分组统计SQL

Itpub上遇到一个求助写SQL的帖子，感觉很有意思，于是写出来看看，要求如下：

有个计划表1， 记录物料的年度计划量
有个实际使用情况表2，记录实际使用情况。
最后要出个统计表，把计划和实际的数据结合到一起进行统计，
有特殊的情况就是可能部分物料有计划但是没有实际使用，有的物料无计划但是有实际使用情况，
要把这些数据合并到一起按照季度进行统计并算差异量，能实现吗？？

于是建表插数据，表结构如下：

create table plan(OBJ varchar2(10),YEAR number(4),PLAN_COUNT number(10));
create table actual(OBJ varchar2(10),YEAR number(4),SEASON number(10),USED number(10));

语句如下：

with t1 as (
select distinct a.obj,p.plan_count,a.year,
case when a.season=1 then sum(a.used) over (partition by a.obj,a.season) else null end as s1,
case when a.season=2 then sum(a.used) over (partition by a.obj,a.season) else null end as s2,
case when a.season=3 then sum(a.used) over (partition by a.obj,a.season) else null end as s3,
case when a.season=4 then sum(a.used) over (partition by a.obj,a.season) else null end as s4,
sum(a.used) over (partition by a.obj) as sum_count 
from actual a left join plan p on a.obj=p.obj order by 1,4
)
select temp.obj,nvl(temp.plan_count,0),temp.year,temp.q1,temp.q2,temp.q3,temp.q4,nvl(temp.sum_count,0),(nvl(temp.plan_count,0)-nvl(temp.sum_count,0)) as diff 
from(
select distinct t1.obj,t1.plan_count,t1.year,
sum(s1) over (partition by t1.obj) as Q1,
sum(s2) over (partition by t1.obj) as Q2,
sum(s3) over (partition by t1.obj) as Q3,
sum(s4) over (partition by t1.obj) as Q4,t1.sum_count from t1
union all
select pl.obj,pl.plan_count,pl.year,null,null,null,null,null from plan pl where pl.obj not in (select distinct obj from actual)
) temp 
order by 1

用到了nvl,case when,not in,over partition,sum,union,left join,with as等写法，常年写SQL的应该有更加效率的写法，有待发现之后补充到此进行完善。

发现一个更简明的写法：

select 
distinct case when a.obj is not null then a.obj else p.obj end as obj,  
nvl(p.plan_count,0) as plan_count,
sum(case when a.season=1 then a.used end) as Q1,
sum(case when a.season=2 then a.used end) as Q2,
sum(case when a.season=3 then a.used end) as Q3,
sum(case when a.season=4 then a.used end) as Q4,
sum(nvl(a.used,0)) as sum_season,
nvl(p.plan_count,0)-sum(nvl(a.used,0)) as diff 
from plan p full join actual a on p.obj=a.obj
group by case when a.obj is not null then a.obj else p.obj end,nvl(p.plan_count,0)
order by 1

 

转载于:https://www.cnblogs.com/leohahah/p/6914300.html