Hive SQL案例分析
数据筹备
7369,SMITH,CLERK,7902,1980-12-17 00:00:00,800,\N,20
7499,ALLEN,SALESMAN,7698,1981-02-20 00:00:00,1600,300,30
7521,WARD,SALESMAN,7698,1981-02-22 00:00:00,1250,500,30
7566,JONES,MANAGER,7839,1981-04-02 00:00:00,2975,\N,20
7654,MARTIN,SALESMAN,7698,1981-09-28 00:00:00,1250,1400,30
7698,BLAKE,MANAGER,7839,1981-05-01 00:00:00,2850,\N,30
7782,CLARK,MANAGER,7839,1981-06-09 00:00:00,2450,\N,10
7788,SCOTT,ANALYST,7566,1987-04-19 00:00:00,1500,\N,20
7839,KING,PRESIDENT,\N,1981-11-17 00:00:00,5000,\N,10
7844,TURNER,SALESMAN,7698,1981-09-08 00:00:00,1500,0,30
7876,ADAMS,CLERK,7788,1987-05-23 00:00:00,1100,\N,20
7900,JAMES,CLERK,7698,1981-12-03 00:00:00,950,\N,30
7902,FORD,ANALYST,7566,1981-12-03 00:00:00,3000,\N,20
7934,MILLER,CLERK,7782,1982-01-23 00:00:00,1300,\N,10
CREATE TABLE t_employee(
empno INT,
ename STRING,
job STRING,
mgr INT,
hiredate TIMESTAMP,
sal DECIMAL(7,2),
comm DECIMAL(7,2),
deptno INT)
row format delimited
fields terminated by ','
collection items terminated by '|'
map keys terminated by '>'
lines terminated by '\n'
stored as textfile;
10,ACCOUNTING,NEW YORK
20,RESEARCH,DALLAS
30,SALES,CHICAGO
40,OPERATIONS,BOSTON
CREATE TABLE t_dept(
DEPTNO INT,
DNAME STRING,
LOC STRING)
row format delimited
fields terminated by ','
collection items terminated by '|'
map keys terminated by '>'
lines terminated by '\n'
stored as textfile;
0: jdbc:hive2://CentOS:10000> select empno,ename,job,mgr,hiredate,sal,comm,deptno from t_employee;
+--------+---------+------------+-------+------------------------+-------+-------+---------+--+
| empno | ename | job | mgr | hiredate | sal | comm | deptno |
+--------+---------+------------+-------+------------------------+-------+-------+---------+--+
| 7369 | SMITH | CLERK | 7902 | 1980-12-17 00:00:00.0 | 800 | NULL | 20 |
| 7499 | ALLEN | SALESMAN | 7698 | 1981-02-20 00:00:00.0 | 1600 | 300 | 30 |
| 7521 | WARD | SALESMAN | 7698 | 1981-02-22 00:00:00.0 | 1250 | 500 | 30 |
| 7566 | JONES | MANAGER | 7839 | 1981-04-02 00:00:00.0 | 2975 | NULL | 20 |
| 7654 | MARTIN | SALESMAN | 7698 | 1981-09-28 00:00:00.0 | 1250 | 1400 | 30 |
| 7698 | BLAKE | MANAGER | 7839 | 1981-05-01 00:00:00.0 | 2850 | NULL | 30 |
| 7782 | CLARK | MANAGER | 7839 | 1981-06-09 00:00:00.0 | 2450 | NULL | 10 |
| 7788 | SCOTT | ANALYST | 7566 | 1987-04-19 00:00:00.0 | 1500 | NULL | 20 |
| 7839 | KING | PRESIDENT | NULL | 1981-11-17 00:00:00.0 | 5000 | NULL | 10 |
| 7844 | TURNER | SALESMAN | 7698 | 1981-09-08 00:00:00.0 | 1500 | 0 | 30 |
| 7876 | ADAMS | CLERK | 7788 | 1987-05-23 00:00:00.0 | 1100 | NULL | 20 |
| 7900 | JAMES | CLERK | 7698 | 1981-12-03 00:00:00.0 | 950 | NULL | 30 |
| 7902 | FORD | ANALYST | 7566 | 1981-12-03 00:00:00.0 | 3000 | NULL | 20 |
| 7934 | MILLER | CLERK | 7782 | 1982-01-23 00:00:00.0 | 1300 | NULL | 10 |
+--------+---------+------------+-------+------------------------+-------+-------+---------+--+
14 rows selected (0.047 seconds)
0: jdbc:hive2://CentOS:10000> select deptno,dname,loc from t_dept;
+---------+-------------+-----------+--+
| deptno | dname | loc |
+---------+-------------+-----------+--+
| 10 | ACCOUNTING | NEW YORK |
| 20 | RESEARCH | DALLAS |
| 30 | SALES | CHICAGO |
| 40 | OPERATIONS | BOSTON |
+---------+-------------+-----------+--+
4 rows selected (0.046 seconds)
CREATE TABLE t_employee_partition(
empno INT,
ename STRING,
job STRING,
mgr INT,
hiredate TIMESTAMP,
sal DECIMAL(7,2),
comm DECIMAL(7,2))
PARTITIONED BY(deptno INT)
row format delimited
fields terminated by ','
collection items terminated by '|'
map keys terminated by '>'
lines terminated by '\n'
stored as textfile;
0: jdbc:hive2://CentOS:10000> set hive.exec.dynamic.partition.mode=nonstrict
0: jdbc:hive2://CentOS:10000> INSERT OVERWRITE TABLE t_employee_partition PARTITION (deptno) SELECT empno,ename,job,mgr,hiredate,sal,comm,deptno FROM t_employee;
SQL查询
单表查询
0: jdbc:hive2://CentOS:10000> select empno,ename,job,mgr,hiredate,sal,comm,deptno from t_employee;
+--------+---------+------------+-------+------------------------+-------+-------+---------+--+
| empno | ename | job | mgr | hiredate | sal | comm | deptno |
+--------+---------+------------+-------+------------------------+-------+-------+---------+--+
| 7369 | SMITH | CLERK | 7902 | 1980-12-17 00:00:00.0 | 800 | NULL | 20 |
| 7499 | ALLEN | SALESMAN | 7698 | 1981-02-20 00:00:00.0 | 1600 | 300 | 30 |
| 7521 | WARD | SALESMAN | 7698 | 1981-02-22 00:00:00.0 | 1250 | 500 | 30 |
| 7566 | JONES | MANAGER | 7839 | 1981-04-02 00:00:00.0 | 2975 | NULL | 20 |
| 7654 | MARTIN | SALESMAN | 7698 | 1981-09-28 00:00:00.0 | 1250 | 1400 | 30 |
| 7698 | BLAKE | MANAGER | 7839 | 1981-05-01 00:00:00.0 | 2850 | NULL | 30 |
| 7782 | CLARK | MANAGER | 7839 | 1981-06-09 00:00:00.0 | 2450 | NULL | 10 |
| 7788 | SCOTT | ANALYST | 7566 | 1987-04-19 00:00:00.0 | 1500 | NULL | 20 |
| 7839 | KING | PRESIDENT | NULL | 1981-11-17 00:00:00.0 | 5000 | NULL | 10 |
| 7844 | TURNER | SALESMAN | 7698 | 1981-09-08 00:00:00.0 | 1500 | 0 | 30 |
| 7876 | ADAMS | CLERK | 7788 | 1987-05-23 00:00:00.0 | 1100 | NULL | 20 |
| 7900 | JAMES | CLERK | 7698 | 1981-12-03 00:00:00.0 | 950 | NULL | 30 |
| 7902 | FORD | ANALYST | 7566 | 1981-12-03 00:00:00.0 | 3000 | NULL | 20 |
| 7934 | MILLER | CLERK | 7782 | 1982-01-23 00:00:00.0 | 1300 | NULL | 10 |
+--------+---------+------------+-------+------------------------+-------+-------+---------+--+
14 rows selected (0.056 seconds)
WHERE查询
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,job,mgr,hiredate,sal,comm,deptno FROM t_employee WHERE empno > 7782 AND deptno = 10;
+--------+---------+------------+-------+------------------------+-------+-------+---------+--+
| empno | ename | job | mgr | hiredate | sal | comm | deptno |
+--------+---------+------------+-------+------------------------+-------+-------+---------+--+
| 7839 | KING | PRESIDENT | NULL | 1981-11-17 00:00:00.0 | 5000 | NULL | 10 |
| 7934 | MILLER | CLERK | 7782 | 1982-01-23 00:00:00.0 | 1300 | NULL | 10 |
+--------+---------+------------+-------+------------------------+-------+-------+---------+--+
2 rows selected (0.067 seconds)
DISTINCT查询
0: jdbc:hive2://CentOS:10000> select distinct(job) from t_employee;
+------------+--+
| job |
+------------+--+
| ANALYST |
| CLERK |
| MANAGER |
| PRESIDENT |
| SALESMAN |
+------------+--+
分区查询
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,job,mgr,hiredate,sal,comm,deptno FROM t_employee_partition e WHERE e.deptno >= 20 AND e.deptno <= 40;
+--------+---------+-----------+-------+------------------------+-------+-------+---------+--+
| empno | ename | job | mgr | hiredate | sal | comm | deptno |
+--------+---------+-----------+-------+------------------------+-------+-------+---------+--+
| 7369 | SMITH | CLERK | 7902 | 1980-12-17 00:00:00.0 | 800 | NULL | 20 |
| 7566 | JONES | MANAGER | 7839 | 1981-04-02 00:00:00.0 | 2975 | NULL | 20 |
| 7788 | SCOTT | ANALYST | 7566 | 1987-04-19 00:00:00.0 | 1500 | NULL | 20 |
| 7876 | ADAMS | CLERK | 7788 | 1987-05-23 00:00:00.0 | 1100 | NULL | 20 |
| 7902 | FORD | ANALYST | 7566 | 1981-12-03 00:00:00.0 | 3000 | NULL | 20 |
| 7499 | ALLEN | SALESMAN | 7698 | 1981-02-20 00:00:00.0 | 1600 | 300 | 30 |
| 7521 | WARD | SALESMAN | 7698 | 1981-02-22 00:00:00.0 | 1250 | 500 | 30 |
| 7654 | MARTIN | SALESMAN | 7698 | 1981-09-28 00:00:00.0 | 1250 | 1400 | 30 |
| 7698 | BLAKE | MANAGER | 7839 | 1981-05-01 00:00:00.0 | 2850 | NULL | 30 |
| 7844 | TURNER | SALESMAN | 7698 | 1981-09-08 00:00:00.0 | 1500 | 0 | 30 |
| 7900 | JAMES | CLERK | 7698 | 1981-12-03 00:00:00.0 | 950 | NULL | 30 |
+--------+---------+-----------+-------+------------------------+-------+-------+---------+--+
11 rows selected (0.123 seconds)
LIMIT查询
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,job,mgr,hiredate,sal,comm,deptno FROM t_employee ORDER BY sal DESC LIMIT 5;
+--------+--------+------------+-------+------------------------+-------+-------+---------+--+
| empno | ename | job | mgr | hiredate | sal | comm | deptno |
+--------+--------+------------+-------+------------------------+-------+-------+---------+--+
| 7839 | KING | PRESIDENT | NULL | 1981-11-17 00:00:00.0 | 5000 | NULL | 10 |
| 7902 | FORD | ANALYST | 7566 | 1981-12-03 00:00:00.0 | 3000 | NULL | 20 |
| 7566 | JONES | MANAGER | 7839 | 1981-04-02 00:00:00.0 | 2975 | NULL | 20 |
| 7698 | BLAKE | MANAGER | 7839 | 1981-05-01 00:00:00.0 | 2850 | NULL | 30 |
| 7782 | CLARK | MANAGER | 7839 | 1981-06-09 00:00:00.0 | 2450 | NULL | 10 |
+--------+--------+------------+-------+------------------------+-------+-------+---------+--+
5 rows selected (14.294 seconds)
GROUP BY查询
0: jdbc:hive2://CentOS:10000> set hive.map.aggr=true;
0: jdbc:hive2://CentOS:10000> SELECT deptno,SUM(sal) as total FROM t_employee GROUP BY deptno;
+---------+--------+--+
| deptno | total |
+---------+--------+--+
| 10 | 8750 |
| 20 | 9375 |
| 30 | 9400 |
+---------+--------+--+
3 rows selected (12.645 seconds)
hive.map.aggr控制程序如何进行聚合。默认值为false。如果设置为true，Hive会在map阶段就执行一次聚合。这可以提高聚合效率，但需要消耗更多内存。
ORDER AND SORT
可以使用ORDER BY或者Sort BY对查询结果进行排序，排序字段可以是整型也可以是字符串：如果是整型，则按照大小排序；如果是字符串，则按照字典序排序。ORDER BY 和 SORT BY 的区别如下：使用ORDER BY时会有一个Reducer对全部查询结果进行排序，可以保证数据的全局有序性；使用SORT BY时只会在每个Reducer中进行排序，这可以保证每个Reducer的输出数据是有序的，但不能保证全局有序。由于ORDER BY的时间可能很长，如果你设置了严格模式(hive.mapred.mode = strict)，则其后面必须再跟一个limit子句。
- sort by
0: jdbc:hive2://CentOS:10000> set mapreduce.job.reduces=2
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,sal from t_employee sort by sal desc;
+--------+---------+-------+--+
| empno | ename | sal |
+--------+---------+-------+--+
| 7902 | FORD | 3000 |
| 7566 | JONES | 2975 |
| 7844 | TURNER | 1500 |
| 7788 | SCOTT | 1500 |
| 7521 | WARD | 1250 |
| 7654 | MARTIN | 1250 |
| 7876 | ADAMS | 1100 |
| 7900 | JAMES | 950 |
| 7369 | SMITH | 800 |
| 7839 | KING | 5000 |
| 7698 | BLAKE | 2850 |
| 7782 | CLARK | 2450 |
| 7499 | ALLEN | 1600 |
| 7934 | MILLER | 1300 |
+--------+---------+-------+--+
14 rows selected (14.474 seconds)
- order by
0: jdbc:hive2://CentOS:10000> set mapreduce.job.reduces=3;
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,sal from t_employee order by sal desc;
+--------+---------+-------+--+
| empno | ename | sal |
+--------+---------+-------+--+
| 7839 | KING | 5000 |
| 7902 | FORD | 3000 |
| 7566 | JONES | 2975 |
| 7698 | BLAKE | 2850 |
| 7782 | CLARK | 2450 |
| 7499 | ALLEN | 1600 |
| 7844 | TURNER | 1500 |
| 7788 | SCOTT | 1500 |
| 7934 | MILLER | 1300 |
| 7654 | MARTIN | 1250 |
| 7521 | WARD | 1250 |
| 7876 | ADAMS | 1100 |
| 7900 | JAMES | 950 |
| 7369 | SMITH | 800 |
+--------+---------+-------+--+
14 rows selected (13.049 seconds)
0: jdbc:hive2://CentOS:10000> set hive.mapred.mode = strict;
No rows affected (0.004 seconds)
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,sal from t_employee order by sal desc;
Error: Error while compiling statement: FAILED: SemanticException 1:48 In strict mode, if ORDER BY is specified, LIMIT must also be specified. Error encountered near token 'sal' (state=42000,code=40000)
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,sal from t_employee order by sal desc limit 5;
+--------+--------+-------+--+
| empno | ename | sal |
+--------+--------+-------+--+
| 7839 | KING | 5000 |
| 7902 | FORD | 3000 |
| 7566 | JONES | 2975 |
| 7698 | BLAKE | 2850 |
| 7782 | CLARK | 2450 |
+--------+--------+-------+--+
5 rows selected (12.468 seconds)
8、HAVING过滤
0: jdbc:hive2://CentOS:10000> SELECT deptno,SUM(sal) total FROM t_employee GROUP BY deptno HAVING SUM(sal)>9000;
+---------+--------+--+
| deptno | total |
+---------+--------+--+
| 30 | 9400 |
| 20 | 9375 |
+---------+--------+--+
2 rows selected (18.361 seconds)
DISTRIBUTE BY
默认情况下，MapReduce程序会对Map输出结果的Key值进行散列，并均匀分发到所有Reducer上。如果想要把具有相同Key值的数据分发到同一个Reducer进行处理，这就需要使用DISTRIBUTE BY字句。需要注意的是，DISTRIBUTE BY虽然能保证具有相同Key值的数据分发到同一个Reducer，但是不能保证数据在Reducer上是有序的。
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,sal, deptno FROM t_employee distribute BY deptno;
+--------+---------+-------+---------+--+
| empno | ename | sal | deptno |
+--------+---------+-------+---------+--+
| 7654 | MARTIN | 1250 | 30 |
| 7900 | JAMES | 950 | 30 |
| 7698 | BLAKE | 2850 | 30 |
| 7521 | WARD | 1250 | 30 |
| 7844 | TURNER | 1500 | 30 |
| 7499 | ALLEN | 1600 | 30 |
| 7934 | MILLER | 1300 | 10 |
| 7839 | KING | 5000 | 10 |
| 7782 | CLARK | 2450 | 10 |
| 7788 | SCOTT | 1500 | 20 |
| 7566 | JONES | 2975 | 20 |
| 7876 | ADAMS | 1100 | 20 |
| 7902 | FORD | 3000 | 20 |
| 7369 | SMITH | 800 | 20 |
+--------+---------+-------+---------+--+
14 rows selected (15.504 seconds)
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,sal, deptno FROM t_employee distribute BY deptno sort by sal desc;
+--------+---------+-------+---------+--+
| empno | ename | sal | deptno |
+--------+---------+-------+---------+--+
| 7698 | BLAKE | 2850 | 30 |
| 7499 | ALLEN | 1600 | 30 |
| 7844 | TURNER | 1500 | 30 |
| 7521 | WARD | 1250 | 30 |
| 7654 | MARTIN | 1250 | 30 |
| 7900 | JAMES | 950 | 30 |
| 7839 | KING | 5000 | 10 |
| 7782 | CLARK | 2450 | 10 |
| 7934 | MILLER | 1300 | 10 |
| 7902 | FORD | 3000 | 20 |
| 7566 | JONES | 2975 | 20 |
| 7788 | SCOTT | 1500 | 20 |
| 7876 | ADAMS | 1100 | 20 |
| 7369 | SMITH | 800 | 20 |
+--------+---------+-------+---------+--+
14 rows selected (16.528 seconds)
CLUSTER BY
如果SORT BY和DISTRIBUTE BY指定的是相同字段，且SORT BY排序规则是ASC，此时可以使用CLUSTER BY进行替换。
0: jdbc:hive2://CentOS:10000> SELECT empno,ename,sal, deptno FROM t_employee cluster by deptno;
+--------+---------+-------+---------+--+
| empno | ename | sal | deptno |
+--------+---------+-------+---------+--+
| 7934 | MILLER | 1300 | 10 |
| 7839 | KING | 5000 | 10 |
| 7782 | CLARK | 2450 | 10 |
| 7876 | ADAMS | 1100 | 20 |
| 7788 | SCOTT | 1500 | 20 |
| 7369 | SMITH | 800 | 20 |
| 7566 | JONES | 2975 | 20 |
| 7902 | FORD | 3000 | 20 |
| 7844 | TURNER | 1500 | 30 |
| 7499 | ALLEN | 1600 | 30 |
| 7698 | BLAKE | 2850 | 30 |
| 7654 | MARTIN | 1250 | 30 |
| 7521 | WARD | 1250 | 30 |
| 7900 | JAMES | 950 | 30 |
+--------+---------+-------+---------+--+
14 rows selected (25.847 seconds)
表Join查询
Hive支持内连接，外连接，左外连接，右外连接，笛卡尔连接，这和传统数据库中的概念是一致的。需要特别强调：JOIN语句的关联条件必须用ON指定，不能用WHERE指定，否则就会先做笛卡尔积，再过滤，这会导致你得不到预期的结果。
- 内连接
0: jdbc:hive2://CentOS:10000> SELECT e.empno,e.ename,e.sal,d.dname,d.deptno FROM t_employee e JOIN t_dept d ON e.deptno = d.deptno WHERE e.empno=7369;
+----------+----------+--------+-----------+-----------+--+
| e.empno | e.ename | e.sal | d.dname | d.deptno |
+----------+----------+--------+-----------+-----------+--+
| 7369 | SMITH | 800 | RESEARCH | 20 |
+----------+----------+--------+-----------+-----------+--+
1 row selected (10.419 seconds)
- 外连接
0: jdbc:hive2://CentOS:10000> SELECT e.empno,e.ename,e.sal,d.dname,d.deptno FROM t_employee e LEFT OUTER JOIN t_dept d ON e.deptno = d.deptno;
+----------+----------+--------+-------------+-----------+--+
| e.empno | e.ename | e.sal | d.dname | d.deptno |
+----------+----------+--------+-------------+-----------+--+
| 7369 | SMITH | 800 | RESEARCH | 20 |
| 7499 | ALLEN | 1600 | SALES | 30 |
| 7521 | WARD | 1250 | SALES | 30 |
| 7566 | JONES | 2975 | RESEARCH | 20 |
| 7654 | MARTIN | 1250 | SALES | 30 |
| 7698 | BLAKE | 2850 | SALES | 30 |
| 7782 | CLARK | 2450 | ACCOUNTING | 10 |
| 7788 | SCOTT | 1500 | RESEARCH | 20 |
| 7839 | KING | 5000 | ACCOUNTING | 10 |
| 7844 | TURNER | 1500 | SALES | 30 |
| 7876 | ADAMS | 1100 | RESEARCH | 20 |
| 7900 | JAMES | 950 | SALES | 30 |
| 7902 | FORD | 3000 | RESEARCH | 20 |
| 7934 | MILLER | 1300 | ACCOUNTING | 10 |
+----------+----------+--------+-------------+-----------+--+
14 rows selected (11.424 seconds)
0: jdbc:hive2://CentOS:10000> SELECT e.empno,e.ename,e.sal,d.dname,d.deptno FROM t_employee e RIGHT OUTER JOIN t_dept d ON e.deptno = d.deptno;
+----------+----------+--------+-------------+-----------+--+
| e.empno | e.ename | e.sal | d.dname | d.deptno |
+----------+----------+--------+-------------+-----------+--+
| 7782 | CLARK | 2450 | ACCOUNTING | 10 |
| 7839 | KING | 5000 | ACCOUNTING | 10 |
| 7934 | MILLER | 1300 | ACCOUNTING | 10 |
| 7369 | SMITH | 800 | RESEARCH | 20 |
| 7566 | JONES | 2975 | RESEARCH | 20 |
| 7788 | SCOTT | 1500 | RESEARCH | 20 |
| 7876 | ADAMS | 1100 | RESEARCH | 20 |
| 7902 | FORD | 3000 | RESEARCH | 20 |
| 7499 | ALLEN | 1600 | SALES | 30 |
| 7521 | WARD | 1250 | SALES | 30 |
| 7654 | MARTIN | 1250 | SALES | 30 |
| 7698 | BLAKE | 2850 | SALES | 30 |
| 7844 | TURNER | 1500 | SALES | 30 |
| 7900 | JAMES | 950 | SALES | 30 |
| NULL | NULL | NULL | OPERATIONS | 40 |
+----------+----------+--------+-------------+-----------+--+
15 rows selected (11.063 seconds)
0: jdbc:hive2://CentOS:10000> SELECT e.empno,e.ename,e.sal,d.dname,d.deptno FROM t_employee e FULL OUTER JOIN t_dept d ON e.deptno = d.deptno;
+----------+----------+--------+-------------+-----------+--+
| e.empno | e.ename | e.sal | d.dname | d.deptno |
+----------+----------+--------+-------------+-----------+--+
| 7934 | MILLER | 1300 | ACCOUNTING | 10 |
| 7839 | KING | 5000 | ACCOUNTING | 10 |
| 7782 | CLARK | 2450 | ACCOUNTING | 10 |
| 7876 | ADAMS | 1100 | RESEARCH | 20 |
| 7788 | SCOTT | 1500 | RESEARCH | 20 |
| 7369 | SMITH | 800 | RESEARCH | 20 |
| 7566 | JONES | 2975 | RESEARCH | 20 |
| 7902 | FORD | 3000 | RESEARCH | 20 |
| 7844 | TURNER | 1500 | SALES | 30 |
| 7499 | ALLEN | 1600 | SALES | 30 |
| 7698 | BLAKE | 2850 | SALES | 30 |
| 7654 | MARTIN | 1250 | SALES | 30 |
| 7521 | WARD | 1250 | SALES | 30 |
| 7900 | JAMES | 950 | SALES | 30 |
| NULL | NULL | NULL | OPERATIONS | 40 |
+----------+----------+--------+-------------+-----------+--+
15 rows selected (24.703 seconds)
12、LEFT SEMI JOIN
LEFT SEMI JOIN （左半连接）是 IN/EXISTS 子查询的一种更高效的实现。
- JOIN 子句中右边的表只能在 ON 子句中设置过滤条件;
- 查询结果只包含左边表的数据，所以只能SELECT左表中的列。
0: jdbc:hive2://CentOS:10000> SELECT e.empno,e.ename,d.dname FROM t_employee e LEFT SEMI JOIN t_dept d ON e.deptno = d.deptno AND d.loc="NEW YORK";
+----------+----------+-----------+--+
| e.empno | e.ename | e.deptno |
+----------+----------+-----------+--+
| 7782 | CLARK | 10 |
| 7839 | KING | 10 |
| 7934 | MILLER | 10 |
+----------+----------+-----------+--+
3 rows selected (10.119 seconds)
JOIN优化
- STREAMTABLE
在多表进行join的时候，如果每个ON子句都使用到共同的列，此时Hive会进行优化，将多表JOIN在同一个map / reduce作业上进行。同时假定查询的最后一个表是最大的一个表，在对每行记录进行JOIN操作时，它将尝试将其他的表缓存起来，然后扫描最后那个表进行计算。因此用户需要保证查询的表的大小从左到右是依次增加的。
SELECT a.val, b.val, c.val FROM a JOIN b ON (a.key = b.key) JOIN c ON (c.key = b.key)
然而用户并非需要总是把最大的表放在查询语句的最后面，Hive提供了/*+ STREAMTABLE() */标志，使用该标识来指出大表，能避免数据表过大导致占用内存过多而产生的问题。示例如下：
0: jdbc:hive2://CentOS:10000> SELECT /*+ STREAMTABLE(e) */ e.empno,e.ename,d.dname,d.deptno FROM t_employee e JOIN t_dept d ON e.deptno = d.deptno WHERE job='CLERK';
+----------+----------+-------------+-----------+--+
| e.empno | e.ename | d.dname | d.deptno |
+----------+----------+-------------+-----------+--+
| 7369 | SMITH | RESEARCH | 20 |
| 7876 | ADAMS | RESEARCH | 20 |
| 7900 | JAMES | SALES | 30 |
| 7934 | MILLER | ACCOUNTING | 10 |
+----------+----------+-------------+-----------+--+
4 rows selected (11.645 seconds)
- MAPJOIN
如果在进行join操作时，有一个表很小，则可以将join操作调整到map阶段执行。这就是典型的极大表和极小表关联问题。有两种解决方式：1.增加**/*+ MAPJOIN(b) */标示；2.设置参数hive.optimize.bucketmapjoin = true**，在
0: jdbc:hive2://CentOS:10000> SELECT /*+ MAPJOIN(d) */ e.empno, e.ename,d.dname FROM t_employee e JOIN t_dept d ON d.deptno = e.deptno;
+----------+----------+-------------+--+
| e.empno | e.ename | d.dname |
+----------+----------+-------------+--+
| 7369 | SMITH | RESEARCH |
| 7499 | ALLEN | SALES |
| 7521 | WARD | SALES |
| 7566 | JONES | RESEARCH |
| 7654 | MARTIN | SALES |
| 7698 | BLAKE | SALES |
| 7782 | CLARK | ACCOUNTING |
| 7788 | SCOTT | RESEARCH |
| 7839 | KING | ACCOUNTING |
| 7844 | TURNER | SALES |
| 7876 | ADAMS | RESEARCH |
| 7900 | JAMES | SALES |
| 7902 | FORD | RESEARCH |
| 7934 | MILLER | ACCOUNTING |
+----------+----------+-------------+--+
14 rows selected (11.416 seconds)
开窗函数
0: jdbc:hive2://CentOS:10000> select e.empno ,e.ename,e.sal,e.deptno,rank() over(partition by e.deptno order by e.sal) as rank from t_employee e;
+----------+----------+--------+-----------+-------+--+
| e.empno | e.ename | e.sal | e.deptno | rank |
+----------+----------+--------+-----------+-------+--+
| 7839 | KING | 5000 | 10 | 1 |
| 7782 | CLARK | 2450 | 10 | 2 |
| 7934 | MILLER | 1300 | 10 | 3 |
| 7902 | FORD | 3000 | 20 | 1 |
| 7566 | JONES | 2975 | 20 | 2 |
| 7788 | SCOTT | 1500 | 20 | 3 |
| 7876 | ADAMS | 1100 | 20 | 4 |
| 7369 | SMITH | 800 | 20 | 5 |
| 7698 | BLAKE | 2850 | 30 | 1 |
| 7499 | ALLEN | 1600 | 30 | 2 |
| 7844 | TURNER | 1500 | 30 | 3 |
| 7654 | MARTIN | 1250 | 30 | 4 |
| 7521 | WARD | 1250 | 30 | 4 |
| 7900 | JAMES | 950 | 30 | 6 |
+----------+----------+--------+-----------+-------+--+
0: jdbc:hive2://CentOS:10000> select e.empno ,e.ename,e.sal,e.deptno,dense_rank() over(partition by e.deptno order by e.sal desc) as rank from t_employee e;
+----------+----------+--------+-----------+-------+--+
| e.empno | e.ename | e.sal | e.deptno | rank |
+----------+----------+--------+-----------+-------+--+
| 7839 | KING | 5000 | 10 | 1 |
| 7782 | CLARK | 2450 | 10 | 2 |
| 7934 | MILLER | 1300 | 10 | 3 |
| 7902 | FORD | 3000 | 20 | 1 |
| 7566 | JONES | 2975 | 20 | 2 |
| 7788 | SCOTT | 1500 | 20 | 3 |
| 7876 | ADAMS | 1100 | 20 | 4 |
| 7369 | SMITH | 800 | 20 | 5 |
| 7698 | BLAKE | 2850 | 30 | 1 |
| 7499 | ALLEN | 1600 | 30 | 2 |
| 7844 | TURNER | 1500 | 30 | 3 |
| 7654 | MARTIN | 1250 | 30 | 4 |
| 7521 | WARD | 1250 | 30 | 4 |
| 7900 | JAMES | 950 | 30 | 5 |
+----------+----------+--------+-----------+-------+--+
14 rows selected (24.262 seconds)
Cube分析
0: jdbc:hive2://CentOS:10000> select e.deptno,e.job,avg(e.sal) avg,max(e.sal) max,min(e.sal) min from t_employee e group by e.deptno,e.job with cube;
+-----------+------------+--------------+-------+-------+--+
| e.deptno | e.job | avg | max | min |
+-----------+------------+--------------+-------+-------+--+
| NULL | ANALYST | 2250 | 3000 | 1500 |
| 10 | CLERK | 1300 | 1300 | 1300 |
| 20 | CLERK | 950 | 1100 | 800 |
| 30 | CLERK | 950 | 950 | 950 |
| 20 | ANALYST | 2250 | 3000 | 1500 |
| NULL | PRESIDENT | 5000 | 5000 | 5000 |
| 10 | PRESIDENT | 5000 | 5000 | 5000 |
| NULL | SALESMAN | 1400 | 1600 | 1250 |
| NULL | MANAGER | 2758.333333 | 2975 | 2450 |
| 30 | SALESMAN | 1400 | 1600 | 1250 |
| 10 | MANAGER | 2450 | 2450 | 2450 |
| 20 | MANAGER | 2975 | 2975 | 2975 |
| 30 | MANAGER | 2850 | 2850 | 2850 |
| NULL | NULL | 1966.071429 | 5000 | 800 |
| NULL | CLERK | 1037.5 | 1300 | 800 |
| 10 | NULL | 2916.666667 | 5000 | 1300 |
| 20 | NULL | 1875 | 3000 | 800 |
| 30 | NULL | 1566.666667 | 2850 | 950 |
+-----------+------------+--------------+-------+-------+--+
18 rows selected (25.037 seconds)
行转列
1,语文,100
1,数学,100
1,英语,100
2,数学,79
2,语文,80
2,英语,100
CREATE TABLE t_student(
id INT,
course STRING,
score double)
row format delimited
fields terminated by ','
collection items terminated by '|'
map keys terminated by '>'
lines terminated by '\n'
stored as textfile;
0: jdbc:hive2://CentOS:10000> select * from t_student;
+---------------+-------------------+------------------+--+
| t_student.id | t_student.course | t_student.score |
+---------------+-------------------+------------------+--+
| 1 | 语文 | 100.0 |
| 1 | 数学 | 100.0 |
| 1 | 英语 | 100.0 |
| 2 | 数学 | 79.0 |
| 2 | 语文 | 80.0 |
| 2 | 英语 | 100.0 |
+---------------+-------------------+------------------+--+
6 rows selected (0.05 seconds)
0: jdbc:hive2://CentOS:10000> select id,max(case course when '语文' then score else 0 end) as chinese,max(case course when '数学' then score else 0 end ) as math,max(case course when '英语' then score else 0 end ) as english from t_student group by id ;
+-----+----------+--------+----------+--+
| id | chinese | math | english |
+-----+----------+--------+----------+--+
| 1 | 100.0 | 100.0 | 100.0 |
| 2 | 80.0 | 79.0 | 100.0 |
+-----+----------+--------+----------+--+
2 rows selected (25.617 seconds)
SELECT id,concat_ws(’,’, collect_set(concat(course, ‘:’, score))) 成绩 FROM t_student GROUP BY id
Hive数据倾斜
数据倾斜是进行大数据计算时最经常遇到的问题之一。当我们在执行HiveQL或者运行MapReduce作业时候，如果遇到一直卡在map100%,reduce99%一般就是遇到了数据倾斜的问题。数据倾斜其实是进行分布式计算的时候，某些节点的计算能力比较强或者需要计算的数据比较少，早早执行完了，某些节点计算的能力较差或者由于此节点需要计算的数据比较多，导致出现其他节点的reduce阶段任务执行完成，但是这种节点的数据处理任务还没有执行完成。
group by,我使用Hive对数据做一些类型统计的时候遇到过某种类型的数据量特别多，而其他类型数据的数据量特别少。当按照类型进行group by的时候，会将相同的group by字段的reduce任务需要的数据拉取到同一个节点进行聚合，而当其中每一组的数据量过大时，会出现其他组的计算已经完成而这里还没计算完成，其他节点的一直等待这个节点的任务执行完成，所以会看到一直map 100% reduce 99%的情况。
解决方法：
set hive.map.aggr=true
set hive.groupby.skewindata=true
原理：
hive.map.aggr=true 这个配置项代表是否在map端进行聚合hive.groupby.skwindata=true 当选项设定为 true，生成的查询计划会有两个 MR Job。第一个 MR Job 中，Map 的输出结果集合会随机分布到 Reduce 中，每个 Reduce 做部分聚合操作，并输出结果，这样处理的结果是相同的 Group By Key 有可能被分发到不同的 Reduce 中，从而达到负载均衡的目的；第二个 MR Job 再根据预处理的数据结果按照 Group By Key 分布到 Reduce 中（这个过程可以保证相同的 Group By Key 被分布到同一个 Reduce 中），最后完成最终的聚合操作。
Hive On Hbase
create external table t_employee(
empno INT,
ename STRING,
job STRING,
mgr INT,
hiredate TIMESTAMP,
sal DECIMAL(7,2),
comm DECIMAL(7,2),
deptno INT)
STORED BY 'org.apache.hadoop.hive.hbase.HBaseStorageHandler'
WITH SERDEPROPERTIES("hbase.columns.mapping" = ":key,cf1:name,cf1:job,cf1:mgr,cf1:hiredate,cf1:sal,cf1:comm,cf1:deptno")
TBLPROPERTIES("hbase.table.name" = "baizhi:t_employee");
需要替换hive-hbase-handler-1.2.2.jar