根據mysql元數據還原hive建表語句

概述

• 公司集羣崩了,hive訪問不上,想要建表語句
• 功能是一次性的(不要有下次了),就用python寫了

適用場景

• Hive版本:Hive release version 1.1.0 其餘自行測試
• CDH版本:1.1.0-cdh5.12.0 其餘自行測試
• 通常的hive表,不帶分桶的,分桶表的沒作
• 壓縮的也沒作,因此有壓縮表的建表語句也不完整

前期準備

• 要對mysql裏的hive數據庫結構有必定的瞭解,請訪問如下網址:Hive的元數據表結構詳解
• mysql的連接信息根據本身的配置更改

代碼

#!/usr/bin/env python
#-*- coding:utf8 -*-
# 從mysql中提取hive建表語句
import os
import fileinput
import datetime
import mysql.connector

def hive_create_table():
    #conn = mysql.connector.connect(host="11.11.11.11",user='root',passwd='root',database='hive')
    conn = mysql.connector.connect(host="localhost",user='root',database='hive')
    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("foo.txt", "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")
      	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 `"+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 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 in result_COLUMN_NAME:
                print(col)
                print(col_type)
         	    	print(len(result_COLUMN_NAME) )
            		# 寫入表的列和列的類型到文件
                fo.write("  `"+str(col)+"`  "+str(col_type))
                if index < len(result_COLUMN_NAME)-1:
	                 index = index + 1
                   fo.write(",\n")
            		elif index == len(result_COLUMN_NAME)-1:
                   fo.write(")")
            # 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 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:
            #if result_PKEY_NAME_TYPE is not None:
               fo.write("\nPARTITIONED BY (\n")
            #elif len(result_PKEY_NAME_TYPE) == 0:
            else :
               fo.write("\n")
            i=0
            for pkey_name,pkey_type in result_PKEY_NAME_TYPE:
                fo.write("  `"+str(pkey_name)+"`  "+str(pkey_type))
                if i < len(result_PKEY_NAME_TYPE)- 1:
	                 i = i + 1
                   fo.write(",\n")
            		elif i == len(result_PKEY_NAME_TYPE) - 1:
                   fo.write(")\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 "Orc" in str(store_format):
                   fo.write("STORED AS ORC;\n")
                elif "RCFile" in str(store_format):
                   fo.write("STORED AS RCFILE;\n")
                else :
                   fo.write("STORED AS TEXTFILE;\n")
                   #fo.write(";\n")

    fo.close()
hive_create_table()