Oracle分析函数

Posted 2020-12-21 yifanSJ

一、分析函数语法
function_name(<argument>,<argument>...) over(<partition_Clause><order by_Clause><windowing_Clause>);
function_name()：函数名称
argument：参数
over( )：开窗函数
partition_Clause：分区子句，数据记录集分组，group by...
order by_Clause：排序子句，数据记录集排序，order by...
windowing_Clause：开窗子句，定义分析函数在操作行的集合，三种开窗方式：rows、range、Specifying
注：使用开窗子句时一定要有排序子句！！！
本篇未涉及开窗子句，开窗子句在另外的文章中单独说明

0、listagg () within group ()，将多行合并成一行
select listagg(t.ename, ‘,‘) within group (order by t.ename) as names from emp t group by t.deptno

1、count() over()  ：统计分区中各组的行数，partition by 可选，order by 可选
select ename,esex,eage,count(*) over() from emp; --总计数
select ename,esex,eage,count(*) over(order by eage) from emp; --递加计数
select ename,esex,eage,count(*) over(partition by esex) from emp; --分组计数
select ename,esex,eage,count(*) over(partition by esex order by eage) from emp;--分组递加计数

2、sum() over()  ：统计分区中记录的总和，partition by 可选，order by 可选
select ename,esex,eage,sum(salary) over() from emp; --总累计求和
select ename,esex,eage,sum(salary) over(order by eage) from emp; --递加累计求和
select ename,esex,eage,sum(salary) over(partition by esex) from emp; --分组累计求和
select ename,esex,eage,sum(salary) over(partition by esex order by eage) from emp; --分组递加累计求和

3、avg() over()  ：统计分区中记录的平均值，partition by 可选，order by 可选
select ename,esex,eage,avg(salary) over() from emp; --总平均值
select ename,esex,eage,avg(salary) over(order by eage) from emp; --递加求平均值
select ename,esex,eage,avg(salary) over(partition by esex) from emp; --分组求平均值
select ename,esex,eage,avg(salary) over(partition by esex order by eage) from emp; --分组递加求平均值

4. min() over() ：统计分区中记录的最小值，partition by 可选，order by 可选
max() over() ：统计分区中记录的最大值，partition by 可选，order by 可选
select ename,esex,eage,salary,min(salary) over() from emp; --求总最小值
select ename,esex,eage,salary,min(salary) over(order by eage) from emp; --递加求最小值
select ename,esex,eage,salary,min(salary) over(partition by esex) from emp; --分组求最小值
select ename,esex,eage,salary,min(salary) over(partition by esex order by eage) from emp; --分组递加求最小值

5、rank() over()  ：跳跃排序，partition by 可选，order by 必选
select ename,eage,rank() over(partition by job order by eage) from emp;
select ename,eage,rank() over(order by eage) from emp;

6、dense_rank() ：连续排序，partition by 可选，order by 必选

    select ename,eage,dense_rank() over(partition by job order by eage) from emp;
    select ename,eage,dense_rank() over(order by eage) from emp;

7、row_number() over() ：排序，无重复值，partition by 可选，order by 必选--表示根据col1分组，在分组内部根据col2排序
select ename,eage,row_number() over(partition by job order by eage) from emp;
select ename,eage,row_number() over(order by eage) from emp;

8、ntile(n) over() ：partition by 可选，order by 必选

    n表示将分区内记录平均分成n份，多出的按照顺序依次分给前面的组

    select ename,salary,ntile(3) over(order by salary desc) from emp;
    select ename,salary,ntile(3) over(partition by job order by salary desc) from emp;

9、first_value() over() ：取出分区中第一条记录的字段值，partition by 可选，order by 可选

     last_value() over() ：取出分区中最后一条记录的字段值，partition by 可选，order by 可选

    select ename,first_value(salary) over() from emp;
    select ename,first_value(salary) over(order by salary desc) from emp;
    select ename,first_value(salary) over(partition by job) from emp;                                                           
    select ename,first_value(salary) over(partition by job order by salary desc) from emp;

    select ename,last_value(ename) over() from emp;
    select ename,last_value(ename) over(order by salary desc) from emp;
    select ename,last_value(ename) over(partition by job) from emp;
    select ename,last_value(ename) over(partition by job order by salary desc) from emp;

10、first ：从DENSE_RANK返回的集合中取出排在最前面的一个值的行

      last ：从DENSE_RANK返回的集合中取出排在最后面的一个值的行

    select job,max(salary) keep(dense_rank first order by salary desc),
    max(salary) keep(dense_rank last order by salary desc) from emp
    group by job;

11、lag() over() ：取出前n行数据，partition by 可选，order by 必选

      lead() over() ：取出后n行数据，partition by 可选，order by 必选

    select ename,eage,lag(eage,1,0) over(order by salary),
    lead(eage,1,0) over(order by salary) from emp;
     
    select ename,eage,lag(eage,1) over(partition by esex order by salary),
    lead(eage,1) over(partition by esex order by salary) from emp;

12、ratio_to_report(a) over(partition by b) ：求按照b分组后a的值在所属分组中总值的占比,a的值必须为数值或数值型字段

      partition by 可选，order by 不可选

    select ename,job,salary,ratio_to_report(1) over() from emp; --给每一行赋值1，求当前行在总值的占比，总是0.1
    select ename,job,salary,ratio_to_report(salary) over() from emp; --当前行的值在所有数据中的占比
    select ename,job,salary,ratio_to_report(1) over(partition by job) from emp; --给每一行赋值1，求当前行在分组后的组内总值的占比
    select ename,job,salary,ratio_to_report(salary) over(partition by job) from emp; --当前行的值在分组后组内总值占比

13、percent_rank() over()  ：partition by 可选，order by 必选

     所在组排名序号-1除以该组所有的行数-1，排名跳跃排序

    select ename,job,salary,percent_rank() over(order by salary) from emp;
    select ename,job,salary,percent_rank() over(partition by job order by salary) from emp;

14、cume_dist() over() ：partition by 可选，order by必选

所在组排名序号除以该组所有的行数，注意对于重复行，计算时取重复行中的最后一行的位置

    select ename,job,salary,cume_dist() over(order by salary) from emp;
    select ename,job,salary,cume_dist() over(partition by job order by salary) from emp;

15、precentile_cont( x ) within group(order by ...) over()    ：over()中partition by可选，order by 不可选

x为输入的百分比，是0-1之间的一个小数，返回该百分比位置的数据，若没有则返回以下计算值（r）：

a=1+( x *(N-1) )  x为输入的百分比，N为分区内的记录的行数

b=ceil ( a )  向上取整

c = floor( a ) 向下取整

r=a * 百分比位置上一条数据 + b * 百分比位置下一条数据

    select ename,job,salary,percentile_cont(0.5) within group(order by salary) over() from emp;
    select ename,job,salary,percentile_cont(0.5) within group(order by salary) over(partition by job) from emp;

16、precentile_disc( x ) within group(order by ...) over()   ：over()中partition by可选，order by 不可选

x为输入的百分比，是0-1之间的一个小数，返回百分比位置对应位置上的数据值，若没有对应数据值，就取大于该分布值的下一个值

    select ename,job,salary,percentile_disc(0.5) within group(order by salary) over()from emp;
    select ename,job,salary,percentile_disc(0.5) within group(order by salary) over(partition by job) from emp;

17、stddev() over()：计算样本标准差，只有一行数据时返回0，partition by 可选，order by 可选

      stddev_samp() over()：计算样本标准差，只有一行数据时返回null，partition by 可选，order by 可选

      stddev_pop() over()：计算总体标准差，partition by 可选，order by 可选

    select stddev(stu_age) over() from student; --计算所有记录的样本标准差
    select stddev(stu_age) over(order by stu_age) from student; --计算递加的样本标准差
    select stddev(stu_age) over(partition by stu_major) from student; --计算分组的样本标准差
    select stddev(stu_age) over(partition by stu_major order by stu_age) from student; --计算分组递加的样本标准差
     
     
    select stddev_samp(stu_age) over() from student; --计算所有记录的样本标准差
    select stddev_samp(stu_age) over(order by stu_age) from student; --计算递加的样本标准差
    select stddev_samp(stu_age) over(partition by stu_major) from student; --计算分组的样本标准差
    select stddev_samp(stu_age) over(partition by stu_major order by stu_age) from student; --计算分组递加的样本标准差
     
     
    select stddev_pop(stu_age) over() from student; --计算所有记录的总体标准差
    select stddev_pop(stu_age) over(order by stu_age) from student; --计算递加的总体标准差
    select stddev_pop(stu_age) over(partition by stu_major) from student; --计算分组的总体标准差
    select stddev_pop(stu_age) over(partition by stu_major order by stu_age) from student;--计算分组递加的总体标准差

18、variance() over()：计算样本方差，只有一行数据时返回0，partition by 可选，order by 可选

       var_samp() over()：计算样本方差，只有一行数据时返回null，partition by 可选，order by 可选

       var_pop() over()：计算总体方差，partition by 可选，order by 可选

    select variance(stu_age) over() from student; --计算所有记录的样本方差
    select variance(stu_age) over(order by stu_age) from student; --计算递加的样本方差
    select variance(stu_age) over(partition by stu_major) from student; --计算分组的样本方差
    select variance(stu_age) over(partition by stu_major order by stu_age) from student; --计算分组递加的样本方差
     
     
    select var_samp(stu_age) over() from student; --计算所有记录的样本方差
    select var_samp(stu_age) over(order by stu_age) from student; --计算递加的样本方差
    select var_samp(stu_age) over(partition by stu_major) from student; --计算分组的样本方差
    select var_samp(stu_age) over(partition by stu_major order by stu_age) from student; --计算分组递加的样本方差
     
     
    select var_pop(stu_age) over() from student; --记录所有就的总体方差
    select var_pop(stu_age) over(order by stu_age) from student; --计算递加的总体方差
    select var_pop(stu_age) over(partition by stu_major) from student; --计算分组的总体方差
    select var_pop(stu_age) over(partition by stu_major order by stu_age) from student;--计算分组递加的样本方差

stddev()=sqrt( variance() )     sqrt()--求开方

stddev_samp()=sqrt( var_samp() )

stddec_pop=sqrt( var_pop() )


19、covar_samp over()：返回一对表达式的样本协方差，partition by 可选，order by 可选

       covar_pop over()： 返回一堆表达式的总体协方差，partition by 可选，order by 可选

    select covar_samp(stu_age,line) over() from student; --计算所有记录的样本协方差
    select covar_samp(stu_age,line) over(order by stu_age) from student; --计算递加的样本协方差
    select covar_samp(stu_age,line) over(partition by stu_major) from student; --计算分组的样本协方差
    select covar_samp(stu_age,line) over(partition by stu_major order by stu_age) from student; --计算分组递加的样本协方差
     
     
    select covar_pop(stu_age,line) over() from student; --计算所有记录的总体协方差
    select covar_pop(stu_age,line) over(order by stu_age) from student; --计算递加的总体协方差
    select covar_pop(stu_age,line) over(partition by stu_major) from student; --计算分组的总体协方差
    select covar_pop(stu_age,line) over(partition by stu_major order by stu_age) from student; --计算分组递加的总体协方差

20、corr() over() ：返回一对表达式的相关系数，partition by 可选，order by 可选

    select corr(stu_age,line) over() from student; --计算所有记录的相关系数
    select corr(stu_age,line) over(order by stu_age) from student; --计算递加的相关系数
    select corr(stu_age,line) over(partition by stu_major) from student; --计算分组的相关系数
    select corr(stu_age,line) over(partition by stu_major order by stu_age) from student; --计算分组递加的相关系数

21、REGR_ (Linear Regression) Functions：这些线性回归函数适合最小二乘法回归线，有9个不同的回归函数可使用