Archives of the TeradataForum

Message Posted: Tue, 01 Dec 2009 @ 20:47:03 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Hi David,

Dieters solution will be the fastes.

It should be also possible if you can use a derived table. Will also work if your group definition is not numeric :-)...

     select t.*.sum(week_change_id) over (order by date rows unbounded preceding and current
     row) as running_week ( select a.*, case when sum(wwtY) over (order by date rows 1
     preceding and 1 preceding) <>sum(wwtY) over (order by date rows current row and current
     row) then 1 else 0 end) as week_change_id from yourdata
     ) as t

I have no access to a db to check the syntax but I guess this should to the calculation....

Ulrich

	<Prev	Next>		<<First	<Prev	Next>	Last>>