考虑具有以下架构的表：
id, location, starred
有许多具有相同位置的记录：
id | location | starred
-----------------------
1 rome yes
2 rome no
3 rome no
4 milan yes
5 milan no
6 bozen no
我希望每个位置最多有一个记录。考虑到加星记录和未加星记录之间的选择，我想要加星。那么什么sql会产生这个表：
id | location | starred
-----------------------
1 rome yes
4 milan yes
6 bozen no
我怀疑这可以通过一些虚拟表或“视图”来完成。
DELETE FROM table
GROUP BY location,
使用分析函数删除重复项。以下代码生成基于 row_number 的位置，并按星号 desc 排序（所以先是）
delete from mytable2 where id in (
select id from
( select id, location,starred,row_number() over ( partition by location order by location, starred desc) row_num
from mytable2
) where row_num >1
)
如果 [started] 只能是是或否，那么这应该有效：
create table data
(
id int identity(1,1),
location varchar(50),
[started] varchar(3)
)
insert into data select 'Rome', 'Yes'
insert into data select 'Rome', 'No'
insert into data select 'Rome', 'No'
insert into data select 'Milan', 'Yes'
insert into data select 'Milan', 'No'
insert into data select 'Bozen', 'No'
WITH locationsRanked (id, location, [started], rank)
AS
(
select min(Id), location, [started],
RANK() OVER (PARTITION BY location ORDER BY location, [started] DESC) AS Rank
from data
group by location, [started]
)
select * from locationsRanked where Rank = 1
order by id
如果你只是想提取数据，这样的东西应该可以工作：
select
[table].*
from
[table]
inner join (select
MIN(id) as id,
location
from
[table]
group by location) as data
on [table].id = data.id
显然，您也可以使用此（或类似查询）的结果来确定要删除的 id 列表。