Question

I’m a beginner in SQL Server 2008 and i have difficulty to find solution for the folowing problem:

I’d like to do a SELECT which eliminate successive rows having the same “stat” and keep only the last occurrence.

To make myself better understood, I give you this example:

My table                                                 Results 
idLigne | id | statut                                    idLigne | id | stat  
   L1     1      A                                          L2     1      A
   L2     1      A                                          L6     1      B
   L3     1      B                                          L7     1      A
   L4     1      B                    ====>                 L8     1      B
   L5     1      B
   L6     1      B
   L7     1      A
   L8     1      B

Thank you.

Answer 1

If you assign row numbers to result set you might connect “current” and “previous” record on row number and row number – 1; if statut is not the same row from t2 will be nulled out. Testing this will give us rows having no predecessors.

; with transformed as (
  select idLigne, 
         statut, 
         row_number () over (order by idLigne) rn
    from myTable
)
select myTable.*
  from myTable
 inner join transformed t1
    on myTable.idLigne = t1.idLigne
  left join transformed t2
    on t1.rn = t2.rn - 1
   and t1.statut = t2.statut
 where t2.rn is null

Demo is at Sql Fiddle.