Question

(a “How it works” – question)

Suppose I have this table :

   Id            val
    ------------------
    1             a       <--
    1             a
    1             a
    2             b       <--
    2             b
    2             b

Now lets say I want the marked rows :

Can I count on this query :

select id,val from (
select id , val , row_number() over (partition by id order by id) as rn
) k where rn=1

to give me the Selected rows ?

(notice the order by clause). will it consider the order as the order they were inserted ?

Answer 1

The behaviour is not guaranteed. Your query will return two rows, but which ones is undefined.

When I experiment with

create table #t (id int, val char(1), t timestamp)
insert #t  (id, val) values (1,'a')
insert #t  (id, val) values (1,'a')
insert #t  (id, val) values (2,'b')

insert #t  (id, val) values (1,'a')
insert #t  (id, val) values (2,'b')
insert #t  (id, val) values (2,'b')
select * from #t 

select * from ( 
select *, row_number() over (partition by id order by id) as rn from #t  
) k where rn=1 


drop table #t

the results are variable, depending on the order of insertion, but not consistently the first inserted rows.