在T-sql中查找具有不同值的重複項？

Question

我得到了從不同的源端系統找到重複的情況。

防爆：我有表如下圖所示：

declare @table table (id int,portnumber int, [sourceport] varchar(50), sourcereportedDate datetime) 
insert into @table values (1, 1111, 'north' , '2016-08-20 09:44:30.847') 
insert into @table values (2, 1111, 'north' , '2016-08-21 09:44:30.847') 
insert into @table values (3, 1111, 'north' , '2016-08-22 09:44:30.847') 

insert into @table values (4, 2222, 'north' , '2016-08-20 09:44:30.847') 
insert into @table values (5, 2222, 'north' , '2016-08-26 09:44:30.847') 
insert into @table values (6, 2222, 'south' , '2016-08-22 09:44:30.847') 

insert into @table values (7, 3333, 'south' , '2016-08-10 09:44:30.847') 
insert into @table values (8, 3333, 'north' , '2016-08-12 09:44:30.847') 

insert into @table values (9, 4444, 'north' , '2016-08-20 09:44:30.847') 

insert into @table values (10, 5555, 'south' , '2016-08-21 09:44:30.847') 
insert into @table values (11, 5555, 'south' , '2016-08-27 09:44:30.847') 

insert into @table values (12, 6666, 'south' , '2016-08-10 09:44:30.847') 
insert into @table values (13, 6666, 'north' , '2016-08-21 09:44:30.847') 
insert into @table values (14, 6666, 'south' , '2016-08-09 09:44:30.847') 

現在我想找到「portnumber」副本應該是相同的，「sourceport」應該是不同的。如果'portnumber'和'sourceport'相同，則不應該重複。 ，也是我需要additional column持有最大的「sourcereportedDate」日期

我想獲取輸出像下面的Id：

(4, 2222, 'north' , '2016-08-20 09:44:30.847',5) 
(5, 2222, 'north' , '2016-08-26 09:44:30.847','latest') 
(6, 2222, 'south' , '2016-08-22 09:44:30.847',5) 
(7, 3333, 'south' , '2016-08-10 09:44:30.847',8) 
(8, 3333, 'north' , '2016-08-12 09:44:30.847','latest') 
(12, 6666, 'south' , '2016-08-10 09:44:30.847',13) 
(13, 6666, 'north' , '2016-08-21 09:44:30.847','latest') 
(14, 6666, 'south' , '2016-08-09 09:44:30.847',13) 

在此先感謝。

Answer 1

請試試這個（我仍然認爲這可以進一步優化） -

;with DupWithMaxDate as (
    select 
     a.portnumber, 
     sourcereportedDate = max(a.sourcereportedDate) 
    from  @table a 
    left join @table b on b.portnumber = a.portnumber and b.sourceport <> a.sourceport 
    where  b.portnumber is not null 
    group by a.portnumber 
), 
DupWithMaxID as (
    select 
     a.portnumber, 
     max_id = a.id 
    from  DupWithMaxDate x 
    inner join @table a on a.portnumber = x.portnumber and a.sourcereportedDate = x.sourcereportedDate 
) 
select 
    a.id, 
    a.portnumber, 
    a.sourceport, 
    a.sourcereportedDate, 
    new_column = case when a.id = x.max_id then 'Latest' else convert(varchar(10), x.max_id) end 
from  DupWithMaxID x 
inner join @table a on a.portnumber = x.portnumber 

更新上面的查詢 -

;with DuplicateWithMaxID as (
    select 
     portnumber = a.portnumber, 
     max_id = a.id, 
     rank_id = row_number() over (partition by a.portnumber order by a.sourcereportedDate desc) 
    from  @table a 
    inner join @table b on b.portnumber = a.portnumber and b.sourceport <> a.sourceport 
) 
select 
    a.id, 
    a.portnumber, 
    a.sourceport, 
    a.sourcereportedDate, 
    new_column = case when a.id = x.max_id then 'Latest' else convert(varchar(10), x.max_id) end 
from 
    DuplicateWithMaxID x 
inner join @table a on a.portnumber = x.portnumber 
where 
    x.rank_id = 1 

Answer 2

試試這個：

; 
with 
dis as 
(
    select distinct portnumber, sourceport 
     from @table 
), 
dup as 
(
    select portnumber 
     from dis 
     group by portnumber 
     having count(1) > 1 
), 
mx as 
(
    select 
     dup.portnumber, 
     max(t.sourcereportedDate) as sourcereportedDate 
    from 
     dup 
    join 
     @table as t 
    on 
     t.portnumber = dup.portnumber 
    group by 
     dup.portnumber 
), 
mxi as 
(
    select 
     mx.portnumber, 
     t.id 
    from 
     mx 
    left join 
     @table as t 
    on 
     t.portnumber = mx.portnumber 
    and t.sourcereportedDate = mx.sourcereportedDate 
) 
select 
    t.id, 
    t.portnumber, 
    t.sourceport, 
    t.sourcereportedDate, 
    case when t.id = mxi.id 
     then 'latest' 
     else cast(mxi.id as varchar(10)) 
     end as latest 
from 
    dup 
join 
    @table as t 
on 
    t.portnumber = dup.portnumber 
join 
    mxi 
on 
    mxi.portnumber = t.portnumber 
left join 
    mx 
on 
    mx.portnumber = t.portnumber 
and mx.sourcereportedDate = t.sourcereportedDate