1. 聚合函数
通俗说,聚合函数是将多行数据导出一条结果
常用的聚合函数有:max,min,count,sum,avg对emp表查询员工的最大、最小、平均工资及所有工资的和
1. hive> select * from emp; 2. OK 3. 7369 SMITH CLERK 7902 1980/12/17 800.0 NULL 20 4. 7499 ALLEN SALESMAN 7698 1981/2/20 1600.0 300.0 30 5. 7521 WARD SALESMAN 7698 1981/2/22 1250.0 500.0 30 6. 7566 JONES MANAGER 7839 1981/4/2 2975.0 NULL 20 7. 7654 MARTIN SALESMAN 7698 1981/9/28 1250.0 1400.0 30 8. 7698 BLAKE MANAGER 7839 1981/5/1 2850.0 NULL 30 9. 7782 CLARK MANAGER 7839 1981/6/9 2450.0 NULL 10 10. 7788 SCOTT ANALYST 7566 1987/4/19 3000.0 NULL 20 11. 7839 KING PRESIDENT NULL 1981/11/17 5000.0 NULL 10 12. 7844 TURNER SALESMAN 7698 1981/9/8 1500.0 0.0 30 13. 7876 ADAMS CLERK 7788 1987/5/23 1100.0 NULL 20 14. 7900 JAMES CLERK 7698 1981/12/3 950.0 NULL 30 15. 7902 FORD ANALYST 7566 1981/12/3 3000.0 NULL 20 16. 7934 MILLER CLERK 7782 1982/1/23 1300.0 NULL 10 17. Time taken: 5.998 seconds, Fetched: 14 row(s) 18. hive> select max(salary),min(salary),avg(salary),sum(salary) from emp; 19. 5000.0 800.0 2073.214285714286 29025.0
查询记录数count
1. hive> select count(*) from emp; 2. hive> select count(1) from emp; 3. 14
使用1和*都是对表进行记录数查询
2. 分组函数(group by)
分组字段要和查询字段匹配
分组函数会有mr
分组函数会产生数据倾斜按照部门进行分组
1. hive> select deptno from emp group by deptno; 2. 结果: 3. 10 4. 20 5. 30
查询每个部门的平均工资
1. hive> select deptno,avg(salary) avg_sal from emp group by deptno; 2. 结果: 3. 10 2916.6666666666665 4. 20 2175.0 5. 30 1566.6666666666667
查询平均工资大于2000的部门
1. hive> select deptno,avg(salary) from emp group by deptno having avg(salary) > 2000; 2. 结果: 3. 10 2916.6666666666665 4. 20 2175.0
分组函数使用条件语句不能使用where,用having
按照部门和入职时间进行分组
1. hive> select deptno,hiredate from emp group by deptno,hiredate; 2. 结果: 3. 10 1981/11/17 4. 10 1981/6/9 5. 10 1982/1/23 6. 20 1980/12/17 7. 20 1981/12/3 8. 20 1981/4/2 9. 20 1987/4/19 10. 20 1987/5/23 11. 30 1981/12/3 12. 30 1981/2/20 13. 30 1981/2/22 14. 30 1981/5/1 15. 30 1981/9/28 16. 30 1981/9/8
按照部门和入职时间进行分组并计算出每组的人数
1. hive> select deptno,hiredate,count(ename) from emp group by deptno,hiredate; 2. 结果: 3. 10 1981/11/17 1 4. 10 1981/6/9 1 5. 10 1982/1/23 1 6. 20 1980/12/17 1 7. 20 1981/12/3 1 8. 20 1981/4/2 1 9. 20 1987/4/19 1 10. 20 1987/5/23 1 11. 30 1981/12/3 1 12. 30 1981/2/20 1 13. 30 1981/2/22 1 14. 30 1981/5/1 1 15. 30 1981/9/28 1 16. 30 1981/9/8 1
3. case when then end(不会跑mr)
查询员工的姓名和工资等级,按如下规则显示
salary小于等于1000,显示LOWER
salaray大于1000且小于等于2000,显示MIDDLE
salaray大于2000小于等于4000,显示MIDDLE
sal大于4000,显示highest1. select ename, salary, 2. case 3. when salary > 1 and salary <= 1000 then 'LOWER' 4. when salary > 1000 and salary <= 2000 then 'MIDDLE' 5. when salary > 2000 and salary <= 4000 then 'HIGH' 6. ELSE 'HIGHEST' 7. end 8. from emp; 9. 结果: 10. SMITH 800.0 LOWER 11. ALLEN 1600.0 MIDDLE 12. WARD 1250.0 MIDDLE 13. JONES 2975.0 HIGH 14. MARTIN 1250.0 MIDDLE 15. BLAKE 2850.0 HIGH 16. CLARK 2450.0 HIGH 17. SCOTT 3000.0 HIGH 18. KING 5000.0 HIGHEST 19. TURNER 1500.0 MIDDLE 20. ADAMS 1100.0 MIDDLE 21. JAMES 950.0 LOWER 22. FORD 3000.0 HIGH 23. MILLER 1300.0 MIDDLE
4. join
- Hive中Join的关联键必须在ON ()中指定,不能在Where中指定,否则就会先做笛卡尔积,再过滤。
创建表
1. hive> create table join_a( 2. > id int, 3. > name string 4. > ) 5. > ROW FORMAT DELIMITED FIELDS TERMINATED BY "\t"; 6. hive> load data local inpath '/home/hadoop/data/join_a.txt' OVERWRITE INTO TABLE join_a; 7. hive> select * from join_a; 8. OK 9. 1 zhangsan 10. 2 lisi 11. 3 wangwu 12. hive> create table join_b( 13. > id int, 14. > age int 15. > ) 16. ROW FORMAT DELIMITED FIELDS TERMINATED BY "\t"; 17. hive> load data local inpath '/home/hadoop/data/join_b.txt' OVERWRITE INTO TABLE join_b; 18. hive> select * from join_b; 19. OK 20. 1 20 21. 2 30 22. 4 40
内连接
内连接即基于on语句,仅列出表1和表2符合连接条件的数据1. hive> select a.id,a.name,b.age from join_a a join join_b b on a.id=b.id; 2. 结果: 3. 1 zhangsan 20 4. 2 lisi 30 5. hive> select * from join_a a join join_b b on a.id=b.id; 6. 结果: 7. 1 zhangsan 1 20 8. 2 lisi 2 30
左连接
左连接是显示左边的表的所有数据,如果有右边表与之对应,则显示;否则显示null1. hive> select a.id,a.name,b.age from join_a a left join join_b b on a.id=b.id; 2. 结果: 3. 1 zhangsan 20 4. 2 lisi 30 5. 3 wangwu NULL
右连接
右连接是显示右边的表的所有数据,如果有左边表与之对应,则显示;否则显示null1. hive> select a.id,a.name,b.age from join_a a right join join_b b on a.id=b.id; 2. 结果: 3. 1 zhangsan 20 4. 2 lisi 30 5. NULL NULL 40
全连接
相当于表1和表2的数据都显示,如果没有对应的数据,则显示Null.- hive > select a.id,a.name,b.age from join_a a full join join_b b on a.id=b.id;
- 结果:
- 1 zhangsan 20
- 2 lisi 30
- 3 wangwu1 NULL
- NULL NULL 40
笛卡尔积
笛卡尔积(没有连接条件)会针对表1和表2的每条数据做连接
join(cross join)1. hive > select a.id,a.name,b.age from join_a a cross join join_b b; 2. 结果: 3. 1 zhangsan 20 4. 1 zhangsan 30 5. 1 zhangsan 40 6. 2 lisi 20 7. 2 lisi 30 8. 2 lisi 40 9. 3 wangwu1 20 10. 3 wangwu1 30 11. 3 wangwu1 40
来自@若泽大数据