Date: Tue, 18 Apr 2000 05:28:25 GMT
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Paul Dorfman <paul_dorfman@HOTMAIL.COM>
Subject: Re: Delete or update the date set: which one is faster? Windows
Content-Type: text/plain; format=flowed
If you recall the ultimate example of SAS inefficiency (dubbed 'going
nowhere' in Aster/Seidman SAS Bible),
data a; set a; run;
there is a valuable lesson to be extracted from this programming nonsense.
The computer does the following work in the presence of NOREPLACE (default)
option: 1) backs up the original of a; 2) creates a new version of a; 3) if
it is successful, erases the original.
Of course, you realize that if there is _no_ original, there is nothing to
back up - so some computer time gets saved right there. In the exaggerated
example above, the original could not be dumped as then there would be
nothing to read, but when one creates a dataset x reading a dataset y, it
can save significant computer time, especially it the dataset is sizable.
That .s02 thingie is just a backup copy. There are gobs of ways in SAS to
kill an unneeded dataset from a library; just off the top of the head,
proc delete data=lib.y; run;
proc datasets library=lib; delete y; run;
proc sql; drop table lib.y; quit;
Of course, you can delete it manually, too, but why bother if it can be
Paul M. Dorfman
>From: paula <icj808@USWEST.NET>
>Reply-To: paula <icj808@USWEST.NET>
>Subject: Delete or update the date set: which one is faster? Windows
> version 6.12
>Date: Mon, 17 Apr 2000 20:28:37 -0400
>The size of the SAS data sets I am dealing with increases dramatically day
>by day. One week ago when I was thrown at the SAS position, the average
>for a data set was 200MB. Thanks to all the helps I have got from this
>newsgroup, I found myself fiddling with SAS data sets of about 1.33 GB now
>Beside getting proficient helps from folks here and those at SAS institute,
>I heard rumors about SAS. One really stakes me here.
>I keep on revising codes in SAS programs. And then resubmit them until I
>everything right. I notice in the folder where the library name points to,
>when compiling is ongoing, there is a temporary file showing *.s02 with the
>same name as the data set I compiled from previous versions of the program.
>I heard that if I manually delete the preexisting version of the date set,
>SAS will run faster in creating the newer version, as compared with leaving
>alone the older version and let SAS automatically replace the older
>How true is this?
Get Your Private, Free Email at http://www.hotmail.com