Heh... I started reading this post and then saw your first reply... I knew you'd use the running balance trick... it just makes sense, here. Nicely done!

I do have to ask the stupid question... you are running the code from the database that has the WRT_% tables in it, aren't you?

4 table idea sounds good to me. Could be 3 tables because a deposit is nothing more than a credit transaction. Payments are certainly different and should be...

Did you ever figure this one out?

Thanks for the explanation... why not just import the XML into a table as "real" data and use DateTime columns? Again, just curious...

Indexes don't interfere with the load if the table is empty at the beginning of the load.

Heh... wasn't meant to be personal, at all.

I'd make the startup size of TempDB whatever size it grew to after a month. 1 GB may not be enough... shoot,...

Thanks for the feedback, Mike. I've been involved with such undocumented legacy systems before and I know the hell you're going to go through. Just remember the old...

Not a problem... thanks for the feedback. 🙂

15 table join with a non-indexable WHERE clause and a killer ORDER BY and we wonder why this code takes a bit to run? First thing to do would...

I think an ALTER instead of a DROP would probably work faster.

Yes... the INSERT/EXEC would be used to save the results in a table. If you get different results, use more than 1 table... nothing wrong with that.

Of course, you...

Actually.... everything I've written about is NOT new... Tally table is old...

On the truncation problem, everyone realizes that the table columns must be in the same cardinal order as the input file unless you use a format file, right?

Do you mean "ISO Standard"?