Oralce 高级 sql 之 rollup 与 cube

在 oracle sql 查询中，oracle 提供了一些高级功能能够简化 sql 查询语句以及提高查询性能。下面将介绍 rollup，cube的用法与区别。

1. rollup

rollup 能够在 select 语句中计算指定分组字段的多个层次的小计以及合计. rollup 非常容易使用并且非常高效。
rollup 从右到左扫描分组字段, 逐步创建更高级别的小计. 最后再创建一行总计. rollup 将创建 n + 1 个层次(级别)的小计, n 指的是 rollup 中的字段数. 

1.1 什么时候使用 rollup

SQL> create table emp as select * from scott.emp; Table created. SQL> create table dept as select * from scott.dept; Table created. SQL> select * from emp; EMPNO ENAME JOB MGR HIREDATE SAL COMM DEPTNO ---------- ---------- --------- ---------- --------- ---------- ---------- ---------- 7369 SMITH CLERK 7902 17-DEC-80 800 20 7499 ALLEN SALESMAN 7698 20-FEB-81 1600 300 30 7521 WARD SALESMAN 7698 22-FEB-81 1250 500 30 7566 JONES MANAGER 7839 02-APR-81 2975 20 7654 MARTIN SALESMAN 7698 28-SEP-81 1250 1400 30 7698 BLAKE MANAGER 7839 01-MAY-81 2850 30 7782 CLARK MANAGER 7839 09-JUN-81 2450 10 7788 SCOTT ANALYST 7566 19-APR-87 3000 20 7839 KING preSIDENT 17-NOV-81 5000 10 7844 TURNER SALESMAN 7698 08-SEP-81 1500 0 30 7876 ADAMS CLERK 7788 23-MAY-87 1100 20 7900 JAMES CLERK 7698 03-DEC-81 950 30 7902 FORD ANALYST 7566 03-DEC-81 3000 20 7934 MILLER CLERK 7782 23-JAN-82 1300 10 14 rows selected. SQL> select * from dept; DEPTNO DNAME LOC ---------- -------------- ------------- 10 ACCOUNTING NEW YORK 20 RESEARCH DALLAS 30 SALES CHICAGO 40 OPERATIONS BOSTON




如果想统计每个部门每个职位的总薪水，sql 语句如下：

SQL> select b.dname, a.job, sum(a.sal)
  2  from emp a, dept b
  3  where a.deptno = b.deptno
  4  group by b.dname, a.job;

DNAME          JOB       SUM(A.SAL)
-------------- --------- ----------
SALES          MANAGER         2850
SALES          CLERK            950
ACCOUNTING     MANAGER         2450
ACCOUNTING     preSIDENT       5000
ACCOUNTING     CLERK           1300
SALES          SALESMAN        5600
RESEARCH       MANAGER         2975
RESEARCH       ANALYST         6000
RESEARCH       CLERK           1900

9 rows selected.


如果想用一条 sql 语句统计每个部门每个职位的总薪水，每个部门的总薪水以及所有部门的总薪水的话，势必要再用一个 group by 语句统计每个部门的总薪水然后再与上面的结果 union 起来才能得到最后的结果，但是这样写的话除了书写会复杂一点外，另外肯定还要多扫描几次 emp 与 dept 表，现在 oracle 提供了 rollup 子句，我们可以先来看看 rollup 子句的结果：


SQL> select b.dname, a.job, sum(a.sal)
  2  from emp a, dept b
  3  where a.deptno = b.deptno
  4  group by rollup(b.dname, a.job);

DNAME          JOB       SUM(A.SAL)
-------------- --------- ----------
SALES          CLERK            950
SALES          MANAGER         2850
SALES          SALESMAN        5600
SALES                          9400
RESEARCH       CLERK           1900
RESEARCH       ANALYST         6000
RESEARCH       MANAGER         2975
RESEARCH                      10875
ACCOUNTING     CLERK           1300
ACCOUNTING     MANAGER         2450
ACCOUNTING     preSIDENT       5000
ACCOUNTING                     8750
                              29025




从上面的结果中可以看出，rollup 子句会为每个部门增加一行小计以及为所有部门增加一行总计，即统计了每个部门的总薪水以及所有部门的总薪水。

通常来说，rollup 往往同 group by 语句一起使用，它是 group by 语句的一种扩展。

• 如果语句为 group by rollup(a, b)，oracle 将会从右到左先对字段 a 和 b 先进行 group by，然后对字段 A 进行 group by，最后对全表进行 group by。
• 如果语句为 group by rollup(a, b, c), oracle 将从右到左先会对字段 a 和 b 和 c 先进行 group by, 然后对字段 a 和 b 进行 group by，然后对字段 a 进行 group by，最后对全表进行 group by。

下面我们将演示一个 rollup 三个字段的例子：

SQL> select b.dname, a.job, to_char(hiredate, 'yyyy'), sum(sal)
  2  from emp a, dept b
  3  where a.deptno = b.deptno
  4  group by b.dname, a.job, to_char(hiredate, 'yyyy')
  5  order by 1, 2, 3;

DNAME          JOB       TO_C   SUM(SAL)
-------------- --------- ---- ----------
ACCOUNTING     CLERK     1982       1300
ACCOUNTING     MANAGER   1981       2450
ACCOUNTING     preSIDENT 1981       5000
RESEARCH       ANALYST   1981       3000
RESEARCH       ANALYST   1987       3000
RESEARCH       CLERK     1980        800
RESEARCH       CLERK     1987       1100
RESEARCH       MANAGER   1981       2975
SALES          CLERK     1981        950
SALES          MANAGER   1981       2850
SALES          SALESMAN  1981       5600

11 rows selected.

SQL> select b.dname, a.job, to_char(hiredate, 'yyyy'), sum(sal)
  2  from emp a, dept b
  3  where a.deptno = b.deptno
  4  group by rollup(b.dname, a.job, to_char(hiredate, 'yyyy'));

DNAME          JOB       TO_C   SUM(SAL)
-------------- --------- ---- ----------
SALES          CLERK     1981        950
SALES          CLERK                 950
SALES          MANAGER   1981       2850
SALES          MANAGER              2850
SALES          SALESMAN  1981       5600
SALES          SALESMAN             5600
SALES                               9400
RESEARCH       CLERK     1980        800
RESEARCH       CLERK     1987       1100
RESEARCH       CLERK                1900
RESEARCH       ANALYST   1981       3000
RESEARCH       ANALYST   1987       3000
RESEARCH       ANALYST              6000
RESEARCH       MANAGER   1981       2975
RESEARCH       MANAGER              2975
RESEARCH                           10875
ACCOUNTING     CLERK     1982       1300
ACCOUNTING     CLERK                1300
ACCOUNTING     MANAGER   1981       2450
ACCOUNTING     MANAGER              2450
ACCOUNTING     preSIDENT 1981       5000
ACCOUNTING     preSIDENT            5000
ACCOUNTING                          8750
                                   29025

24 rows selected.




1.3 部分 rollup(Partial rollup)




当你只想统计部分字段时, 可以使用部分 rollup. 例如, group by a, rollup(b, c), 这条语句将创建三个(2 + 1)级别的小计. 分别为级别 (a, b, c), 级别 (a, b) 以及级别 (a).


SQL> select b.dname, a.job, to_char(hiredate, 'yyyy'), sum(sal)
  2  from emp a, dept b
  3  where a.deptno = b.deptno
  4  group by b.dname, rollup(a.job, to_char(hiredate, 'yyyy'));

DNAME          JOB       TO_C   SUM(SAL)
-------------- --------- ---- ----------
SALES          CLERK     1981        950
SALES          CLERK                 950
SALES          MANAGER   1981       2850
SALES          MANAGER              2850
SALES          SALESMAN  1981       5600
SALES          SALESMAN             5600
SALES                               9400
RESEARCH       CLERK     1980        800
RESEARCH       CLERK     1987       1100
RESEARCH       CLERK                1900
RESEARCH       ANALYST   1981       3000
RESEARCH       ANALYST   1987       3000
RESEARCH       ANALYST              6000
RESEARCH       MANAGER   1981       2975
RESEARCH       MANAGER              2975
RESEARCH                           10875
ACCOUNTING     CLERK     1982       1300
ACCOUNTING     CLERK                1300
ACCOUNTING     MANAGER   1981       2450
ACCOUNTING     MANAGER              2450
ACCOUNTING     preSIDENT 1981       5000
ACCOUNTING     preSIDENT            5000
ACCOUNTING                          8750

23 rows selected.




从上面的结果中可以看出，部分 rollup 产生的结果:

• 普通的汇总行是由 group by 产生而不是 rollup
• 不会产生总计




2. cube




cube 可以为指定的列创建各种不同组合的小计. 如果指定的列的数量为 n, group by cube 将创建 2 * n 个层次的小计. cube 是一种比 rollup 更细粒度的分组统计语句。先看看 cube 语句的结果：




2.1 什么时候使用 cube

• 当需求中有类似 cross-tabular report (交叉报表)时
• 对于数据仓库中的统计汇总表, rollup 能够简化统计汇总表并且提高查询统计汇总表的速度




2.2 cube 例子

SQL> select b.dname, a.job, sum(a.sal)
  2  from emp a, dept b
  3  where a.deptno = b.deptno
  4  group by cube(b.dname, a.job);

DNAME          JOB       SUM(A.SAL)
-------------- --------- ----------
                              29025
               CLERK           4150
               ANALYST         6000
               MANAGER         8275
               SALESMAN        5600
               preSIDENT       5000
SALES                          9400
SALES          CLERK            950
SALES          MANAGER         2850
SALES          SALESMAN        5600
RESEARCH                      10875
RESEARCH       CLERK           1900
RESEARCH       ANALYST         6000
RESEARCH       MANAGER         2975
ACCOUNTING                     8750
ACCOUNTING     CLERK           1300
ACCOUNTING     MANAGER         2450
ACCOUNTING     preSIDENT       5000

18 rows selected.


从上面的结果可以看出，group by cube(b.dname, a.job) 语句首先统计所有部门的总薪水，然后统计每个职位(a.job)的总薪水，然后统计每个部门(b.dname)的总薪水,最后统计每个部门每个职位(b.dname, a.job)的总薪水。
• 如果语句为 group by cube(a, b)，oracle 首先对字段 a 和 b 进行 group by，然后对字段 a 进行 group by，然后对字段 b 进行 group by，最后对全表进行 group by。
• 如果语句为 group by cube(a, b, c)，oracle 进行分组的字段分别为 (a, b, c)，(a, b)，(a, c)，(b, c)，(a)，(b)，(c)，最后对全表的总计

下面演示 cube 三个字段的例子：

SQL> select b.dname, a.job, to_char(hiredate, 'yyyy'), sum(sal)
  2  from emp a, dept b
  3  where a.deptno = b.deptno
  4  group by cube(b.dname, a.job, to_char(hiredate, 'yyyy'));

DNAME          JOB       TO_C   SUM(SAL)
-------------- --------- ---- ----------
                                   29025
                         1980        800
                         1981      22825
                         1982       1300
                         1987       4100
               CLERK                4150
               CLERK     1980        800
               CLERK     1981        950
               CLERK     1982       1300
               CLERK     1987       1100
               ANALYST              6000
               ANALYST   1981       3000
               ANALYST   1987       3000
               MANAGER              8275
               MANAGER   1981       8275
               SALESMAN             5600
               SALESMAN  1981       5600
               preSIDENT            5000
               preSIDENT 1981       5000
SALES                               9400
SALES                    1981       9400
SALES          CLERK                 950
SALES          CLERK     1981        950
SALES          MANAGER              2850
SALES          MANAGER   1981       2850
SALES          SALESMAN             5600
SALES          SALESMAN  1981       5600
RESEARCH                           10875
RESEARCH                 1980        800
RESEARCH                 1981       5975
RESEARCH                 1987       4100
RESEARCH       CLERK                1900
RESEARCH       CLERK     1980        800
RESEARCH       CLERK     1987       1100
RESEARCH       ANALYST              6000
RESEARCH       ANALYST   1981       3000
RESEARCH       ANALYST   1987       3000
RESEARCH       MANAGER              2975
RESEARCH       MANAGER   1981       2975
ACCOUNTING                          8750
ACCOUNTING               1981       7450
ACCOUNTING               1982       1300
ACCOUNTING     CLERK                1300
ACCOUNTING     CLERK     1982       1300
ACCOUNTING     MANAGER              2450
ACCOUNTING     MANAGER   1981       2450
ACCOUNTING     preSIDENT            5000
ACCOUNTING     preSIDENT 1981       5000

48 rows selected.


2.3 部分 cube (partial cube)
部分 cube 类似于部分 rollup, 把列放在 cube 操作符的外面能够限制生成列组合的小计. 例如, group by a, cube(b, c), 这条语句将产生 4 (2 * 2) 个层次的小计, 分别为层次 (a, b, c), 层次 (a, b), 层次 (a, c), 层次 (a). 



SQL> select b.dname, a.job, to_char(hiredate, 'yyyy'), sum(sal)
  2  from emp a, dept b
  3  where a.deptno = b.deptno
  4  group by b.dname, cube(a.job, to_char(hiredate, 'yyyy'));

DNAME          JOB       TO_C   SUM(SAL)
-------------- --------- ---- ----------
SALES                               9400
SALES                    1981       9400
SALES          CLERK                 950
SALES          CLERK     1981        950
SALES          MANAGER              2850
SALES          MANAGER   1981       2850
SALES          SALESMAN             5600
SALES          SALESMAN  1981       5600
RESEARCH                           10875
RESEARCH                 1980        800
RESEARCH                 1981       5975
RESEARCH                 1987       4100
RESEARCH       CLERK                1900
RESEARCH       CLERK     1980        800
RESEARCH       CLERK     1987       1100
RESEARCH       ANALYST              6000
RESEARCH       ANALYST   1981       3000
RESEARCH       ANALYST   1987       3000
RESEARCH       MANAGER              2975
RESEARCH       MANAGER   1981       2975
ACCOUNTING                          8750
ACCOUNTING               1981       7450
ACCOUNTING               1982       1300
ACCOUNTING     CLERK                1300
ACCOUNTING     CLERK     1982       1300
ACCOUNTING     MANAGER              2450
ACCOUNTING     MANAGER   1981       2450
ACCOUNTING     preSIDENT            5000
ACCOUNTING     preSIDENT 1981       5000

29 rows selected.