Oracle 分析函数

分析函数语法解析

分析函数的语法很简单，但使用时却发现有很多令人迷惑的地方。本文就但从语法层次来分析其中的迷惑之处。
1. 分析函数整体语法
分析函数的语法如下：
analytic_function([ arguments ]) OVER (analytic_clause)
这里：
n        analytic_function 是分析函数的名称；
n        arguments 是分析函数的参数；
n        over 用来标识函数是一个分析函数；
n        analytic_clause 用来确定分析函数的操作规则。包括query_partition_clause、order_by_clause和windowing_clause三个子句。
2. Analytic_function
分析函数的名称。Oracle10gR2带的内置分析函数有39个；对于用户自定义的分析函数，分析函数名称需要满足标识符规则。
3. Arguments
分析函数所带参数，内置分析函数一般带0-3个参数。 参数可以是任何数字类型或是可以隐式转换为数字类型的数据类型 。Oracle根据最高数字优先级别确定函数参数，并且隐式地将需要处理的参数转换为数字类型。
用户自定义分析函数的参数，可以根据实际情况使用。
4. Over
用以标识函数是一个分析函数，对于既可作为聚集函数又可作为分析函数的函数，Oracle无法识别，必须用over来标识此函数为分析函数。也不是说只可用作分析函数的函数就可以省略该关键字，over关键字是必须的，是分析函数就必须使用。
Over后面的()中是analytic_clause，即使analytic_clause的三个部分全不使用，()也必不可省，否则Oracle因不能识别函数为分析函数而报错。
5. Analytic_clause
Analytic_clause的语法如下：
[ query_partition_clause ] [ order_by_clause [ windowing_clause ] ]
这里：
n        query_partition_clause 是查询分组子句；
n        order_by_clause 是分组排序子句；
n        windowing_clause 是窗口范围子句。
分析函数在查询结果集确定之后才开始进行计算，Analytic_clause就是用来定义函数怎样对查询结果集进行分组计算的。
根据Oracle对查询和分析函数的处理方法可知， 在select和order by子句中都可以使用分析函数。
query_partition_by、order_by_clause和windowing_clause三个子句都是可选的，将三个子句分别简记为P，O，W。
合法的组合方式有如下6种：
1).        POW
(query_partition_clause order_by_clause windowing_clause)
分组，排序，定义窗口范围
2).        PO
(query_partition_clause order_by_clause)
分组，排序，窗口默认为range between unbounded preceding and current row
3).        P
(query_partition_clause)
分组，不排序，没有窗口(range between unbounded preceding and unbounded following)
4).        OW
(order_by_clause windowing_clause)
分组为整个查询结果集，排序，定义窗口范围
5).        O
(order_by_clause)
分组为整个查询结果集，排序，窗口默认为range between unbounded preceding and current row
6).        NULL
()
分组为整个查询结果集，不排序，没有窗口(range between unbounded preceding and unbounded following)
因为只有存在order_by_clause，才能有windowing_clause，故不存在如下两种形式的组合：
PW(query_partition_clause windowing_clause)
W(windowing_clause)
总结：
1).        对于是否存在order_by_clause，分析函数可以分为两类，含有order_by_clause的一般称为windowing function，不含的称为reporting function。
2).        Windowing function，对查询结果集进行分组，排序，根据窗口范围计算分组中每一行的函数结果。
3).        Reporting function，对查询结果集进行分组，不排序，窗口范围为整个分组，在每一个分组内，计算整个分组的函数值，再将函数值分别赋给分组内的每一行。
注意事项：
1).        Analytic_clause内不能包含任何分析函数。
2).        用户自定义分析函数和内置函数分析函数都可以使用over (analytic_clause)。
5.1. Query_partition_clause
query_partition_clause的语法如下：
PARTITION BY value_expr[, value_expr ]...
这里：
n        partition by 是关键字，表示要将查询结果集分组；
n        value_expr 是分组表达式；
n        Partition by 子句根据一个或多个value_expr将查询结果集分成若干组。分组的原理与group by分组的原理类似，将value_expr相同的行集中在一起。若不使用该子句，那末函数将整个查询结果集作为一个分组；
n        value_expr 可以为常量、表列、非分析函数、函数表达式，或者前面这些元素的任意组合表达式。
5.2. Order_by_clause
Order_by_clause的语法如下：
ORDER BY expr [ ASC | DESC ] [ NULLS FIRST | NULLS LAST ]
[, expr  [ ASC | DESC ] [ NULLS FIRST | NULLS LAST ] ]...
这里：
n        order by 是关键字，表示要将分组排序；
n        expr 是排序表达式；
n        asc|desc 和nulls first|null last是排序的附加规则；
n        Order_by_clause 根据一个或多个expr对分组进行排序；
n        Expr 只能为表达式，不能为位置编号或列别名。
注意事项：
1).        当窗口范围为
Range between unbounded preceding and current row 或
Range between current row and unbounded following 时，可以在order_by_clause中使用多个expr对分组进行排序。其他range窗口范围只能使用一个expr。使用rows定义窗口范围无此限制。
2).        asc|desc 指定排序是按expr升序还是降序排序，若不指定则默认按asc进行升序排序。
3).        nulls first|nulls last 指定若返回行包含空值，该值应该出现在排序序列的开始还是末尾。升序排序的默认值是nulls last，降序排序的默认值是nulls first。
5.3. Windowing_clause
Windowing_clause的语法如下：
{ ROWS | RANGE }
{ BETWEEN
{ UNBOUNDED PRECEDING | CURRENT ROW | value_expr { PRECEDING | FOLLOWING } } AND
   { UNBOUNDED FOLLOWING | CURRENT ROW | value_expr { PRECEDING | FOLLOWING } }
| { UNBOUNDED PRECEDING | CURRENT ROW  | value_expr PRECEDING  }
}
这里，
n        rows 是关键字，指定窗口由物理行构成；
n        range 是关键字，指定窗口由逻辑偏移量构成；
n        between…and 是关键字，用来指定窗口的起始点和终结点；
n        Unbounded preceding 指明窗口开始于分组的第一行；
n        Current row ，作为起始点，指明窗口开始于当前行或当前值；作为终结点，指明窗口结束于当前行或当前值；
n        Unbounded following 指明窗口结束于分组的最后一行；
n        Value_expr 为物理或逻辑偏移量表达式。

Windowing_clause用来指定分组中当前行的计算范围。

不论rows还是range窗口，窗口总是在分组中从上至下滑动的。

窗口范围可以由between…and限定，也可以不用between…and，不用的都表示窗口到当前行结束。

n        ROWS 窗口
是由分组排序后分组中若干连续的行构成的窗口。

以下是全部合法的ROWS窗口范围:
1.        Rows between unbounded preceding and unbounded following
窗口开始于分组第一行，结束于分组最后一行。

2.        Rows [between] unbounded preceding [and current row]
窗口开始于分组第一行，结束于当前行。

3.        Rows between unbounded preceding and value_expr preceding
窗口开始于分组第一行，结束于当前行前value_expr行。

4.        Rows between unbounded preceding and value_expr following
窗口开始于分组第一行，结束于当前行后value_expr行。

5.        Rows between current row and unbounded following
窗口开始于当前行，结束于分组最后一行。

6.        Rows [between current row and] current row
窗口开始于当前行，结束于当前行。

7.        Rows between current row and value_expr following
窗口开始于当前行，结束于当前行后value_expr行。

8.        Rows between value_expr preceding and unbounded following
窗口开始于当前行前value_expr行，结束于分组最后一行。

9.        Rows [between value_expr] preceding [and current row]
窗口开始于当前行前value_expr行，结束于当前行。

10.        Rows between value_expr1 preceding and value_expr2 preceding
窗口开始于当前行前value_expr1行，结束于当前行前value_expr2行。

11.        Rows between value_expr1 preceding and value_expr2 following
窗口开始于当前行前value_expr1行，结束于当前行后value_expr2行。

12.        Rows between value_expr following and unbounded following
窗口开始于当前行后value_expr行，结束于分组最后一行。

13.        Rows between value_expr1 following and value_expr2 following
窗口开始于当前行后value_expr1行，结束于当前行后value_expr2行。

14.        Rows unbounded preceding
与2等价。

15.        Rows current row
与6等价。

16.        Rows value_expr preceding
与9等价。

n        RANGE 窗口
(1).对于单个expr，相当于给order_by_clause中的expr加一个where限定条件，即where expr between a and b构成了一个逻辑窗口，此窗口在expr包含的行上滑动求值。
a和b的值可结合windowing_clause来求值确定。order_by_clause子句为asc和desc，a和b的求值可能有所不同。

假设分组排序后第一行的值为first_value，最后一行的值为last_value。当前行的值为current_value。
unbounded preceding = first_value
unbounded following = last_value
current row = current_value
value_expr preceding = current_value -/+ value_expr
value_expr following = current_value +/- value_expr
下面是具体情况分类分析：
1.        Range between unbounded preceding and unbounded following
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between first_value and last_value

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between last_value and first_value

2.        Range [between] unbounded preceding [and current row]
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between first_value and current_value

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between current_value and first_value

3.        Range between unbounded preceding and value_expr preceding
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between first_value and current_value-value_expr

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between current_value+value_expr and first_value

4.        Range between unbounded preceding and value_expr following
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between first_value and current_value+value_expr

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between current_value-value_expr and first_value

5.        Range between current row and unbounded following
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between current_value and last_value

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between last_value and current_value

6.        Range [between current row and] current row
不论order_by_clause使用何种排序时，逻辑窗口如下：
Where expr = current_value

7.        Range between current row and value_expr following
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between current_value and current_value+value_expr

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between current_value-value_expr and current_value

8.        Range between value_expr preceding and unbounded following
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between current_value-value_expr and last_value

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between last_value and current_value+value_expr

9.        Range [between value_expr] preceding [and current row]
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between current_value-value_expr and current_value

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between current_value and current_value+value_expr

10.        Range between value_expr1 preceding and value_expr2 preceding
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between current_value-value_expr1
and current_value-value_expr2

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between current_value+value_expr2
and current_value+value_expr1

11.        Range between value_expr1 preceding and value_expr2 following
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between current_value-value_expr1
and current_value+value_expr2

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between current_value+value_expr2
and current_value-value_expr1

12.        Range between value_expr following and unbounded following
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between current_value+value_expr and last_value

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between last_value and current_value-value_expr

13.        Range between value_expr1 following and value_expr2 following
当order_by_clause使用asc排序时，逻辑窗口如下：
Where expr between current_value+value_expr1
and current_value+value_expr2

当order_by_clause使用desc排序时，逻辑窗口如下：
Where expr between current_value-value_expr2
and current_value-value_expr1

14.        Range unbounded preceding
与2等价。

15.        Range current row
与6等价。

16.        Range value_expr preceding
与9等价。

(2).对于在order_by_clause中可以使用多个expr的窗口：
u        Range between unbounded preceding and current row
u        Range between current row and unbounded following

当它们使用多个expr排序时(注意不是一个)分别等价于：
u        Rows between unbounded preceding and current row
u        Rows between current row and unbounded following

注意事项：
1.若windowing_clause由rows指定，则：
(1)value_expr是物理偏移量，它必须是常量或值为非负数的表达式。
(2)若value_expr是起点的一部分，那末它必须在终点之前对行求值。
2.若windowing_clause由range指定，则:
(1)value_expr是逻辑偏移量。它必须是常量或值为非负数的表达式或时间间隔文字常量。
(2)若value_expr值为一个数字，那末order_by_clause中 expr必须为数字或date类型。
(3)若value_expr为一个间隔值，那末order_by_clause中expr必须是一个date类型。
3.若完全忽略windowing_clause，那末默认的窗口范围为 range between unbounded preceding and current row 。
6. 分析函数简例
下面的例子都基于scott模式下的emp表。
SELECT * FROM emp;

EMPNO        ENAME            JOB             MGR        HIREDATE            SAL           COMM         DEPTNO
7839        KING          PRESIDENT                    1981-11-17 5000.00                       10
7698        BLAKE          MANAGER        7839        1981-5-1         2850.00                       30
7782        CLARK          MANAGER        7839        1981-6-9         2450.00                       10
7566        JONES          MANAGER        7839        1981-4-2         2975.00                       20
7654        MARTIN          SALESMAN        7698        1981-9-28  1250.00 1400.00           30
7499        ALLEN          SALESMAN        7698        1981-2-20  1600.00  300.00           30
7844        TURNER  SALESMAN        7698        1981-9-8         1500.00          0.00           30
7900        JAMES          CLERK            7698        1981-12-3   950.00                            30
7521        WARD          SALESMAN        7698        1981-2-22  1250.00  500.00           30
7902        FORD          ANALYST                  7566        1981-12-3  3000.00                           20
7369        SMITH          CLERK            7902        1980-12-17          800.00                           20
7788        SCOTT          ANALYST                  7566        1982-12-9   3000.00                           20
7876        ADAMS          CLERK            7788        1983-1-12   1100.00                           20
7934        MILLER          CLERK            7782        1982-1-23   1300.00                           10
6.1. E1
SELECT SUM(sal) over() sal_sum FROM emp;

1.        聚集函数用作分析函数时，不需要group by子句；
2.        分析函数作为表列可以有别名，但必须放在over()之后。
6.2. E2
SELECT empno, ename, job, SUM(sal) over(PARTITION BY mgr ORDER BY job)
  FROM emp;

1.        分析函数的分组与选择列也没有任何关系，比如上例，在选择列中可以有mgr，也可以没有mgr；
2.        分析函数的排序与选择列也没有任何关系，比如上例，在选择列中可以没有job。
6.3. E3
SELECT empno, ename, mgr, sal
  FROM emp
ORDER BY empno,
          row_number() over(PARTITION BY mgr ORDER BY sal),
          SUM(1) over(PARTITION BY empno)

1.        分析函数可用于order by子句中；
2.        Order by后可以用多个分析函数，也可以与其他排序列混合使用。
6.4. E4(pow)
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following)
  FROM emp;

           EMPNO        ENAME        MGR              SAL             HIREDATE        SAL_SUM
7902        FORD        7566        3000.00        1981-12-3         6000
7788        SCOTT        7566        3000.00        1982-12-9         6000
7499        ALLEN        7698        1600.00        1981-2-20         6550
7521        WARD        7698        1250.00        1981-2-22         6550
7844        TURNER7698        1500.00        1981-9-8         6550
7654        MARTIN7698        1250.00        1981-9-28         6550
7900        JAMES        7698        950.00            1981-12-3         6550
7934        MILLER        7782        1300.00        1982-1-23         1300
7876        ADAMS        7788        1100.00        1983-1-12         1100
7566        JONES        7839        2975.00        1981-4-2         8275
7698        BLAKE        7839        2850.00        1981-5-1         8275
7782        CLARK        7839        2450.00        1981-6-9         8275
7369        SMITH        7902        800.00            1980-12-17  800
7839        KING                    5000.00        1981-11-17 5000

按mgr分组，hiredate排序，窗口范围为整个分组。
分析函数sum对每个窗口中的所有雇员薪水求和，再将它赋给每一个雇员，故相同mgr的雇员的sal_sum都是相同的。
6.5. E5(po)
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate) sal_sum
  FROM emp;

           EMPNO        ENAME         MGR          SAL            HIREDATE         SAL_SUM
7902        FORD        7566        3000.00        1981-12-3          3000
7788        SCOTT        7566        3000.00        1982-12-9          6000
7499        ALLEN        7698        1600.00        1981-2-20          1600
7521        WARD        7698        1250.00        1981-2-22          2850
7844        TURNER7698        1500.00        1981-9-8          4350
7654        MARTIN7698        1250.00        1981-9-28          5600
7900        JAMES        7698        950.00            1981-12-3          6550
7934        MILLER        7782        1300.00        1982-1-23          1300
7876        ADAMS        7788        1100.00        1983-1-12          1100
7566        JONES        7839        2975.00        1981-4-2          2975
7698        BLAKE        7839        2850.00        1981-5-1          5825
7782        CLARK        7839        2450.00        1981-6-9          8275
7369        SMITH        7902        800.00            1980-12-17  800
7839        KING                    5000.00        1981-11-17  5000

按mgr分组，hiredate排序，没有指定窗口范围，默认为range between unbounded preceding and current row。
注意窗口是由hiredate确定，默认的窗口范围为不晚于当前雇员雇佣日期的hiredate。因又按hiredate升序，故sum对分组中当前雇员前面(包括当前雇员)的所有雇员求和,再赋给当前雇员。
可以将其补齐为pow的形式，等价于：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND CURRENT ROW) sal_sum
  FROM emp;

再看一个例子：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       COUNT(1) over(PARTITION BY mgr ORDER BY hiredate DESC) sal_sum
  FROM emp;

   EMPNO        ENAME         MGR          SAL            HIREDATE        SAL_SUM
7788        SCOTT        7566        3000.00        1982-12-9           1
7902        FORD        7566        3000.00        1981-12-3           2
7900        JAMES        7698        950.00            1981-12-3           1
7654        MARTIN7698        1250.00        1981-9-28           2
7844        TURNER7698        1500.00        1981-9-8           3
7521        WARD        7698        1250.00        1981-2-22           4
7499        ALLEN        7698        1600.00        1981-2-20           5
7934        MILLER        7782        1300.00        1982-1-23           1
7876        ADAMS        7788        1100.00        1983-1-12           1
7782        CLARK        7839        2450.00        1981-6-9           1
7698        BLAKE        7839        2850.00        1981-5-1           2
7566        JONES        7839        2975.00        1981-4-2           3
7369        SMITH        7902        800.00            1980-12-17   1
7839        KING                    5000.00        1981-11-17   1

这个例子统计相同经理下当前雇员雇佣日期不早于自己的雇员个数。
6.6. E6(p)
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       COUNT(1) over(PARTITION BY mgr) sal_sum
  FROM emp;

           EMPNO        ENAME         MGR          SAL            HIREDATE        SAL_SUM
7788        SCOTT        7566        3000.00        1982-12-9           2
7902        FORD        7566        3000.00        1981-12-3           2
7844        TURNER7698        1500.00        1981-9-8           5
7521        WARD        7698        1250.00        1981-2-22           5
7900        JAMES        7698        950.00            1981-12-3           5
7499        ALLEN        7698        1600.00        1981-2-20           5
7654        MARTIN7698        1250.00        1981-9-28           5
7934        MILLER        7782        1300.00        1982-1-23           1
7876        ADAMS        7788        1100.00        1983-1-12           1
7698        BLAKE        7839        2850.00        1981-5-1           3
7566        JONES        7839        2975.00        1981-4-2           3
7782        CLARK        7839        2450.00        1981-6-9           3
7369        SMITH        7902        800.00            1980-12-17   1
7839        KING                    5000.00        1981-11-17   1

按mgr分组，不排序，没有规定窗口范围。
这是个reporting function，不排序表明没有窗口，计算范围则为整个分组。窗口范围等价于range/rows between unbounded preceding and unbounded following。分析函数计算分组中的雇员总数，再将其赋给每一个雇员。
可以将其补齐为pow形式，等价于：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       COUNT(1) over(PARTITION BY mgr ORDER BY NULL ROWS BETWEEN unbounded preceding AND unbounded following) sal_sum
  FROM emp;
6.7. E7(ow)
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(ORDER BY hiredate RANGE BETWEEN 30 preceding AND 30 following) sal_sum
  FROM emp;

           EMPNO        ENAME         MGR           SAL            HIREDATE          SAL_SUM
7369        SMITH        7902         800.00        1980-12-17                800
7499        ALLEN        7698        1600.00        1981-2-20            2850
7521        WARD        7698        1250.00        1981-2-22            2850
7566        JONES        7839        2975.00        1981-4-2            5825
7698        BLAKE        7839        2850.00        1981-5-1            5825
7782        CLARK        7839        2450.00        1981-6-9            2450
7844        TURNER7698        1500.00        1981-9-8            2750
7654        MARTIN7698        1250.00        1981-9-28            2750
7839        KING                    5000.00        1981-11-17                8950
7900        JAMES        7698        950.00            1981-12-3            8950
7902        FORD        7566        3000.00        1981-12-3            8950
7934        MILLER        7782        1300.00        1982-1-23            1300
7788        SCOTT        7566        3000.00        1982-12-9            3000
7876        ADAMS        7788        1100.00        1983-1-12            1100

分组为整个查询结果集，按hiredate排序，窗口范围为hiredate between current_hiredate ? 30 and current_hiredate + 30。
在整个查询结果集中，分析函数计算与当前雇员日期相差在30天的雇员的工资总和，再将结果赋给当前雇员。
补齐pow形式，等价如下：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY NULL ORDER BY hiredate RANGE BETWEEN 30 preceding AND 30 following) sal_sum
  FROM emp;
6.8. E7(o)
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(ORDER BY hiredate) sal_sum
  FROM emp;

           EMPNO        ENAME         MGR          SAL            HIREDATE          SAL_SUM
7369        SMITH        7902        800.00            1980-12-17                800
7499        ALLEN        7698        1600.00        1981-2-20            2400
7521        WARD        7698        1250.00        1981-2-22            3650
7566        JONES        7839        2975.00        1981-4-2            6625
7698        BLAKE        7839        2850.00        1981-5-1            9475
7782        CLARK        7839        2450.00        1981-6-9            11925
7844        TURNER7698        1500.00        1981-9-8            13425
7654        MARTIN7698        1250.00        1981-9-28            14675
7839        KING                    5000.00        1981-11-17                19675
7900        JAMES        7698        950.00            1981-12-3            23625
7902        FORD        7566        3000.00        1981-12-3            23625
7934        MILLER        7782        1300.00        1982-1-23            24925
7788        SCOTT        7566        3000.00        1982-12-9            27925
7876        ADAMS        7788        1100.00        1983-1-12            29025

分组为整个查询结果集，按hiredate排序，窗口默认为range between unbounded preceding and current row。
在整个查询结果集中，分析函数计算当前雇员雇佣日期不晚于自己的雇员薪水总和，再将其赋给当前雇员。
补齐pow形式，等价于：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY NULL ORDER BY hiredate RANGE BETWEEN unbounded preceding AND CURRENT ROW) sal_sum
  FROM emp;
6.9. E8()
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over() sal_sum
  FROM emp;

           EMPNO        ENAME          MGR          SAL             HIREDATE           SAL_SUM
7369        SMITH        7902        800.00            1980-12-17                29025
7499        ALLEN        7698        1600.00        1981-2-20            29025
7521        WARD        7698        1250.00        1981-2-22            29025
7566        JONES        7839        2975.00        1981-4-2            29025
7654        MARTIN7698        1250.00        1981-9-28            29025
7698        BLAKE        7839        2850.00        1981-5-1            29025
7782        CLARK        7839        2450.00        1981-6-9            29025
7788        SCOTT        7566        3000.00        1982-12-9            29025
7839        KING                    5000.00        1981-11-17                29025
7844        TURNER7698        1500.00        1981-9-8            29025
7876        ADAMS        7788        1100.00        1983-1-12            29025
7900        JAMES        7698        950.00            1981-12-3            29025
7902        FORD        7566        3000.00        1981-12-3            29025
7934        MILLER        7782        1300.00        1982-1-23            29025

分组为整个查询结果集，不排序，没有规定窗口。
这是个reporting function。分析函数计算整个查询结果集中雇员薪水总和，再将其值赋给每一雇员。
补齐pow形式等价于：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY NULL ORDER BY NULL RANGE BETWEEN unbounded preceding AND unbounded following) sal_sum
  FROM emp;
续上，下面是一些range窗口的例子

7. Range窗口详例
假设分组第一行值为first_value，最后一行的值last_value，当前行的值为current_value。
7.1. Range between unbounded preceding and unbounded following
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       first_value(hiredate) OVER(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following) first_value,
       last_value(hiredate) OVER(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following) last_value,     
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND unbounded following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       last_value(hiredate) OVER(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND unbounded following) first_value,
       first_value(hiredate) OVER(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND unbounded following) last_value,     
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND unbounded following) sal_sum
  FROM emp;

相当于
hiredate BETWEEN first_value(hiredate) AND last_value(hiredate) --asc
hiredate BETWEEN last_value(hiredate) AND first_value(hiredate)--desc
7.2. Range [between] unbounded preceding [and current row]
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND CURRENT ROW) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       first_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following) first_value,
       hiredate current_value,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND CURRENT ROW) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND CURRENT ROW) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate current_value,
       first_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND unbounded following) first_value,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND CURRENT ROW) sal_sum
  FROM emp;

相当于
hiredate BETWEEN first_value(hiredate) AND current_value(hiredate) --asc
hiredate BETWEEN current_value(hiredate) AND first_value(hiredate) --desc

多个排序键值的例子：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY mgr, hiredate DESC, sal RANGE BETWEEN unbounded preceding AND CURRENT ROW) sal_sum
  FROM emp
ORDER BY mgr, hiredate DESC, sal;
等价于：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY mgr, hiredate DESC, sal rows BETWEEN unbounded preceding AND CURRENT ROW) sal_sum
  FROM emp
ORDER BY mgr, hiredate DESC, sal;
7.3. Range between unbounded preceding and value_expr preceding
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND 20 preceding) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       first_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following) first_value,
       hiredate - 20,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND 20 preceding) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND 20 preceding) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate + 20,
       first_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND unbounded following) first_value,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND 20 preceding) sal_sum
  FROM emp;

相当于
hiredate BETWEEN first_value(hiredate) AND current_value(hiredate) -20 --asc
hiredate BETWEEN current_value(hiredate)+20 AND first_value(hiredate) --desc
7.4. Range between unbounded preceding and value_expr following
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND 30 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       first_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following) first_value,
       hiredate + 30,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND 30 following) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND 30 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate - 30,
       first_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND unbounded following) first_value,     
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND 30 following) sal_sum
  FROM emp;
相当于
hiredate BETWEEN first_value(hiredate) AND current_value(hiredate) +30 --asc
hiredate BETWEEN current_value(hiredate)-30 AND first_value(hiredate)  --desc
7.5. Range between current row and unbounded following
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN CURRENT ROW AND unbounded following) sal_sum
  FROM emp;

-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate current_value,
       last_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following) last_value,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN CURRENT ROW AND unbounded following) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN CURRENT ROW AND unbounded following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       last_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND unbounded following) last_value,     
       hiredate current_value,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN CURRENT ROW AND unbounded following) sal_sum
  FROM emp;

相当于
hiredate BETWEEN current_value(hiredate) AND last_value(hiredate) --asc
hiredate BETWEEN last_value(hiredate) AND current_value(hiredate) --desc

多个排序键值的例子：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY mgr, hiredate DESC, sal RANGE BETWEEN CURRENT ROW AND unbounded following) sal_sum
  FROM emp
ORDER BY mgr, hiredate DESC, sal;
等价于：
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY mgr, hiredate DESC, sal rows BETWEEN CURRENT ROW AND unbounded following) sal_sum
  FROM emp
ORDER BY mgr, hiredate DESC, sal;
7.6. Range [between current row and] current row
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE CURRENT ROW) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate current_date,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE CURRENT ROW) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE  CURRENT ROW) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate current_date,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE  CURRENT ROW) sal_sum
  FROM emp;

相当于
hiredate = current_value(hiredate) --asc/desc
7.7. Range between current row and value_expr following
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN CURRENT ROW AND 20 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate current_value,
       hiredate + 20 "CURRENT_VALUE + 20",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN CURRENT ROW AND 20 following) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN CURRENT ROW AND 20 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate - 20 "CURRENT_VALUE - 20",
       hiredate current_value,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN CURRENT ROW AND 20 following) sal_sum
  FROM emp;

相当于
hiredate BETWEEN current_value(hiredate) AND current_value(hiredate)+20 --asc
hiredate BETWEEN current_value(hiredate)-20 AND current_value(hiredate)
--desc
7.8. Range between value_expr preceding and unbounded following
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 50 preceding AND 20 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate - 50 "CURRENT_VALUE - 50",
       hiredate + 20 "CURRENT_VALUE + 20",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 50 preceding AND 20 following) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 50 preceding AND 20 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate - 20 "CURRENT_VALUE - 20",
       hiredate + 50 "CURRENT_VALUE + 50",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 50 preceding AND 20 following) sal_sum
  FROM emp;

相当于
hiredate BETWEEN current_value(hiredate)-50 AND current_value(hiredate)+20 --asc
hiredate BETWEEN current_value(hiredate)-20 AND current_value(hiredate)+50 --desc
7.9. Range [between value_expr] preceding [and current row]
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 20 preceding AND CURRENT ROW) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate - 20 "CURRENT_VALUE - 20",
       hiredate CURRENT_VALUE,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 20 preceding AND CURRENT ROW) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 20 preceding AND CURRENT ROW) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate CURRENT_VALUE,
       hiredate + 20 "CURRENT_VALUE + 20",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 20 preceding AND CURRENT ROW) sal_sum
  FROM emp;

相当于
hiredate BETWEEN current_value(hiredate)-20 AND current_value(hiredate) --asc
hiredate BETWEEN current_value(hiredate) AND current_value(hiredate)+20
--desc
7.10. Range between value_expr1 preceding and value_expr2 preceding
这里一定要满足value_expr1>=value_expr2。
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 80 preceding AND 20 preceding) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate - 80 "CURRENT_VALUE - 80",
       hiredate - 20 "CURRENT_VALUE - 20",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 80 preceding AND 20 preceding) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 80 preceding AND 20 preceding) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate + 20 "CURRENT_VALUE + 20",
       hiredate + 80 "CURRENT_VALUE + 80",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 80 preceding AND 20 preceding) sal_sum
  FROM emp;

相当于
hiredate BETWEEN current_value(hiredate)-80 AND current_value(hiredate) -20  --asc
hiredate BETWEEN current_value(hiredate)+20 AND current_value(hiredate) +80  --desc
7.11. Range between value_expr1 preceding and value_expr2 following
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 20 preceding AND 10 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate - 20 "CURRENT_VALUE - 20",
       hiredate + 10 "CURRENT_VALUE + 10",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 20 preceding AND 10 following) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 20 preceding AND 10 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate + 20 "CURRENT_VALUE + 20",
       hiredate - 10 "CURRENT_VALUE - 10",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 20 preceding AND 10 following) sal_sum
  FROM emp;

相当于
hiredate BETWEEN current_value(hiredate)-20 AND current_value(hiredate) +10 --asc
hiredate BETWEEN current_value(hiredate)+10 AND current_value(hiredate) -20 --desc
7.12. Range between value_expr following and unbounded following
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 20 following AND unbounded following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate + 20 "CURRENT_VALUE + 20",
       last_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN unbounded preceding AND unbounded following) last_value,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 20 following AND unbounded following) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 20 following AND unbounded following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       last_value(hiredate) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN unbounded preceding AND unbounded following) last_value,
       hiredate - 20 "CURRENT_VALUE - 20",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 20 following AND unbounded following) sal_sum
  FROM emp;

相当于
hiredate BETWEEN current_value(hiredate)+20 AND last_value(hiredate)             
hiredate BETWEEN last_value(hiredate) AND current_value(hiredate) - 20 
7.13. Range between value_expr1 following and value_expr2 following
--asc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 20 following AND 90 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate + 20 "CURRENT_VALUE + 20",
       hiredate + 90 "CURRENT_VALUE + 90",
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate RANGE BETWEEN 20 following AND 90 following) sal_sum
  FROM emp;
--desc
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 20 following AND 90 following) sal_sum
  FROM emp;
-->
SELECT empno,
       ename,
       mgr,
       sal,
       hiredate,
       hiredate - 90 "CURRENT_VALUE - 90",
       hiredate - 20 "CURRENT_VALUE - 20",     
       SUM(sal) over(PARTITION BY mgr ORDER BY hiredate DESC RANGE BETWEEN 20 following AND 90 following) sal_sum
  FROM emp;
相当于
hiredate BETWEEN current_value(hiredate)+20 AND current_value(hiredate)+90 --asc
hiredate BETWEEN current_value(hiredate)-90 AND current_value(hiredate)-20 --desc