Heh... "What works for them all the time", could break tomorrow...   Then, a new "Localized Best Practice" will be born getting them closer...

I pull more than a half-million rows from Oracle to SQL using a linked server and an INSERT/SELECT in less than 2 minutes... sure you want to stick with DTS?

P.S.  I'll kindly with-hold the lecture about making such mixed character sequences

As much as I love number (Tally) tables, there's absolutely no need for one here... nore is there the need for a bunch of functions, etc.  I don't remember where...

Heh... I agree... but that's not the debate... "It Depends" is not the debate... the debate is "What is the best practice?" and it also seems to be coming "What the...

Heh... I knew it ...  the term "Best Practice" is a bloody Management term ... AND, it...

Like Jackie Chan's Uncle would say "Oh! One more thing"...

I keep a "Tally" table of 11,000 numbers as a permanent table... good for more than 30 years of...

Oh yeah... almost forgot... if you want a nice clustered primary key on both, the test and test results are as follows:

--===== Setup the test environment

    SET NOCOUNT...

If you always want the numbers table to start at 1, this demonstrates both an alternative to the function and a comparison of run times from 10 rows to 10 million rows...

Uh-huh... so... whatcha gonna do with values that could look like this?

SELECT ISNUMERIC('13e05')

SELECT ISNUMERIC('13d05')

Got part way there using a limited triangular join... it spawned 2 times as many rows as the original count of both tables and that's with just a couple of dates... I...

Eugene... wanna post some data from the file, please?

Guess you can't actually look into a book on Amazon anymore, huh

Ummm.... not quite that simple... need to distribute a batch of QTY 15 issues against 4 batches of QTY 5 each and do the calculation and forward the remainder.  Lot's...

That's why the stored proc method is better.