At 11:23 25/03/99 -0500, Bernard Tremblay wrote:

> You could use a view for the input of the sort and it would >achieve what your want or simply replace the DSN by an sql view including >the sort ...

Yes, but that's still an extra (DATA or SQL) step to write (to create the view).

I can but presume that any approach (even the extension to 'BY variables' I was proposing) would actually require a complete new copy of the dataset, with a 'new variable' introduced, to be created before the sorting actually started (deriving functions or evaluating expressions 'on the fly' during the sort would surely be very inefficient) - so I don't suppose there would be much to choose between the approaches in terms of performance, but the code would be quicker to write, and easier to read, if I could, for example, simply code: "PROC SORT ; BY UPCASE(var) ; run ;".

Regards

John

---------------------------------------------------------------- Dr John Whittington, Voice: +44 (0) 1296 730225 Mediscience Services Fax: +44 (0) 1296 738893 Twyford Manor, Twyford, E-mail: medisci@powernet.com Buckingham MK18 4EL, UK mediscience@compuserve.com ----------------------------------------------------------------