一、版本
Python版本 x64 2.7.12
Oracle(x64 12.1.0.2.0)和Elasticsearch(2.2.0)
python编辑器 PyCharm
下载安装请选择适合自己机器的版本
二、下载模块
通过官网下载和安装cx_Oracle和pyes模块,分别用于操作Oracle数据库和ES。安装fcntl模块用于解决python脚本单例执行问题。
如果是远程连接数据库和ES,请一定注意安装的模块或包版本。务必选择相应的版本,不然会遇到问题。
三、安装过程中会遇到的问题
cx_Oracle在本地安装过程中出现的一些问题:
1、安装c++for python的环境
2、安装Oracle数据库(或者安装API接口中需要的文件而不必下载配置整个oracle环境)
3、打开数据库工具 oracle SQL developor 按要求创建连接,并新建用户(创建数据库用户名时以c##开头,不然会提示)
4、oracle连接不上远程的服务器,检查版本是否匹配
fcntl在windows上安装时出现的问题:
1、用pip install fcntl 报错:indentationerror: unexpected indent(模块版本有问题)
四、源码
- # -*- coding: utf-8 -*-
- """
- 作者:陈龙
- 日期:2016-7-22
- 功能:oracle数据库到ES的数据同步
- """
- import os
- import sys
- import datetime, time
- # import fcntl
- import threading
- import pyes # 引入pyes模块,ES接口
- import cx_Oracle # 引入cx_Oracle模块,Oracle接口
- os.environ['NLS_LANG'] = 'SIMPLIFIED CHINESE_CHINA.UTF8' # 中文编码
- reload(sys) # 默认编码设置为utf-8
- sys.setdefaultencoding('utf-8')
- # 创建ES连接 并返回连接参数
- def connect_ES(addr):
- try:
- global conn
- conn = pyes.ES(addr) # 链接ES '127.0.0.1:9200'
- print 'ES连接成功'
- return conn
- except:
- print 'ES连接错误'
- pass
- # 创建ES映射mapping 注意各各个字段的类型
- def create_ESmapping():
- global spiderInfo_mapping, involveVideo_mapping, involveCeefax_mapping,keyWord_mapping,sensitiveWord_mapping
- spiderInfo_mapping = {'tableName': {'index': 'not_analyzed', 'type': 'string'},
- 'tableId': {'index': 'not_analyzed', 'type': 'integer'},
- 'title': {'index': 'analyzed', 'type': 'string'},
- 'author': {'index': 'not_analyzed', 'type': 'string'},
- 'content': {'index': 'analyzed', 'type': 'string'},
- 'publishTime': {'index': 'not_analyzed', 'type': 'string'},
- 'browseNum': {'index': 'not_analyzed', 'type': 'integer'},
- 'commentNum': {'index': 'not_analyzed', 'type': 'integer'},
- 'dataType': {'index': 'not_analyzed', 'type': 'integer'}} # 除去涉我部分内容的ES映射结构
- involveVideo_mapping = {'tableName': {'index': 'not_analyzed', 'type': 'string'},
- 'tableId': {'index': 'not_analyzed', 'type': 'integer'},
- 'title': {'index': 'analyzed', 'type': 'string'},
- 'author': {'index': 'not_analyzed', 'type': 'string'},
- 'summary': {'index': 'analyzed', 'type': 'string'},
- 'publishTime': {'index': 'not_analyzed', 'type': 'string'},
- 'url': {'index': 'not_analyzed', 'type': 'string'},
- 'imgUrl': {'index': 'not_analyzed', 'type': 'string'},
- 'ranking': {'index': 'not_analyzed', 'type': 'integer'},
- 'playNum': {'index': 'not_analyzed', 'type': 'integer'},
- 'dataType': {'index': 'not_analyzed', 'type': 'integer'}} # 涉我视音频内容的ES映射结构
- involveCeefax_mapping = {'tableName': {'index': 'not_analyzed', 'type': 'string'},
- 'tableId': {'index': 'not_analyzed', 'type': 'integer'},
- 'title': {'index': 'analyzed', 'type': 'string'},
- 'author': {'index': 'not_analyzed', 'type': 'string'},
- 'content': {'index': 'analyzed', 'type': 'string'},
- 'publishTime': {'index': 'not_analyzed', 'type': 'string'},
- 'keyWords': {'index': 'not_analyzed', 'type': 'string'},
- 'popularity': {'index': 'not_analyzed', 'type': 'integer'},
- 'url': {'index': 'not_analyzed', 'type': 'string'},
- 'dataType': {'index': 'not_analyzed', 'type': 'integer'}} # 涉我图文资讯内容的ES映射结构
- keyWord_mapping = {'id':{'index': 'not_analyzed', 'type': 'integer'},
- 'keywords':{'index': 'not_analyzed', 'type': 'string'}}
- sensitiveWord_mapping = {'id':{'index': 'not_analyzed', 'type': 'integer'},
- 'sensitiveType':{'index': 'not_analyzed', 'type': 'string'},
- 'sensitiveTopic': {'index': 'not_analyzed', 'type': 'string'},
- 'sensitiveWords': {'index': 'not_analyzed', 'type': 'string'}}
- # 创建ES相关索引和索引下的type
- def create_ESindex(ES_index, index_type1,index_type2,index_type3,index_type4,index_type5):
- if conn.indices.exists_index(ES_index):
- pass
- else:
- conn.indices.create_index(ES_index) # 如果所有Str不存在,则创建Str索引
- create_ESmapping()
- conn.indices.put_mapping(index_type1, {'properties': spiderInfo_mapping},[ES_index]) # 在索引pom下创建spiderInfo的_type "spiderInfo"
- conn.indices.put_mapping(index_type2, {'properties': involveVideo_mapping},[ES_index]) # 在索引pom下创建involveVideo的_type "involveVideo"
- conn.indices.put_mapping(index_type3, {'properties': involveCeefax_mapping},[ES_index]) # 在索引pom下创建involveCeefax的_type "involveCeefax"
- conn.indices.put_mapping(index_type4, {'properties': keyWord_mapping}, [ES_index])
- conn.indices.put_mapping(index_type5, {'properties': sensitiveWord_mapping}, [ES_index])
- # conn.ensure_index
- # 创建数据库连接 并返回连接参数
- def connect_Oracle(name, password, address):
- try:
- global conn1
- # conn1 = cx_Oracle.connect('c##chenlong','1234567890','localhost:1521/ORCL') #链接本地数据库
- conn1 = cx_Oracle.connect(name, password, address) # 链接远程数据库 "pom","Bohui@123","172.17.7.118:1521/ORCL"
- print 'Oracle连接成功'
- return conn1
- except:
- print 'ES数据同步脚本连接不上数据库,请检查connect参数是否正确,或者模块版本是否匹配'
- pass
- def fetch_account(accountcode): # 取两个‘_’之间的账号名称
- end = accountcode.find('_')
- return accountcode[0:end].strip()
- # 根据表的个数创建不同的对象
- # 从记录文档中读取各个表的记录ID,判断各个表的ID是否有变化
- # 分别读取各个表中的相关数据
- # 读取各个表的ID与记录的ID(记录在文本或者数据库中)并判断
- """def read_compare_ID():
- global tuple_tableName_IdNum
- global cur
- tuple_tableName_IdNum = {}
- tablename = []
- cur = conn1.cursor()
- result1 = cur.execute("select * from tabs") ##执行数据库操作 读取各个表名
- row = result1.fetchall()
- for x in row:
- tablename.append(x[0]) # 将表名取出并赋值给tablename数组
- result2 = cur.execute('select {}_ID from {}'.format(x[0], x[0]))
- ID_num = result2.fetchall()
- tuple_tableName_IdNum[x[0]] = ID_num"""
- def readOracle_writeES(tableName, ES_index, index_type):
- global cc
- cur = conn1.cursor()
- #result_AlltableNames = cur.execute("select * from tabs")
- result_latestId = cur.execute("select max({}_Id) from {} ".format(tableName,tableName))
- num1 = result_latestId.fetchone() #当前表中的最大ID
- print '当前表中的最大ID{}'.format(num1[0])
- result_rememberId = cur.execute("select tableId from T_REMEMBERID where tableName='{}'".format(tableName.upper())) #通过数据库表拿到更新的ID tablename 都转化成大写
- num2 = result_rememberId.fetchone() #上次记录的更新ID
- print '上次记录的更新ID{}'.format(num2[0])
- if tableName.upper() == 'T_SOCIAL':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute("select {}_ID,title,author,content,publishTime,browseNum,likeNum,forwardNum,commentNum,accountCode from {} where {}_ID > {} and rownum<=40 ".format(tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchall() #之前是因为数据量太大,超过了变量的内存空间,所以用fetchmany取40条 后来大神建议数据库中限制查询数 然后fetchall,这样查询更有效率
- for i in result_tuple1: #一条一条写入ES,这个速度太慢,改进 通过bulk接口导入
- aa= (i[5]+i[6])
- bb= (i[7]+i[8])
- if conn.index(
- {'tableName': tableName, 'tableId': i[0], 'title': unicode(i[1]), 'author': unicode(i[2]),
- 'content': unicode(i[3]), 'publishTime': str(i[4]), 'browseNum': aa,
- 'commentNum':bb, 'dataType':fetch_account(i[9])}, ES_index, index_type,bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0] #如果写入成功才赋值
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId,tableName))
- conn1.commit()
- result_rememberId = cur.execute("select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读{}写成功".format(tableName,index_type)
- if tableName.upper() == 'T_HOTSEARCH':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute("select {}_ID,accountCode,title,publishTime from {} where {}_ID > {} and rownum<=40 ".format(tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchall() # 之前是因为数据量太大,超过了变量的内存空间,所以用fetchmany取40条 后来大神建议数据库中限制查询数 然后fetchall,这样查询更有效率
- for i in result_tuple1: #一条一条写入ES,这个速度太慢,改进 通过bulk接口导入
- if conn.index(
- {'tableName': tableName, 'tableId': i[0], 'title': unicode(i[2]),'author': '','content': '', 'publishTime': str(i[3]), 'browseNum': 0,
- 'commentNum': 0, 'dataType': fetch_account(i[1])}, ES_index, index_type,bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId, tableName))
- conn1.commit()
- result_rememberId = cur.execute("select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读{}写成功".format(tableName, index_type)
- if tableName.upper() == 'T_VIDEO_HOT':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute("select {}_ID,accountCode,title,Author,publishTime from {} where {}_ID > {} and rownum<=40 ".format(tableName,tableName,tableName,num2[0]))
- result_tuple1 = result_readOracle.fetchall() # 之前是因为数据量太大,超过了变量的内存空间,所以用fetchmany取40条 后来大神建议数据库中限制查询数 然后fetchall,这样查询更有效率
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- if conn.index(
- {'tableName': tableName, 'tableId': i[0], 'title': unicode(i[2]),'author': unicode(i[3]),
- 'content': '', 'publishTime': str(i[4]), 'browseNum': 0,
- 'commentNum': 0, 'dataType': fetch_account(i[1])}, ES_index, index_type, bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId, tableName))
- conn1.commit()
- result_rememberId = cur.execute("select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读写成功".format(tableName)
- if tableName.upper() == 'T_PRESS':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute(
- "select {}_ID,accountCode,title,Author,PublishDate,Content from {} where {}_ID > {} and rownum<=40 ".format(
- tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchall() # 之前是因为数据量太大,超过了变量的内存空间,所以用fetchmany取40条 后来大神建议数据库中限制查询数 然后fetchall,这样查询更有效率
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- if conn.index(
- {'tableName': tableName, 'tableId': i[0], 'title': unicode(i[2]),'author': unicode(i[3]),
- 'content': unicode(i[5]), 'publishTime': str(i[4]), 'browseNum': 0,
- 'commentNum': 0, 'dataType': fetch_account(i[1])}, ES_index, index_type,bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId, tableName))
- conn1.commit()
- result_rememberId = cur.execute(
- "select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读写成功".format(tableName)
- if tableName.upper() == 'T_INDUSTRY':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute(
- "select {}_ID,accountCode,title,Author,PublishTime,Content,BrowseNum from {} where {}_ID > {} and rownum<=40 ".format(
- tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchall() # 之前是因为数据量太大,超过了变量的内存空间,所以用fetchmany取40条 后来大神建议数据库中限制查询数 然后fetchall,这样查询更有效率
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- if conn.index(
- {'tableName': tableName, 'tableId': i[0], 'title': unicode(i[2]),'author': unicode(i[3]),
- 'content': unicode(i[5]), 'publishTime': str(i[4]), 'browseNum': i[6],
- 'commentNum': 0, 'dataType': fetch_account(i[1])}, ES_index, index_type,bulk=True) : # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId, tableName))
- conn1.commit()
- result_rememberId = cur.execute(
- "select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读写成功".format(tableName)
- if tableName.upper() == 'T_SOCIAL_SITESEARCH':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute('select {}_ID,title,author,content,publishTime,keyWords,browseNum,likeNum,forwardNum,commentNum,url,accountCode from {} where ({}_ID > {})'.format(tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchmany(50) #因为数据量太大,超过了变量的内存空间,所以一次性取40条
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- popularity = (i[6] + i[7] + i[8] * 2 + i[9] * 2)
- if conn.index(
- {'tableName': tableName,'tableId':i[0],'title': unicode(i[1]),'author':unicode(i[2]),
- 'content':unicode(i[3]),'publishTime':str(i[4]),'keyWords':unicode(i[5]),
- 'popularity':popularity,'url': i[10],
- 'dataType':fetch_account(i[11])}, ES_index, index_type, bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId,tableName))
- conn1.commit()
- result_rememberId = cur.execute("select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读写成功".format(tableName)
- if tableName.upper() == 'T_REALTIME_NEWS':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute("select {}_ID,title,author,content,publishTime,browseNum,commentNum,accountCode,url from {} where {}_ID > {} and rownum<=40 ".format(tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchall() # 之前是因为数据量太大,超过了变量的内存空间,所以用fetchmany取40条 后来大神建议数据库中限制查询数 然后fetchall,这样查询更有效率
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- popularity = (i[5] + i[6] * 2)
- if conn.index(
- {'tableName': tableName,'tableId':i[0],'title': unicode(i[1]),'author':unicode(i[2]),
- 'content':unicode(i[3]),'publishTime':str(i[4]),'keyWords':unicode(''),
- 'popularity':popularity,'url': i[8],'dataType':fetch_account(i[7])}, ES_index, index_type, bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId, tableName))
- conn1.commit()
- result_rememberId = cur.execute(
- "select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读{}写成功".format(tableName, index_type)
- if tableName.upper() == 'T_KEY_NEWS':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute("select {}_ID,title,author,content,publishTime,browseNum,commentNum,accountCode,url from {} where {}_ID > {} and rownum<=40 ".format(tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchall() # 之前是因为数据量太大,超过了变量的内存空间,所以用fetchmany取40条 后来大神建议数据库中限制查询数 然后fetchall,这样查询更有效率
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- popularity = (i[5] + i[6] * 2)
- if conn.index(
- {'tableName': tableName,'tableId':i[0],'title': unicode(i[1]),'author':unicode(i[2]),
- 'content':unicode(i[3]),'publishTime':str(i[4]),'keyWords':unicode(''),
- 'popularity':popularity,'url': i[8],'dataType':fetch_account(i[7])}, ES_index, index_type, bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId, tableName))
- conn1.commit()
- result_rememberId = cur.execute(
- "select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读{}写成功".format(tableName, index_type)
- if tableName.upper() == 'T_LOCAL_NEWS':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute("select {}_ID,title,author,content,publishTime,browseNum,commentNum,accountCode,url from {} where {}_ID > {} and rownum<=40 ".format(tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchall() # 之前是因为数据量太大,超过了变量的内存空间,所以用fetchmany取40条 后来大神建议数据库中限制查询数 然后fetchall,这样查询更有效率
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- popularity = (i[5] + i[6] * 2)
- if conn.index(
- {'tableName': tableName, 'tableId': i[0], 'title': unicode(i[1]), 'author': unicode(i[2]),
- 'content': unicode(i[3]), 'publishTime': str(i[4]), 'keyWords': unicode(''),
- 'popularity': popularity, 'url': i[8], 'dataType': fetch_account(i[7])}, ES_index, index_type,bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId, tableName))
- conn1.commit()
- result_rememberId = cur.execute(
- "select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读{}写成功".format(tableName, index_type)
- if tableName.upper() == 'T_VIDEO_SITESEARCH':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute("select {}_ID,accountCode,title,Author,publishTime,url,imgUrl,playNum,keyWords from {} where {}_ID > {} and rownum<=40 ".format(tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchall() # 之前是因为数据量太大,超过了变量的内存空间,所以用fetchmany取40条 后来大神建议数据库中限制查询数 然后fetchall,这样查询更有效率
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- if conn.index(
- {
- 'tableName': tableName, 'tableId': i[0], 'title': unicode(i[2]), 'author': unicode(i[3]),
- 'summary': unicode('0'), 'publishTime': str(i[4]), 'browseNum': i[7],'url':i[5],'imgUrl':i[6],'ranking':0,
- 'playNum': 0, 'dataType': fetch_account(i[1])}, ES_index, index_type,bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId, tableName))
- conn1.commit()
- result_rememberId = cur.execute(
- "select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读{}写成功".format(tableName,index_type)
- if tableName.upper() == 'T_BASE_KEYWORDS':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute('select {}_ID,keywords from {} where {}_ID > {} and rownum<=50'.format(tableName, tableName, tableName, num2[0]))
- result_tuple1 = result_readOracle.fetchall() #因为数据量太大,超过了变量的内存空间,所以一次性取40条
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- if conn.index({'id': i[0], 'keywords': i[1]}, ES_index, index_type,bulk=True): # 将数据写入索引pom的spiderInfo
- cc += 1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId,tableName))
- conn1.commit()
- result_rememberId = cur.execute("select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读写成功".format(tableName)
- if tableName.upper() == 'T_BASE_SENSITIVEWORDS':
- while num2[0] < num1[0]:
- result_readOracle = cur.execute('select {}_ID,SensitiveType,SensitiveTopic,SensitiveWords from {} where {}_ID > {} and rownum<=50'.format(tableName, tableName, tableName,num2[0]))
- result_tuple1 = result_readOracle.fetchall() # 因为数据量太大,超过了变量的内存空间,所以一次性取40条
- for i in result_tuple1: # 一条一条写入ES,这个速度太慢,强烈需要改进 通过bulk接口导入?
- if conn.index({'id':i[0],
- 'sensitiveType':unicode(i[1]),
- 'sensitiveTopic': unicode(i[2]),
- 'sensitiveWords':unicode(i[3])}, ES_index, index_type, bulk=True): # 将数据写入索引pom的spiderInfo
- cc +=1
- print 'bulk导入后的ID:{}'.format(i[0])
- rememberId = i[0]
- cur.execute("update T_REMEMBERID set tableId = {} where tableName = '{}'".format(rememberId, tableName))
- conn1.commit()
- result_rememberId = cur.execute("select tableId from T_REMEMBERID where tableName='{}'".format(tableName)) # 通过数据库表拿到更新的ID
- num2 = result_rememberId.fetchone()
- print "{}读写成功".format(tableName)
- else:
- pass
- def ww(a):
- while True:
- print a
- time.sleep(0.5) #用于多线程的一个实验函数
- if __name__ == "__main__":
- cc = 0
- connect_ES('172.17.5.66:9200')
- # conn.indices.delete_index('_all') # 清除所有索引
- create_ESindex("pom", "spiderInfo", "involveVideo", "involveCeefax","keyWord","sensitiveWord")
- connect_Oracle("pom", "Bohui@123", "172.17.7.118:1521/ORCL")
- # thread.start_new_thread(readOracle_writeES,("T_SOCIAL","pom","spiderInfo"),)#创建一个多线程
- # thread.start_new_thread(readOracle_writeES,("T_SOCIAL_SITESEARCH", "pom", "spiderInfo"),)#创建一个多线程
- mm = time.clock()
- readOracle_writeES("T_SOCIAL", "pom", "spiderInfo") #表名虽然在程序中设置了转化为大写,但是还是全大写比较好
- readOracle_writeES("T_HOTSEARCH", "pom", "spiderInfo")
- readOracle_writeES("T_VIDEO_HOT", "pom", "spiderInfo")
- readOracle_writeES("T_PRESS", "pom", "spiderInfo")
- readOracle_writeES("T_INDUSTRY", "pom", "spiderInfo")
- readOracle_writeES("T_VIDEO_SITESEARCH", "pom", "involveVideo")
- readOracle_writeES("T_REALTIME_NEWS", "pom", "involveCeefax")
- readOracle_writeES("T_KEY_NEWS", "pom", "involveCeefax")
- readOracle_writeES("T_LOCAL_NEWS", "pom", "involveCeefax")
- readOracle_writeES("T_SOCIAL_SITESEARCH", "pom", "involveCeefax")
- readOracle_writeES("T_BASE_KEYWORDS", "pom", "keyWord")
- readOracle_writeES("T_BASE_SENSITIVEWORDS", "pom", "sensitiveWord")
- nn = time.clock()
- # conn.indices.close_index('pom')
- conn1.close()
- print '数据写入耗时:{} 成功写入数据{}条'.format(nn-mm,cc)
- #实验多线程
- """
- while a < 100:
- conn.index(
- {'tableName': 'T_base_account', 'type': '1', 'tableId': '123', 'title': unicode('陈龙'), 'author': 'ABC',
- 'content': 'ABC', 'publishTime': '12:00:00', 'browseNum': '12', 'commentNum': '12', 'dataType': '1'},
- "pom", "spiderInfo", ) # 将数据写入索引pom的spiderInfo
- a += 1
- print time.ctime()
- """
- """
- threads = []
- t1 = threading.Thread(target=readOracle_writeES,args=("T_SOCIAL","pom","spiderInfo"))
- threads.append(t1)
- #t3 = threading.Thread(target=ww,args=(10,))
- #threads.append(t3)
- #t2 = threading.Thread(target=readOracle_writeES,args=("T_SOCIAL_SITESEARCH", "pom", "spiderInfo"))
- #threads.append(t2)
- print time.ctime()
- for t in threads:
- t.setDaemon(True)
- t.start()
- t.join()
- """
五、编译过程的问题
1、直接print游标cur.execute ( ) 将不能得到我们想要的结果
result2 = cur.execute('select T_SOCIAL_ID from T_SOCIAL')
print result2
返回:<__builtin__.OracleCursor on <cx_Oracle.Connection to pom@172.17.7.118:1521/ORCL>>
result2 = cur.execute('select T_SOCIAL_ID from T_SOCIAL')
print result2
num = result2.fetchall()
print num
for i in num:
print i[0]
返回:[(55,), (56,), (57,), (58,), (59,), (60,), (61,), (62,), (63,), (64,), (65,), (66,), (67,), (68,), (69,), (70,)]
55
注意:用fetchall()得到的数据为:[(55,), (56,), (57,), (58,), (59,)] 元组而不是数字。
用 变量[num] 的方式取出具体的数值
2、cx_Oracle中文编码乱码问题
显示中文乱码:������DZ��� �����������
或者显示未知的编码:('\xce\xd2\xd5\xe6\xb5\xc4\xca\xc7\xb1\xea\xcc\xe2',)
需要注意一下几个地方,将数据库中的中文编码转化成utf-8编码,并将中文写入elasticsearch
os.environ['NLS_LANG'] = 'SIMPLIFIED CHINESE_CHINA.UTF8' #中文编码
reload(sys) #默认编码设置为utf-8 一定需要reload(sys)
sys.setdefaultencoding('utf-8')
'title':unicode('中文')
python传递给js的列表中文乱码怎么解决?
json.dumps(dictionary,ensure_ascii=False)
3、远程连接不上Oracle数据库的问题
第一:确保connect()中各个参数的值都正确。例如
conn1 = cx_Oracle.connect("username","password","172.17.7.118:1521/ORCL") #连接远程数据库
conn1 = cx_Oracle.connect('username','password','localhost:1521/ORCL') #连接本地数据库
conn2 = pyes.ES('127.0.0.1:9200') #连接ES
第二:确保安装的版本都符合要求,包括模块的版本。
4、提示TypeError: 'NoneType' object is not callable
确保mapping中的各个字段类型都设置正确
检查索引和映射是否都书写正确
5、脚本同时读取多个数据库表
涉及到Python中多线程的问题,给每一个表起一个线程,同时给每一个线程加锁
编译时碰到问题:AssertionError: group argument must be None for now(检查函数是否书写正确,读写冲突)
AttributeError: 'builtin_function_or_method' object has no attribute 'setDaemon'
cx_Oracle.ProgrammingError: LOB variable no longer valid after subsequent fetch(fetchall数据量过大,溢出 设置一次取数据库中 rownum数)
TypeError: 'NoneType' object has no attribute '__getitem__' (注意数据库查询对应的大小写)
No handlers could be found for logger "pyes" 可能是连接超时
AttributeError: 'tuple' object has no attribute 'append' tuple不能直接用append
TypeError: 'tuple' object does not support item assignment tuple不能赋值
数据库批量读取
就多线程问题咨询了大神,大神建议用多进程来实现会比较简单
6、脚本定时触发问题
Linux crontab定时执行任务,crontab防止脚本周期内未执行完重复执行
7、单实例的问题。防止脚本没有执行完再次触发
刚开始设想在脚本中完成,后来知道这个可以在系统中设定
8、数据同步插件
网上有大量的关于同步关系型数据库的有关插件 logstash-input-jdbc 不太好安装,不知道如何使用。
MySQL和ES同步插件的介绍,例如elasticsearch-river-jdbc
在这儿启用的是bulk接口,批量导入。数据同步的速度大大提高
9、判断数据是否同步成功
这个是之前一直没有注意的问题,但其实在数据传输的时候是非常重要的。
目前的判断方法是看ES中的数据量到底有多少,然后对照统计量进行判断分析,,这也是在后期发现有部分数据没有同步过去的方法。
10、统计写入了多少数据
UnboundLocalError: local variable 'cc' referenced before assignment
定义了全局变量cc,但是在局部进行了修改,所以报错 修改同名的全局变量,则认为是一个局部变量
五、源码改进
因为数据写入的速度太慢(40条数据 800Kb大小 写入花费2S左右),所有在原来的基础上,修改了读取数据库中未写入内容的策略和ES写入的策略。
插入完成的源码
调试问题:
1、pip install elasticsearch 引入helpers函数模块,使用bulk函数批量导入。
2、AttributeError: 'ES' object has no attribute 'transport' 因为原来使用的是pyes模块 现在换成了elasticsearch,所以改成对应模块
conn2 = Elasticsearch("127.0.0.1:9200")
其他常见错误
SerializationError:JSON数据序列化出错,通常是因为不支持某个节点值的数据类型
RequestError:提交数据格式不正确
ConflictError:索引ID冲突
TransportError:连接无法建立
最后通过了解其实是找到了数据同步的插件 logstash-input-jdbc 能够实现数据的同步增删改查,按照网上的教程能够很轻松的实现,遇到的问题就是插件同步过去的字段都必须是小写。
------------
Python中cx_Oracle的一些函数:
commit() 提交
rollback() 回滚
rollback() 回滚
cursor用来执行命令的方法:
callproc(self, procname, args):用来执行存储过程,接收的参数为存储过程名和参数列表,返回值为受影响的行数
execute(self, query, args):执行单条sql语句,接收的参数为sql语句本身和使用的参数列表,返回值为受影响的行数
executemany(self, query, args):执行单挑sql语句,但是重复执行参数列表里的参数,返回值为受影响的行数
nextset(self):移动到下一个结果集
callproc(self, procname, args):用来执行存储过程,接收的参数为存储过程名和参数列表,返回值为受影响的行数
execute(self, query, args):执行单条sql语句,接收的参数为sql语句本身和使用的参数列表,返回值为受影响的行数
executemany(self, query, args):执行单挑sql语句,但是重复执行参数列表里的参数,返回值为受影响的行数
nextset(self):移动到下一个结果集
cursor用来接收返回值的方法:
fetchall(self):接收全部的返回结果行.
fetchmany(self, size=None):接收size条返回结果行.如果size的值大于返回的结果行的数量,则会返回cursor.arraysize条数据.
fetchone(self):返回一条结果行.
scroll(self, value, mode='relative'):移动指针到某一行.如果mode='relative',则表示从当前所在行移动value条,如果 mode='absolute',则表示从结果集的第一行移动value条.
MySQL中关于中文编码的问题
conn = MySQLdb.Connect(host='localhost', user='root', passwd='root', db='python') 中加一个属性:
conn = MySQLdb.Connect(host='localhost', user='root', passwd='root', db='python',charset='utf8')
charset是要跟你数据库的编码一样,如果是数据库是gb2312 ,则写charset='gb2312'。