Oracle高级分组

--oracle中GROUP BY中使用的列必须与select语句中没有使用聚合函数的列相匹配

  SELECTd.dname,COUNT(e.empno)

    FROMdeptd,empe

   WHEREe.deptno(+)=D.DEPTNO

GROUPBYD.DNAME;

  SELECTd.dname,COUNT(E.EMPNO)

    FROMdeptd LEFTOUTERJOINempeONE.DEPTNO=D.DEPTNO

GROUPBYd.dname;

--GROUP BY 所产生的结果可以通过HAVING子句中所给出的筛选标准来进行限制

--需求:从有雇员以来的第一个完整年开始所有至少雇佣了5名员工的部门

SELECT D.DNAME,TRUNC(e.hiredate,'yyyy'),COUNT(e.empno)

FROMdeptd,empe

WHERE D.DEPTNO=E.DEPTNO

GROUPBYD.DNAME,TRUNC(e.hiredate,'yyyy')

  HAVING     COUNT(e.empno)>5

         ANDTRUNC(e.hiredate,'yyyy')BETWEEN(SELECTMIN(hiredate)

                                                  FROMemp)

                                            AND(SELECTMAX(hiredate)

                                                  FROMemp);

--CUBE将会使得对每一行都要考虑包含CUBE参数中所有可能的元素组合

CREATETABLEtestCube

(

   num1   NUMBER,

   num2   NUMBER

);

INSERTINTOtestCube

     VALUES(1,2);

COMMIT;

  SELECT*

    FROMtestCube

GROUPBYCUBE(num1,num2);

--GROUPING可以排除cube生成的空值

  SELECTDECODE(GROUPING(num1),1,'NUM1',num1),

         DECODE(GROUPING(num2),1,'NUM2',num2)

    FROMtestCube

GROUPBYCUBE(num1,num2);

--GROUPING_ID确定其参数中的哪一行被用来生成超聚合行，然后创建一个位矢量以数值返回

WITH testGroupid AS(SELECT1 bit_1,0 bit_0 FROM DUAL),

     cubed

     AS(  SELECTGROUPING_ID(bit_1,bit_0) gid,

                 TO_CHAR(GROUPING(bit_1))bv_1,

                 TO_CHAR(GROUPING(bit_0))bv_0,

                 DECODE(GROUPING(bit_1),1,'GRP BIT1') gb_1,

                 DECODE(GROUPING(bit_0),1,'GRP BIT0') gb_0

             FROMtestGroupid

         GROUPBYCUBE(bit_1,bit_0))

  SELECTgid,

         bv_1||bv_0 bit_vector,

         gb_1,

         gb_0

    FROMcubed

ORDERBYgid;

GROUP BY 局限性

1、LOB列，嵌套表和数组不能使用GROUPBY

2、标量子查询表达式不允许

3、GROUP BY引用对象类型列不能并行化