本博文介紹三種方法，推薦使用第三種，前兩種都是嘗試。
方法一：
現在的導出還是有缺陷的，導出的文件中還是存在其他不必要的信息
#!/bin/bash ##獲取數據庫 databases=$(hive -e "show databases; exit;") for database in $databases; do #獲取hive建表語句 tables=$(hive -e "use $database; show tables;") for table in $tables; do echo "--=========== db: $database , table: $table ===========" >> $database.sql echo "$(hive -e "use $database;show create table $table;");" >> $database.sql done done
還沒有找到其他方法。有其他解決方法，可以Mark一下我
方法二：
20191108今天有想出來一個方法，有點繁雜，但也是可以的，用impala-shell
1.先准備一個文件（tables_name.txt），我們會讀這個文件
[root@bigdata zw]# more tables_name.txt show create table cdata.c01_bill_distr_stat show create table cdata.c01_bill_distr_stat_temp1 show create table cdata.c01_bill_pro_bal show create table cdata.c01_bill_repay_stat show create table cdata.c01_bill_repay_stat_temp1
2.一個小腳本
#!/usr/bin/python # -*- coding:utf-8 -*- import time,sys import os reload(sys) sys.setdefaultencoding("utf8") file=open("tables_name.txt") send_file = file.readlines() for i in send_file: os_cmd1 = "impala-shell -q 'use cdata' " os_cmd2 = "impala-shell -q '"+ i.strip('\n') +"'" os.system(os_cmd2) file.close()
都放在一個目錄下，運行python腳本，這時候，日志會打印到屏幕上，需要獲取屏幕上的日志內容即可。
我用的xshell工具
這個時候，所有的日志都會打印到文件中（bigdata_2019-11-08_17-20-11），可以找到自己想要的內容。
方法三：
#!/usr/bin/env python #-*- coding:utf8 -*- # 從mysql中提取hive建表語句 import os,sys import fileinput import datetime import mysql.connector reload(sys) sys.setdefaultencoding("utf8") def hive_create_table(): conn = mysql.connector.connect(host="192.168.xxx.xxx",user='hive',passwd='123456',database='hive',charset='utf8') mycursor = conn.cursor() # 獲取DB_ID select_DB_ID = "select DB_ID from DBS;" mycursor.execute(select_DB_ID) result_DB_ID = mycursor.fetchall() fo = open("create_tab.sql", "w") for dir_DB_ID in result_DB_ID : # 獲取數據庫名 DB_ID = str(dir_DB_ID)[1:].split(',')[0] print(DB_ID) select_DB_NAME = "select NAME from DBS where DB_ID="+DB_ID+";" print(select_DB_NAME ) mycursor.execute(select_DB_NAME) result_DB_NAME = mycursor.fetchone() fo.write("\n===========數據庫:"+str(result_DB_NAME).split('\'')[1]+"===========\n") DBname=str(result_DB_NAME).split('\'')[1] print '數據庫名字：' + DBname print(result_DB_NAME) # 獲取表名 select_table_name_sql = "select TBL_NAME from TBLS where DB_ID="+DB_ID+";" mycursor.execute(select_table_name_sql) result_table_names = mycursor.fetchall() for table_name in result_table_names : fo.write("\nCREATE TABLE "+DBname +'.`'+str(table_name).split('\'')[1]+"`(\n") # 根據表名獲取SD_ID select_table_SD_ID = "select SD_ID from TBLS where tbl_name='"+str(table_name).split('\'')[1]+"' and DB_ID="+DB_ID+";" print(select_table_SD_ID) mycursor.execute(select_table_SD_ID) result_SD_ID = mycursor.fetchone() print(result_SD_ID ) # 根據SD_ID獲取CD_ID SD_ID=str(result_SD_ID)[1:].split(',')[0] select_table_CD_ID = "select CD_ID from SDS where SD_ID="+str(result_SD_ID)[1:].split(',')[0]+";" print(select_table_CD_ID) mycursor.execute(select_table_CD_ID) result_CD_ID = mycursor.fetchone() print(result_CD_ID) # 根據CD_ID獲取表的列 CD_ID=str(result_CD_ID)[1:].split(',')[0] select_table_COLUMN_NAME = "select COLUMN_NAME,TYPE_NAME,COMMENT from COLUMNS_V2 where CD_ID="+str(result_CD_ID)[1:].split(',')[0]+" order by INTEGER_IDX;" print(select_table_COLUMN_NAME) mycursor.execute(select_table_COLUMN_NAME) result_COLUMN_NAME = mycursor.fetchall() print(result_COLUMN_NAME) index=0 for col,col_type,col_name in result_COLUMN_NAME: print(col) print(col_type) print(col_name) print(len(result_COLUMN_NAME) ) # 寫入表的列和列的類型到文件 if col_name is None: fo.write(" `"+str(col)+"` "+str(col_type)) else: fo.write(" `"+str(col)+"` "+str(col_type) + " COMMENT '" + str(col_name) + "'") if index < len(result_COLUMN_NAME)-1: index = index + 1 fo.write(",\n") elif index == len(result_COLUMN_NAME)-1: fo.write("\n)") # 根據表名獲取TBL_ID select_table_SD_ID = "select TBL_ID from TBLS where tbl_name='"+str(table_name).split('\'')[1]+"' and DB_ID="+DB_ID+";" print(select_table_SD_ID) mycursor.execute(select_table_SD_ID) result_TBL_ID = mycursor.fetchone() print(result_TBL_ID) # 根據TBL_ID獲取分區信息 select_table_PKEY_NAME_TYPE = "select PKEY_NAME,PKEY_TYPE,PKEY_COMMENT from PARTITION_KEYS where TBL_ID="+str(result_TBL_ID)[1:].split(',')[0]+" order by INTEGER_IDX;" print(select_table_PKEY_NAME_TYPE) mycursor.execute(select_table_PKEY_NAME_TYPE) result_PKEY_NAME_TYPE = mycursor.fetchall() print(result_PKEY_NAME_TYPE) if len(result_PKEY_NAME_TYPE) > 0: fo.write("\nPARTITIONED BY (\n") else : fo.write("\n") i=0 for pkey_name,pkey_type,PKEY_COMMENT in result_PKEY_NAME_TYPE: if str(PKEY_COMMENT) is None: fo.write(" `"+str(pkey_name)+"` "+str(pkey_type)) else: fo.write(" `"+str(pkey_name)+"` "+str(pkey_type) + " COMMENT '" + str(PKEY_COMMENT) + "'\n") if i < len(result_PKEY_NAME_TYPE)- 1: i = i + 1 fo.write(",") elif i == len(result_PKEY_NAME_TYPE) - 1: fo.write(")\n") # 根據表TBL_ID 獲得中文名稱 select_PARAM_VALUE01 = "select PARAM_VALUE from TABLE_PARAMS WHERE TBL_ID=( select TBL_ID from TBLS where tbl_name='"+str(table_name).split('\'')[1]+"' and DB_ID="+DB_ID+") and PARAM_KEY='comment';" print(select_PARAM_VALUE01) mycursor.execute(select_PARAM_VALUE01) result_PARAM_VALUE01 = mycursor.fetchone() print result_PARAM_VALUE01 if result_PARAM_VALUE01 is None: print '未設置表名' elif not result_PARAM_VALUE01[0]: print '表名為空' else: fo.write("COMMENT '" + str(result_PARAM_VALUE01[0]) +"' \n" ) # 根據SD_ID和CD_ID獲取SERDE_ID select_SERDE_ID = "select SERDE_ID from SDS where SD_ID="+SD_ID+" and CD_ID="+CD_ID+";" print(select_SERDE_ID) mycursor.execute(select_SERDE_ID) result_SERDE_ID = mycursor.fetchone() print(result_SERDE_ID) # 根據SERDE_ID獲取PARAM_VALUE(列分隔符) select_PARAM_VALUE = "select PARAM_VALUE from SERDE_PARAMS where SERDE_ID="+str(result_SERDE_ID)[1:].split(",")[0]+" and PARAM_KEY='field.delim';" print(select_PARAM_VALUE) mycursor.execute(select_PARAM_VALUE) result_PARAM_VALUE = mycursor.fetchone() print(result_PARAM_VALUE) if result_PARAM_VALUE is not None: fo.write("ROW FORMAT DELIMITED\n") fo.write("FIELDS TERMINATED BY '"+str(result_PARAM_VALUE).split('\'')[1]+"'\n") # 根據SERDE_ID獲取PARAM_VALUE(行分隔符) select_PARAM_HNAG = "select PARAM_VALUE from SERDE_PARAMS where SERDE_ID="+str(result_SERDE_ID)[1:].split(",")[0]+" and PARAM_KEY='line.delim';" print(select_PARAM_HNAG) mycursor.execute(select_PARAM_HNAG) RESULT_PARAM_HNAG = mycursor.fetchone() print(RESULT_PARAM_HNAG) if RESULT_PARAM_HNAG is not None: fo.write("LINES TERMINATED BY '"+str(RESULT_PARAM_HNAG).split('\'')[1]+"'\n") # 根據SD_ID和CD_ID獲取輸入輸出格式 select_table_STORE_FORMAT = "select INPUT_FORMAT from SDS where SD_ID="+SD_ID+" and CD_ID="+CD_ID+";" print(select_table_STORE_FORMAT) mycursor.execute(select_table_STORE_FORMAT) result_table_STORE_FORMAT= mycursor.fetchall() print(result_table_STORE_FORMAT) for store_format in result_table_STORE_FORMAT: if "org.apache.hadoop.hive.ql.io.orc.OrcInputFormat" in str(store_format): fo.write("STORED AS ORC;\n") elif "org.apache.hadoop.mapred.TextInputFormat" in str(store_format): fo.write("STORED AS TEXTFILE;\n") elif "org.apache.hadoop.hive.ql.io.parquet.MapredParquetInputFormat" in str(store_format): fo.write("STORED AS PARQUET;\n") elif "org.apache.kudu.mapreduce.KuduTableInputFormat" in str(store_format): fo.write("STORED AS KuduTable;\n") else : fo.write("STORED AS null;\n") fo.close() hive_create_table()
直接生成建表腳本的SQL文件。可以直接運行建表