前言
最近工作需要从Oracle数据库SQL、存储过程迁移到hive上,两个之间还是有些差异的,特写此文,共大家参考学习。有问题欢迎指正。
1、oracle中的(+)写法
1.1、区分左右连接
左连接:
select a.*,b.*
from table1 a, table2 b
where a.id = b.id(+)
右连接:
select a.*,b.*
from table1 a, table2 b
where a.id(+) = b.id
1.2、hive中的写法
以左连接进行讲解,右连接同理
建表和插入数据
CREATE TABLE Test_Departments
(
depID NUMBER(38,0),
depName VARCHAR2(20),
delFlag NUMBER(1,0)
);
--生成员工表
CREATE TABLE Test_Employees
(
empID NUMBER(38,0),
empName VARCHAR2(20),
depID NUMBER(38,0),
delFlag NUMBER(1,0)
);
--插入测试数据
INSERT INTO Test_Departments VALUES(1,'FI',0);
INSERT INTO Test_Departments VALUES(2,'MA',0);
INSERT INTO Test_Departments VALUES(3,'HR',1);
INSERT INTO Test_Departments VALUES(4,'IT',0);
INSERT INTO Test_Employees VALUES(1,'wbq',1,0);
INSERT INTO Test_Employees VALUES(2,'czh',2,0);
INSERT INTO Test_Employees VALUES(3,'chh',1,0);
INSERT INTO Test_Employees VALUES(4,'wal',2,0);
INSERT INTO Test_Employees VALUES(5,'ddd',3,0);
a、最常用
--Oracle中的写法
SELECT A.*, B.*
FROM Test_Departments A, Test_Employees B
WHERE A.depID = B.depID(+);
--Hive中的写法
SELECT A.*, B.*
FROM Test_Departments A
LEFT JOIN Test_Employees B
ON A.depID = B.depID;
b、副表带条件
--Oracle中的写法:
SELECT A.*, B.*
FROM Test_Departments A, Test_Employees B
WHERE A.depID = B.depID(+)
AND (B.depID(+) = 3);
--Hive中的写法:
SELECT A.*, B.*
FROM Test_Departments A
LEFT JOIN Test_Employees B ON A.depID = B.depID
AND B.depID = 3;
这里涉及到一个 left join 后面的过滤条件是on和where 的问题
select A.,B. from A left join B on A.id=B.id and B.id=3
返回结果:显示A表所有数据,B.id=3的数据,B.id不等于3的默认为空
select A.,B. from A,B where A.id=B.id(+) and B.id=3
返回结果:仅仅显示B.id=3的一条数据
c、只显示过滤条件的数据
--Oracle
SELECT A.*, B.*
FROM Test_Departments A, Test_Employees B
WHERE A.depID = B.depID(+)
AND (B.depID = 3);
--Hive
SELECT A.*, B.*
FROM Test_Departments A
LEFT JOIN Test_Employees B ON A.depID = B.depID
WHERE B.depID = 3;
这种方式也可以选择使用join实现
以上这段参考:
oracle (+)学习-CSDN博客
2、select中含有子查询
--Oracle:
select a.id, (select b.id from b where b.name=a.id) from a
--hive 是不支持select 里面子查询 修改如下:
select a.id ,b.id from a left join b on a.id=b.name
3、oracle的decode函数
--Oracle:
decode('key',if1,then1 ,if2,then2...thenN)
--Hive:
--1、一般来改为:
case when key = if1 then then1
when key = if2 then then2
...
else thenN end
--2、如果decode比较简单 可以直接改为 :
if('key'=if1,then1,then2)
--复杂的改为:case when
注意hive有个decode函数是编码函数,不是用来处理null值的
4、oracle的时间转化
某字符串yyyyMM获取上个月时间
--oracle:
select to_char(add_months(to_date('202202','yyyymm'),-1),'yyyymm')
from dual;
SELECT 'yyyyMM', to_char (SYSDATE,'yyyyMM') FROM dual UNION ALL
SELECT 'yyyy-MM', to_char (SYSDATE,'yyyy-MM') FROM dual UNION ALL
SELECT 'yyyy-MM-dd', to_char (SYSDATE,'yyyy-MM-dd') FROM dual UNION ALL
SELECT 'yyyy/MM/dd', to_char (SYSDATE,'yyyy/MM/dd') FROM dual;
TO_CHAR(TO_DATE(REPLACE(ADJ.VAR1, '-',''), 'YYYYMMDD'), 'yyyy')
--hive:
select DATE_FORMAT(current_timestamp,'yyyy-MM-dd'); --转换为字符串格式
select to_date(current_timestamp); --日期格式
select 'yyyyMM', DATE_FORMAT(current_timestamp,'yyyyMM') union all
select 'yyyy-MM', DATE_FORMAT(current_timestamp,'yyyy-MM') union all
select 'yyyy-MM-dd', DATE_FORMAT(current_timestamp,'yyyy-MM-dd') union all
select 'yyyy/MM/dd', DATE_FORMAT(current_timestamp, 'yyyy/MM/dd');
with tmp as (
select '2023-12-23' VAR1
)
select from_unixtime(unix_timestamp(replace(ADJ.VAR1, '-',''),'yyyyMMdd'),'yyyy')
from tmp adj;
5、oracle的trunc函数
oracle的trunc函数改为hive的函数_hive对应oracle的trunc函数-CSDN博客
--hive
select TRUNC(current_date,'YYYY') year,
TRUNC(current_date,'MM') month,
last_day(current_date);
--季初:
select 'floor_quarter',date_format(floor_quarter(timestamp('2024-09-23')),'yyyy-MM-dd HH:mm:ss');
--当前时间属于哪个季度:
select quarter(current_date);
6、oracle instr函数
Oracle中的instr()函数 详解及应用_oracle instr-CSDN博客
--oracle
SELECT instr('1234567890123456789','3') FROM dual -- 3
SELECT instr('1234567890123456789','3',1) FROM dual -- 3 ,从第1位开始查找第一个3
SELECT instr('1234567890123456789','3',1,2) FROM dual --13 从第1位开始查找第二个3
SELECT instr('1234567890123456789','3',4) FROM dual -- 13 从第4位开始查找第一个3
SELECT instr('1234567890123456789','3',4,1) FROM dual --13 从第4位开始查找第一个3
SELECT instr('1234567890123456789','3',4,2) FROM dual --0 从第4位开始查找第二个3
--select instr('被查找的字符串','我们需要查找的字符',从第几位开始 首位是0,查找第几个出现的)
--hive
--instr(str, substr) Returns the index of the first occurance of substr in str
SELECT instr('1234567890123456789','3'); -- 3
--locate函数
select locate('3','12345123',4) --8
select locate('3','12345123',1) --3
-- 这个locate函数也是找到字符串的下标 locate('要找的字符','被找的字符串',' 从下标多少开始找')。
7、截取
字段格式是: xx1.xxxxx2.xx3.xxx4.xx5.xx6 我们需要xx3格式的数据
缺省.受限制现金-人民币-风险准备金专户.工行高新支行321413RMB(财付通专用).缺省.缺省.缺省.缺省.缺省 -> 工行高新支行321413RMB(财付通专用)
--Oracle:
SELECT
SEGMENT_NAME_MERGE,
SUBSTR(T.SEGMENT_NAME_MERGE,
INSTR(T.SEGMENT_NAME_MERGE, '.', 1, 2) + 1,
(INSTR(T.SEGMENT_NAME_MERGE, '.', 1, 3) - INSTR(T.SEGMENT_NAME_MERGE, '.', 1, 2)) - 1
)
FROM ODSERPDATA.ODS_CE_GL_ACCOUNT_Q T
--hive
with tmp as(
select '缺省.受限制现金-人民币-风险准备金专户.工行高新支行321413RMB(财付通专用).缺省.缺省.缺省.缺省.缺省' text
)
select SUBSTRING_INDEX(substring_index(a.text,'.',3),'.',-1)
from tmp a;
select replace(substring_index(a,'.',3),substring_index(a,'.',2)||'.',''),
substr(a,length(substring_index(a,'.',2))+2,length(substring_index(a,'.',3))-length(substring_index(a,'.',2))-1),
regexp_extract(a,'.*?\\..*?\\.(.*?)\\.+',1)
from (select '缺省.受限制现金-人民币-风险准备金专户.工行高新支行321413RMB(财付通专用).缺省.缺省.缺省.缺省.缺省' a )t
--上面是三种办法。 1.是替换 2.是截取 3是正则
7、临时表名
--Oracle:
SELECT * FROM (SELECT 1,2 FROM dual );
--hive:必须临时表名
select * from (select 1, 2 )t; --正确
select * from (select 1, 2 ); --错误
8、with插入用法
--Oracle:
INSERT INTO TEST.CC_STUDENT_02
WITH tmp AS (SELECT * FROM TEST.CC_STUDENT_02 cs )
SELECT * FROM tmp;
--Hive
WITH tmp AS (SELECT * FROM TEST.CC_STUDENT_02 cs )
INSERT INTO TEST.CC_STUDENT_02
SELECT * FROM tmp;
9、计算语法 或者||用法不同
--Oracle:
SELECT substr('202212', 1, 4) - 1 || 'aa' FROM dual -- 2021aa
SELECT 1||NULL||2 FROM dual -- 12
--Hive
SELECT substr('202212', 1, 4) - 1 || 'aa' -- 2021.0aa
--解决办法:
SELECT cast(substr('202212', 1, 4) - 1 as int)|| 'aa' -- 2021aa
SELECT cast(substr('202212', 1, 4) as int) - 1|| 'aa' -- 2021aa
--因为int-int=int 。 string-int 和int-string=double
10、日期格式不标准的转换
--Oracle
SELECT to_date('2017-3-31', 'yyyy/mm/dd') FROM dual UNION ALL
SELECT to_date('2018/11/6', 'yyyy/mm/dd') FROM dual UNION ALL
SELECT to_date('2017/6/20', 'yyyy/mm/dd') FROM dual UNION ALL
SELECT to_date('2017-06-20', 'yyyy/mm/dd') FROM dual UNION ALL
SELECT to_date('20170620', 'yyyy/mm/dd') FROM dual
--Hive
with tmp as
(SELECT '2017-3-31' var UNION ALL
SELECT '2018/11/6' UNION ALL
SELECT '20170620'
)
select case when length(a.var) = 8 then date(from_unixtime(unix_timestamp(a.var,'yyyyMMdd'),'yyyy-MM-dd'))
else date(replace(a.var,'/','-')) end
from tmp a;
11、WM_CONCAT
--Oracle:
SELECT t.id ,to_char(WM_CONCAT(name)),WM_CONCAT(name)
FROM (
SELECT 1 AS id ,1 as name FROM dual UNION ALL
SELECT 1 AS id ,2 as name FROM dual UNION ALL
SELECT 1 AS id ,1 as name FROM dual UNION ALL
SELECT 2 AS id ,4 as name FROM dual UNION ALL
SELECT 2 AS id ,5 as name FROM dual
)t
GROUP BY t.id;
--Hive
with tmp as (
SELECT 1 AS id ,'1' as name UNION ALL
SELECT 1 AS id ,'2' as name UNION ALL
SELECT 1 AS id ,'1' as name UNION ALL
SELECT 2 AS id ,'4' as name UNION ALL
SELECT 2 AS id ,'5' as name
)
select id
,concat_ws(',',collect_list(name)) --未去重
,concat_ws(',',collect_set(name)) --去重
from tmp a
group by id;
12、日期格式化
--Oracle:
SELECT to_date('2023-01-02 15:55:03', 'yyyy-mm-dd hh24:mi:ss')
,TO_DATE('2023-01-02 15:55:03', 'YYYY-MM-DD HH24:MI:SS')FROM dual
--Hive
select DATE_FORMAT('2023-01-02 15:55:03','yyyy-MM-dd HH:mm:ss')
13、类型兼容问题
--Oracle:
with t as (
select '1' a union all
select '11'a union all
select '2' a union all
select '3' a
)
select t.a from t order by a;
--Hive:
with t as (
select '1' a union all
select '11'a union all
select '2' a union all
select '3' a
)
select t.a from t order by cast(a as int );
14、Oracle中pivot/和unpivot函数
oracle 学习之 unpivot/pivot函数及hive实现该功能_cclovezbf的博客-CSDN博客
15、RPAD LPAD (左右填充函数)
--Oracle:
SELECT RPAD ('1234',10),LENGTH(RPAD ('1234',10)) FROM dual
-- 1234 10 --注意 1234后面又6个空格
SELECT RPAD ('1234',10,'a'),LENGTH(RPAD ('1234',10,'a')) FROM dual
--1234aaaaaa 10
--hive:
SELECT RPAD ('1234',10,' '),LENGTH(RPAD ('1234',10,' ')),RPAD ('1234',10,'a'),lPAD ('1234',10,'a')
16、时间相减
--Oracle
SELECT SYSDATE -1 FROM dual --获取昨天
SELECT to_date('2023-05-03','YYYY-MM-DD') -to_date('2023-05-02','YYYY-MM-DD') FROM dual --1 两天时间差
SELECT SYSDATE -to_date('2023-05-02','YYYY-MM-DD') FROM dual
--9.64637731481481481481481481481481481481 时间差 计算了时分秒
---hive
select `current_timestamp`(), `current_timestamp`() -to_date('2023-05-10')
-- 2023-05-11 15:33:00.779000000,1 15:33:00.779000000 但是不推荐 因为看起来既直观又不直观
--计算时间差一般采用的是datediff函数
select datediff(current_timestamp(),to_date('2023-05-10')) -- 1
计--算前几天后几天采用date_add date_sub
select current_date,date_add(current_date,1),date_sub(current_date(),1)
---相差的秒数
select (unix_timestamp('2024-10-23 14:12:34') - unix_timestamp('2024-10-23 12:12:34'))*1.00000
---相差的月数
SELECT months_between('2022-12-01', '2022-01-01') AS months_difference
---相差小时
select (unix_timestamp('2024-10-23 14:13:34') - unix_timestamp('2024-10-23 12:12:34'))*1.00000 / 3600
---相差天数
select (unix_timestamp('2024-10-24 14:13:34') - unix_timestamp('2024-10-23 12:12:34'))*1.00000 / 86400
以上参考文章:oracle和hive之间关于sql的语法差异及转换_hivesql 与oracle 语句区别-CSDN博客
17、INSERT语句
------ORACLE:
EXECUTE IMMEDIATE 'TRUNCATE TABLE TABLE_A';
INSERT /*+APPEND*/ INTO A NOLOGGING (
COL_1,
COL_2,
COL_3,
);
------Hive:
INSERT OVERWRITE TABLE TABLE_A;
18、HIVE特性
在迁移的时候,会遇到一些HIVE的特性导致无法顺利迁移,如下ORACLE语句:
SELECT *
FROM A
GROUP BY A.COL_1
HAVING COUNT(DISTINCT A.COL_2) = 2;
在HIVE中运行会遇到如下报错:
FAILED: SemanticException [Error 10002]: Line 19:22 Invalid column reference 'COL_2'
去掉DISTINCT后可以正常执行。
属于HIVE的特性导致的问题,HAVING子句中无法使用DISTINCT关键字,在迁移中需要进行取舍。
版权归原作者 三生暮雨渡瀟瀟 所有, 如有侵权,请联系我们删除。