Query to collect data from previous rows - sql-server

Question

I have a table with records as, in example data below a CO.Nr are TH-123,Th-456 and so on... I need to collect the data..
Nr. CO.Nr Employee Resp Description Date
1 TH-123 ABC NULL HELLO 10.05.2010
2 TH-123 NULL S14 NULL 18.05.2010
3 TH-123 DEF NULL 13.05.2010
4 TH-456 XYZ NULL NULL 1.07.2010
5 TH-456 NULL S19 SOME NULL
6 TH-456 TEXT 08.05.2010
7 TH-456 NULL 28.05.2010
For TH-123,
If Nr. is maximum, that is the record i need to start with group by CO.Nr, so it is the record with Nr as 3,
if the value in the other columns is NULL or space, go to a record above that is the record with Nr as 2, even if it has null value go to a record above that record with Nr. as 1 in this case.
In the 3 records i need to take the maximum of date.
For the above data, i need to have output as,
CO.Nr Employee Resp Description Date
TH-123 DEF S14 HELLO 18.05.2010
TH-456 XYZ S19 TEXT 01.07.2010
Thanks in advance!

Answer 1

You can do it many ways
select [co.nr],
(select top(1) employee from mytable b where b.[co.nr]=a.[co.nr] and
employee is not null order by nr desc) as employee,
(select top(1) resp from mytable b where b.[co.nr]=a.[co.nr] and
resp is not null order by nr desc) as resp,
(select top(1) description from mytable b where b.[co.nr]=a.[co.nr] and
description is not null order by nr desc) as description,
(select max([date]) from mytable b where b.[co.nr]=a.[co.nr]) as Date
from (
select distinct [co.nr]
from mytable ) as a

Answer 2

You can use a subselect to choose the record you want, then join on that. Something like the following for the employees one (I'll leave the rest of the columns as an exercise):
SELECT MyTable.[CO.Nr], Employees.Employee
FROM MyTable
LEFT OUTER JOIN (SELECT FIRST(Employee) as Employee, [CO.Nr]
FROM MyTable
WHERE Employee IS NOT NULL AND Employee <> ''
GROUP BY [CO.Nr]
ORDER BY [Nr.] DESC) Employees
ON MyTable.[CO.Nr] = Employees.[CO.Nr]
GROUP BY MyTable.[CO.Nr]
Or, if FIRST() is not a valid aggregate function, as mentioned in your comments, you can try subselects in your SELECT clause, like:
SELECT t.MyTable.[CO.Nr],
(SELECT TOP(1) x.Employee
FROM MyTable x
WHERE x.[CO.Nr] = t.[CO.Nr]
AND x.Employee IS NOT NULL AND x.Employee <> ''
ORDER BY [Nr.] DESC) as Employee
FROM MyTable t
GROUP BY t.[CO.Nr]