现在的位置: 首页 > 综合 > 正文

oracle rollup和cube分析

2012年07月11日 ⁄ 综合 ⁄ 共 8691字 ⁄ 字号 评论关闭

rollup是对group by的扩展,会进行小计和合计,而cube包含rollup,是粒度更精细的小计和合计。当只有一个字段时,rollup和cube是一样的。

    可用grouping对rollup和cube进行测试:0看;1不看[看表示列有显示,不看表示没有显示]

             ① rollup
                  ㈠ 什么都不看
                  ㈡ 从左往右  ┌ 先看第1个
                                        │ 再看前2个
                                        └ 后看前3个
             ② cube:0、1全排列

   rollup

    rollup后面指定的列以逗号分隔,rollup的计算结果和其后面指定的列的顺序有关,因为rollup的分组具有方向性。若指定n列,则有n+1种分组方式。可改变列的顺序,达到不同的业务需求。一定要牢记列的顺序对结果的影响

    例子:

  1. 20:32:51 scott@ORCL (^ω^) select a.dname,b.job,sum(b.sal) sum_sal,grouping(a.dname),grouping(b.job) 
  2. 20:33:06   2    from dept a,emp b 
  3. 20:33:06   3   where a.deptno=b.deptno 
  4. 20:33:06   4   group by
    rollup(a.dname,b.job) 
  5. 20:33:08   5  / 
  6.  
  7. DNAME      JOB           SUM_SAL GROUPING(A.DNAME)
    GROUPING(B.JOB) 
  8. ---------- ---------- ---------- ----------------- --------------- 
  9. SALES      CLERK             950                 0               0 
  10. SALES      MANAGER          2850                 0               0 
  11. SALES      SALESMAN         5600                 0               0 
  12. SALES                       9400                 0               1 
  13. RESEARCH   CLERK            1200                 0               0 
  14. RESEARCH   ANALYST          7000                 0               0 
  15. RESEARCH   MANAGER          2975                 0               0 
  16. RESEARCH                   11175                 0               1 
  17. ACCOUNTING CLERK            1300                 0               0 
  18. ACCOUNTING MANAGER          2450                 0               0 
  19. ACCOUNTING PRESIDENT        5000                 0               0 
  20. ACCOUNTING                  8750                 0               1 
  21.                            29325                 1               1 
  22.  
  23. 已选择13行。 

    解释:什么都不看(1,1);从左往右,先看第一个(0,1)

    可以将不需要进行小计和合计的列移出rollup,要小计的留在rollup里面。

    比如:

  1. 20:48:37 scott@ORCL (^ω^) select a.dname,b.job,sum(b.sal) sum_sal,grouping(a.dname),grouping(b.job) 
  2. 20:57:43   2    from dept a,emp b 
  3. 20:57:43   3   where a.deptno=b.deptno 
  4. 20:57:43   4   group by a.dname,rollup(b.job) 
  5. 20:57:45   5  / 
  6.  
  7. DNAME                        JOB                   SUM_SAL
    GROUPING
    (A.DNAME) GROUPING(B.JOB) 
  8. ---------------------------- ------------------ ---------- ----------------- --------------- 
  9. SALES                        CLERK                     950                 0               0 
  10. SALES                        MANAGER                  2850                 0               0 
  11. SALES                        SALESMAN                 5600                 0               0 
  12. SALES                                                 9400                 0               1 
  13. RESEARCH                     CLERK                    1200                 0               0 
  14. RESEARCH                     ANALYST                  7000                 0               0 
  15. RESEARCH                     MANAGER                  2975                 0               0 
  16. RESEARCH                                             11175                 0               1 
  17. ACCOUNTING                   CLERK                    1300                 0               0 
  18. ACCOUNTING                   MANAGER                  2450                 0               0 
  19. ACCOUNTING                   PRESIDENT                5000                 0               0 
  20. ACCOUNTING                                            8750                 0               1 
  21.  
  22. 已选择12行。 

  

   cube

    和rollup不同,cube的计算结果和顺序无关。若n列,则分组方式有2的n次方种。

  1. 20:57:46 scott@ORCL (^ω^) select a.dname,b.job,sum(b.sal) sum_sal,grouping(a.dname),grouping(b.job) 
  2. 21:27:03   2    from dept a,emp b 
  3. 21:27:03   3   where a.deptno=b.deptno 
  4. 21:27:03   4   group by
    cube(a.dname,b.job) 
  5. 21:27:04   5  / 
  6.  
  7. DNAME                        JOB                   SUM_SAL
    GROUPING
    (A.DNAME) GROUPING(B.JOB) 
  8. ---------------------------- ------------------ ---------- ----------------- --------------- 
  9.                                                      29325                 1               1 
  10.                              CLERK                    3450                 1               0 
  11.                              ANALYST                  7000                 1               0 
  12.                              MANAGER                  8275                 1               0 
  13.                              SALESMAN                 5600                 1               0 
  14.                              PRESIDENT                5000                 1               0 
  15. SALES                                                 9400                 0               1 
  16. SALES                        CLERK                     950                 0               0 
  17. SALES                        MANAGER                  2850                 0               0 
  18. SALES                        SALESMAN                 5600                 0               0 
  19. RESEARCH                                             11175                 0               1 
  20. RESEARCH                     CLERK                    1200                 0               0 
  21. RESEARCH                     ANALYST                  7000                 0               0 
  22. RESEARCH                     MANAGER                  2975                 0               0 
  23. ACCOUNTING                                            8750                 0               1 
  24. ACCOUNTING                   CLERK                    1300                 0               0 
  25. ACCOUNTING                   MANAGER                  2450                 0               0 
  26. ACCOUNTING                   PRESIDENT                5000                 0               0 
  27.  
  28. 已选择18行。 

    注释:和rollup的结果相比,cube的所有可能的分组都走一遍。

    可以去掉合计和某些不需要的小计,通过部分cube实现。部分cube比部分rollup来得有用多了。

  1. 21:27:06 scott@ORCL (^ω^) select a.dname,b.job,sum(b.sal) sum_sal,grouping(a.dname),grouping(b.job) 
  2. 21:32:20   2    from dept a,emp b 
  3. 21:32:20   3   where a.deptno=b.deptno 
  4. 21:32:20   4   group by a.dname,cube(b.job) 
  5. 21:32:21   5  / 
  6.  
  7. DNAME                        JOB                   SUM_SAL
    GROUPING
    (A.DNAME) GROUPING(B.JOB) 
  8. ---------------------------- ------------------ ---------- ----------------- --------------- 
  9. SALES                                                 9400                 0               1 
  10. SALES                        CLERK                     950                 0               0 
  11. SALES                        MANAGER                  2850                 0               0 
  12. SALES                        SALESMAN                 5600                 0               0 
  13. RESEARCH                                             11175                 0               1 
  14. RESEARCH                     CLERK                    1200                 0               0 
  15. RESEARCH                     ANALYST                  7000                 0               0 
  16. RESEARCH                     MANAGER                  2975                 0               0 
  17. ACCOUNTING                                            8750                 0               1 
  18. ACCOUNTING                   CLERK                    1300                 0               0 
  19. ACCOUNTING                   MANAGER                  2450                 0               0 
  20. ACCOUNTING                   PRESIDENT                5000                 0               0 
  21.  
  22. 已选择12行。 

原文地址:http://blog.csdn.net/linwaterbin/article/details/7985717

抱歉!评论已关闭.