SUMSEN

Oracle&Sql爱好者,用友NC管理员

  博客园 :: 首页 :: 博问 :: 闪存 :: 新随笔 :: 联系 :: 订阅 订阅 :: 管理 ::

oracle爱好者和群snowg的问题

上面的这个,有站点stationid,year,month,day和每天记录的day_tmin字段。
现在要求统计处每个stationid下面每月每日的最小day_tmin字段,因为不关注year,所以sql这样写

select   stationid,  month,day,min(day_tmin)  tmin
from history.history_day_tem_new  
group by stationid month,day

然后要求在上述结果行增加年,需要知道是那一年的这个月日的最低温度.

select aa.stationid,aa.year,aa.month,aa.day,bb.tmin from
(select   stationid,yearmonth,day,day_tmin
from history.history_day_tem_new )aa,

(select   stationid,  month,day,min(day_tmin)  tmin
from history.history_day_tem_new  
group by stationid month,day ) bb
where aa.stationid=bb.stationid

and aa.month=bb.month
  and aa.day=bb.day
  and aa.day_tmin=bb.tmin
  order by  aa.stationid,aa.year,aa.month,aa.day

我aa表统计出所有的带有year的,然后和下面的min分组统计的进行表连接。
可是感觉复杂了,这是查询的结果。
求另外的写法,主要是这个写法感觉不靠谱,如果每天的统计有min(day_tmin)  有重复的最小就肯定不对了。

更新:我本来是使用partition

 

select distinct stationid, yearmin(day_tmin) over(partition by station order by month,day ) tmin
from history.history_day_tem_new order by stationid

经过仔细群里的辰影sql侠,他使用partition,首先建表(这个方法很好,下次不要create table了)

with w1 as(select '1' id,'2012' year,'12' month,'1' day,to_number('-10','999') tmin from dual
                union all
               select '1' id,'2012' year,'11' month,'1' day,to_number('-11','999') tmin from dual
               union all
               select '1' id,'2012' year,'10' month,'1' day,to_number('-12','999') tmin from dual
                union all
               select '1' id,'2012' year,'12' month,'1' day,to_number('-13','999') tmin from dual
               union all
                select '1' id,'2012' year,'11' month,'1' day,to_number('-6','999') tmin from dual
             union all
               select '1' id,'2012' year,'10' month,'1' day,to_number('-6','999') tmin from dual
              union all
              select '2' id,'2012' year,'12' month,'1' day,to_number('-4','999') tmin from dual
               union all
               select '2' id,'2011' year,'12' month,'1' day,to_number('-4','999') tmin from dual
               union all
          select '1' id,'2011' year,'12' month,'1' day,to_number('-20','999') tmin from dual)
select *  from   w1  order by id,month,day,tmin

之后的sql

重要总结:可以看出partition by的字段是需要统计分析(温度)的影响字段,id,month,day(因为对年不敏感,因此不增加),后面的order by(默认升序)字段是实际统计的的

partition by的字段为主语,或者说要分析的对象,然后order by是分析对象的熟悉(最大最小)

 select id,year,month,day,tmin
       from (select w1.*,dense_rank()over(partition by id,month,day order by tmin) rn from w1)
       where rn=1

再比如下面的sql:求各个公司最新月完成凭证情况表,order by 是最新月 year period,partition是unitname

select *
  from (select UNITCODE,
               unitname,
               year,
               period,
               num1,
               row_number() over(partition by UNITNAME order by YEAR desc,period desc) mm
          from (select BD_CORP.UNITCODE,
                       bd_corp.unitname,
                       gl_voucher.year,
                       gl_voucher.period,
                       count(*) num1
                  from gl_voucher, bd_corp
                 where bd_corp.pk_corp = gl_voucher.pk_corp
                      and gl_voucher.dr='0'
                 group by BD_CORP.UNITCODE,
                          bd_corp.unitname,
                          gl_voucher.year,
                          gl_voucher.period))

where mm = 1 ORDER BY UNITCODE;

再比如下面的这个sql:求出每个病人诊断好id最小下的疾病类别

 

select patiend_id,diagnosis_type,min(diagnosis_no)over(partition by patiend_id,diagnosis_type) diagnosis_no, diagnosis_class from (Select patient_id, visit_id, diagnosis_class from diagnostic_category a,DIAGNOSIS_DICT b
where a.diagnosis= b.diagnosis(+)
and diagnosis_class is not null and diagnosis_type='3'
)

 

这里因为min(diagnosis_no),因此其后面的partition后面就不需要order by了

 

rank()应该也是对的

ROW_NUMBER() 会去掉重复一行

用min partition 不行
select distinct id,year,month,day,tmin,min(tmin)over(partition by id,month,day order by tmin)   from w1        order by id,month,day,tmin

扩展一

2012-12-14 更新

按班级 查出 年龄最大的学生信息
学生表如下

 select * from (select xx.*,rank()over(partition by bj order by age desc) rn from xx) where rn=1
select s_no, name, age, bj
  from xx, (select bj || max(age) a from xx group by bj) yy--甲15 乙18
 where bj || age = a

平均工资的最大值(附加上ename):

select * from

( select ename,avgs,dense_rank()over( order by avgs desc)rn from

( select ename, avg(sal) avgs from emp group by ename))

where rn=1

扩展二

引用 Oracle:Rank,Dense_Rank,Row_Number比较

 

Oracle:Rank,Dense_Rank,Row_Number比较

一个员工信息表

Create Table EmployeeInfo (CODE Number(3) Not Null,EmployeeName varchar2(15),DepartmentID Number(3),Salary NUMBER(7,2),

Constraint PK_EmployeeInfo Primary Key (CODE));

Select * From EMPLOYEEINFO

 

现执行SQL语句:

Select EMPLOYEENAME,SALARY,

RANK() OVER (Order By SALARY Desc)  "RANK",

DENSE_RANK() OVER (Order By SALARY Desc ) "DENSE_RANK",

ROW_NUMBER() OVER(Order By SALARY Desc) "ROW_NUMBER"

 From EMPLOYEEINFO

结果如下:

 

Rank,Dense_rank,Row_number函数为每条记录产生一个从1开始至N的自然数,N的值可能小于等于记录的总数。这3个函数的唯一区别在于当碰到相同数据时的排名策略。

①ROW_NUMBER:

Row_number函数返回一个唯一的值,当碰到相同数据时,排名按照记录集中记录的顺序依次递增。

②DENSE_RANK:

Dense_rank函数返回一个唯一的值,除非当碰到相同数据时,此时所有相同数据的排名都是一样的。

③RANK:

Rank函数返回一个唯一的值,除非遇到相同的数据时,此时所有相同数据的排名是一样的,同时会在最后一条相同记录和下一条不同记录的排名之间空出排名。

同时也可以分组排序,也就是在Over从句内加入Partition by groupField:

 Select DEPARTMENTID,EMPLOYEENAME,SALARY,

RANK() OVER ( Partition By DEPARTMENTID Order By SALARY Desc)  "RANK",

DENSE_RANK() OVER ( Partition By DEPARTMENTID Order By SALARY Desc ) "DENSE_RANK",

ROW_NUMBER() OVER( Partition By DEPARTMENTID Order By SALARY Desc) "ROW_NUMBER"

 From EMPLOYEEINFO

结果如下:

 

现在如果插入一条工资为空的记录,那么执行上述语句,结果如下:

 

会发现空值的竟然排在了第一位,这显然不是想要的结果。解决的办法是在Over从句Order By后加上 NULLS Last即:

Select EMPLOYEENAME,SALARY,

RANK() OVER (Order By SALARY Desc  Nulls Last)  "RANK",

DENSE_RANK() OVER (Order By SALARY Desc Nulls Last) "DENSE_RANK",

ROW_NUMBER() OVER(Order By SALARY Desc Nulls Last ) "ROW_NUMBER"

 From EMPLOYEEINFO

结果如下:

 

 

 

 

posted on 2012-12-10 19:33  sumsen  阅读(2573)  评论(0编辑  收藏  举报