|Date: ||Thu, 12 Nov 1998 19:50:19 +0000|
|Sender: ||"SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>|
|Subject: ||Comparison between macros and SCL for submitting base code|
|Content-Type: ||text/plain; charset=US-ASCII|
I'm looking for evidence, anecdotal or benchmarking on the relative merits of
using either SCL methods or macros for the most efficient way to pass lots of
parameters and base SAS code.
I've always thought, and experienced that the macro processor adds extra
resource utilisation to base SAS code, but when push comes to shove, I can't
recall where to go looking for hard cold facts.
I'm on a site where they are using many many uncompiled macros and voluminous
quantites of macro variables as a data extraction and build process for a
warehouse. The platform is IBM MBS and the performance is poor(currently days
to load this stuff, and
it's only in the order of hundreds of thousands of records, though it's
insurance data, and the merge process is very complex). The system resources
are unchangeable at this point in time, so I'm looking for coding improvements
to reduce CPU and IO.
The lashings of macros approach is an obvious candidate for improvement.
I've always presumed that scl lists and variables , and scl methods are less
resource intensive, because that's what I've noticed and that was the
information I got from SI technical support.
I have done a few tests, and notice that using an SCL method which is a rewrite
of a macro, and passes exactly the same base code takes less CPU and less
The developers on site are sceptical, being macro addicts, so if anyone else has
done some tests, or has some reference material, or merely some anecdotes, I'd
be grateful, particularly any info on resource comparison.
I'll be grateful to anyone who can give this their attention.
PC Analysis Services