for me it is nearly clear, or let's say most of the time clear, what a data-step does. For some sql's it is not really clear what happens. Also one little change in the sql code and some intermediate files "explode", or runtime is 10 times longer than before, or the results are not as expected.

However for me it is necessary sometimes to use sql when I have to do sophisticated joins. I will not say that they are not possible with a data step, but one must be very careful and sometimes it seems to be very hard to get what's needed. In sql that is sometime quite easy to do. You simply "say" what's to be done and sql does. Nearly all other things are for me more easy to understand in a data-step and also much more easy to write. Some data steps are not easy to understand, but some of them I see clear after a while. With sql I don't always understand what happens. I see the results and I can see if they are correct, but the way might stay kind of magic...

Gerhard

On Mon, 5 Jul 2010 16:01:59 -0700, Arthur Tabachneck <art297@NETSCAPE.NET> wrote:

>I agree with Tom but, at the same time, slightly disagree. This topic >was just sort of addressed on the listserv side of SAS-l (see: >http://www.listserv.uga.edu/cgi-bin/wa?A2=ind1007a&L=sas-l&D=1&O=D&P=12516 >). > >Since my job requires me to manage SAS programmers and statisticians, >I'm concerned not only with what works best but, at the same time, >with what my staff understand. Sometimes that translates into what >must the staff learn but, at the same time, the question of whether >they can continue to do their job (while learning) is even more >important. > >Can we all be expected to know everything about SAS NOW? I don't >think that could ever be a reasonable proposition! > >Dependent upon the problem, such SAS modules, functions or >capabilities like datastep merges, enterprise miner, iml, sql, hash, >dow loops (and many, many others) may provide the optimal solution. >Which one should one use, though, I think is dependent upon other >factors (like how long will it take before everyone understands what >they are expected to know). > >Art >------------- >On Jul 2, 7:51 pm, Tom Abernathy <tom.aberna...@gmail.com> wrote: >> On Jul 2, 12:36 pm, member2010 <nagasri.naga...@gmail.com> wrote: >> >> > Hi All, >> >> > Can somebody explain me the advantages of using data step over proc >> > sql in SAS? >> >> > thank you >> >> It really depends what you want to do. >> You cannot create multiple output datasets, read data in from text >> files, write data to text files, build hash objects or use dozens of >> other features of the SAS programming language in PROC SQL.