SQL分组取每组前几条记录，剩下的列为其他

/****** SQLserver 写法  ******/
SELECT  *
FROM [lx].[dbo].[全国销售额] where province='北京' --group by  [province],[name] 

--按销售额排序 取每个省的前4名,剩下的为其他
select * from (
select a.province,a.name ,a.value from 全国销售额 a where value in (select top 4 value from 全国销售额 where province=a.province order by value desc) 
union all
select a.province,'其他',sum(a.value) value from 全国销售额 a where value not in (select top 4 value from 全国销售额 where province=a.province order by value desc) 
group by a.province
) a order by a.province,a.value desc

--oracle 写法（大于5的列到其他）
SELECT a.bm,'其他' mdd ,a.jssj,sum(b.jp) jp,sum(b.hc) hc ,sum(b.bzfy) bzfy,sum(b.zsfy) zsfy,sum(b.jtqt) jtqt,sum(b.qt) qt,sum(b.jtgj) jtgj,
sum(b.bmj) bmj,sum(b.ygj) ygj ,sum(b.gsj) gsj,sum(b.zjj) zjj,sum(b.ccts) ccts ,sum(b.shje) shje  FROM  
(
        SELECT a.bm,a.mdd,a.jssj,jtgj, ROW_NUMBER() OVER(PARTITION BY a.bm,a.jssj
        ORDER BY jtgj desc) AS RN FROM NSC_OA_CLHZ a
        
) a left join NSC_OA_CLHZ b on nvl(a.bm,0) = nvl(b.bm,0) and nvl(a.mdd,0) = nvl(b.mdd,0) and nvl(a.jssj,0)=nvl(b.jssj,0)
and  a.RN >5
group by  a.bm,a.jssj