',\'\',\'\',\'Auto Draft\',\'\',\'inherit\',\'revision\',\'open\',\'open\',\'\',\'100-revision\',\'\',\'\',\'2011-06-01 22:12:25\',\'2011-05-09 04:12:25\',\'%s\',\'0\',\'\')" %bianhao
     sql2="UPDATE wp_posts SET post_author = 1, post_date = \'2011-06-01 22:12:25\', post_date_gmt = \'2011-06-01 22:12:25\', post_content =\'%s\', post_content_filtered = \'\', post_title = \'%s\', post_excerpt = \'\', post_status = \'publish\', post_type = \'post\', comment_status = \'open\', ping_status = \'open\', post_password = \'\', post_name = \'%s\', to_ping = \'\', pinged = \'\', post_modified = \'2011-06-01 22:12:25\', post_modified_gmt = \'2011-05-09 04:12:30\', post_parent = 0, menu_order = 0, guid = \'http://www.moncleronlineshops.com/" %(Str2,titles,titles,bianhao,bianhao)
     cursor.execute(sql)
     cursor.execute(sql2)
     cursor.close()
     conn.close()
     sys.exit()

#! /usr/bin/python
 import os,urllib2,re,time,MySQLdb,sys #加载本程序需要调用的相模块
reTitle          = re.compile('<font[^>]*>(.*"\\n---------------- BLOG OF YAO" #这个没用，开始是准备加到文章里的，后来没加进去。
bianhao=2859   #这里是wordpress 的文章编号，直接查看wp-posts表的id 字段的最后一个数字。

for i in range(1,1500): #循环1500遍，也就是采集1500篇文章。
    Str2="" #先赋值给Str2 空值
    ltime=time.localtime(shijian)  
     timeStr=time.strftime("%Y%m%d",ltime) #这两句将上面的时间戳改为时间，样式为19700101这样的格式
    url="http://www.jokeswarehouse.com/cgi-bin/viewjoke2.cgi" %timeStr #定义要采集的网站，将转化后的时间放在这个url的最后。
    a=urllib2.urlopen(url).read() #将这个网页的源代码读出来，赋值给a;
     Title=reTitle.findall(a)
 #使用 reTitle这个正则提取出标题
    print "=========================================================================================================="
     for titles in map(None,Title): #上面提取出来的标题前后都有一个 [] 
所以我们要写个for循环把前后的[]去掉，并转义成能直接插入mysql库的格式。
        titles=MySQLdb.escape_string(titles)
     Neiron=re.findall(reNeiron,a) #先用reNeiron，取个大概的内容模型出来。这些都是以逗号分隔的数组。
    for i in map(None,Neiron): # 我们来循环读出Neiron这个数组里的每个值。
        x=re.findall(retiqu,i)#并用 retiqu这个正则提出精细出的内容。
        for str in x:
             str=MySQLdb.escape_string(str)
             Str2 += str+"\\n"
 #利用这个循环，我们把内容加到一起，并赋值给Str2这个变量，这个　Str2这个变量就是所有的文章内容。
    shijian += 86400 #每循环一次，就把shijian这个变量加上一天。
    bianhao += 1   #每循环一次，就把bianhao这个变量加上一
    try:
 #下面是用mysqldb连接数据库，并尝试连接是否成功。       conn=MySQLdb.connect("XXXX.XXXX.XXXX.XXXX","user","passwd","dbname",charset="utf8", init_command="set names utf8")
     except MySQLdb.OperationalError,message:
         print "like error"
     cursor=conn.cursor()
 #下面是插入wordpress数据库的两条语句，我是从mysqlbinlog里面导出来的，测试是可以插入数据库，并能正常把内容显示在网页的。变量都写在这两条语句里。
    sql="INSERT INTO wp_posts (post_author,post_date,post_date_gmt,post_content,post_content_filtered,post_title,post_excerpt,post_status,post_type,comment_status,ping_status,post_password,post_name,to_ping,pinged,post_modified,post_modified_gmt,post_parent,menu_order,guid) VALUES (\'1\',\'2011-06-01 22:12:25\',\'2011-05-09 04:12:25\',\'\',\'\',\'Auto Draft\',\'\',\'inherit\',\'revision\',\'open\',\'open\',\'\',\'100-revision\',\'\',\'\',\'2011-06-01 22:12:25\',\'2011-05-09 04:12:25\',\'%s\',\'0\',\'\')" %bianhao
     sql2="UPDATE wp_posts SET post_author = 1, post_date = \'2011-06-01 22:12:25\', post_date_gmt = \'2011-06-01 22:12:25\', post_content =\'%s\', post_content_filtered = \'\', post_title = \'%s\', post_excerpt = \'\', post_status = \'publish\', post_type = \'post\', comment_status = \'open\', ping_status = \'open\', post_password = \'\', post_name = \'%s\', to_ping = \'\', pinged = \'\', post_modified = \'2011-06-01 22:12:25\', post_modified_gmt = \'2011-05-09 04:12:30\', post_parent = 0, menu_order = 0, guid = \'http://www.moncleronlineshops.com/" %(Str2,titles,titles,bianhao,bianhao)
     cursor.execute(sql)
     cursor.execute(sql2) #连接数据库并执行这两条语句。
    cursor.close()
     conn.close()  #关闭数据库。
    sys.exit()