>>I need to upsize my application to SQL-server or Oracle, and I'm looking for advice.
>>
>>What is the best way to write a program that has to run through each record (or maybe a filtered subset) within a file, and do some processing and possibly updating of the record? Should I be selecting all the records into a cursor first, and then SCAN that, and send back updates one record at a time? Or maybe a cursor of keys only?
>>
>>With VFP data bases this is easy. What is the best (most efficient) approach in client-server?
>
>The most efficient way is to code the procedure in T-SQL. However, I have found it difficult so far to be able to do what I want in T-SQL. Experts say it can be done, but it isn't pretty. And debugging is much more difficult.
>
>I think getting all the records you need, processing them in VFP, then sending the updates back to the server is more efficient.

So you would create local updatable view, run the processing against that, and then tell VFP to send all the changes back at the end? If I have 1000 records updated, that's going to create one helluva SQL update statement to send back to the server. Shouldn't I get just the keys, and then do one record at a time?