#! /usr/bin/python
 import os,urllib2,re,time,MySQLdb,sys
 reTitle          = re.compile('<font[^>]*>(.*?)<\/font><font[^>]*')
 reNeiron         = re.compile('[1-9|A-Z|a-z].*')
 retiqu          = re.compile('^(?!MARGINWIDTH|BR).*.[^>|}]$')
 rezhong          = re.compile('^[^[].*')
 shijian=1190944000
 Str1="\\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?id=%s" %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/?p=%s\' WHERE ID = %s" %(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 #加載本程序需要調(diào)用的相模塊
reTitle          = re.compile('<font[^>]*>(.*?)<\/font> <font[^>]*') # 定義一下取文章標(biāo)題的正則
reNeiron         = re.compile('[1-9|A-Z|a-z].*') 
 #定義一個(gè)取提取文章內(nèi)容的正則(注：這里提取出來的不是很精細(xì)，需要在下面的正則里，再進(jìn)行提取，這里只是取一個(gè)大概)
retiqu          = re.compile('^(?!MARGINWIDTH|BR).*.[^>|}]$')
 #這里定義一個(gè)正則，將上面reNeiron提取出來的字符，再進(jìn)行細(xì)化。

shijian=1190944000  #這里字義了一個(gè)時(shí)間戳，
Str1="\\n---------------- BLOG OF YAO" #這個(gè)沒用，開始是準(zhǔn)備加到文章里的，后來沒加進(jìn)去。
bianhao=2859   #這里是wordpress 的文章編號(hào)，直接查看wp-posts表的id 字段的最后一個(gè)數(shù)字。

for i in range(1,1500): #循環(huán)1500遍，也就是采集1500篇文章。
    Str2="" #先賦值給Str2 空值
    ltime=time.localtime(shijian)  
     timeStr=time.strftime("%Y%m%d",ltime) #這兩句將上面的時(shí)間戳改為時(shí)間，樣式為19700101這樣的格式
    url="http://www.jokeswarehouse.com/cgi-bin/viewjoke2.cgi?id=%s" %timeStr #定義要采集的網(wǎng)站，將轉(zhuǎn)化后的時(shí)間放在這個(gè)url的最后。
    a=urllib2.urlopen(url).read() #將這個(gè)網(wǎng)頁的源代碼讀出來，賦值給a;
     Title=reTitle.findall(a)
 #使用 reTitle這個(gè)正則提取出標(biāo)題
    print "=========================================================================================================="
     for titles in map(None,Title): #上面提取出來的標(biāo)題前后都有一個(gè) [] 
所以我們要寫個(gè)for循環(huán)把前后的[]去掉，并轉(zhuǎn)義成能直接插入mysql庫的格式。
        titles=MySQLdb.escape_string(titles)
     Neiron=re.findall(reNeiron,a) #先用reNeiron，取個(gè)大概的內(nèi)容模型出來。這些都是以逗號(hào)分隔的數(shù)組。
    for i in map(None,Neiron): # 我們來循環(huán)讀出Neiron這個(gè)數(shù)組里的每個(gè)值。
        x=re.findall(retiqu,i)#并用 retiqu這個(gè)正則提出精細(xì)出的內(nèi)容。
        for str in x:
             str=MySQLdb.escape_string(str)
             Str2 += str+"\\n"
 #利用這個(gè)循環(huán)，我們把內(nèi)容加到一起，并賦值給Str2這個(gè)變量，這個(gè)　Str2這個(gè)變量就是所有的文章內(nèi)容。
    shijian += 86400 #每循環(huán)一次，就把shijian這個(gè)變量加上一天。
    bianhao += 1   #每循環(huán)一次，就把bianhao這個(gè)變量加上一
    try:
 #下面是用mysqldb連接數(shù)據(jù)庫，并嘗試連接是否成功。       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數(shù)據(jù)庫的兩條語句，我是從mysqlbinlog里面導(dǎo)出來的，測(cè)試是可以插入數(shù)據(jù)庫，并能正常把內(nèi)容顯示在網(wǎng)頁的。變量都寫在這兩條語句里。
    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/?p=%s\' WHERE ID = %s" %(Str2,titles,titles,bianhao,bianhao)
     cursor.execute(sql)
     cursor.execute(sql2) #連接數(shù)據(jù)庫并執(zhí)行這兩條語句。
    cursor.close()
     conn.close()  #關(guān)閉數(shù)據(jù)庫。
    sys.exit()