Date: Thu, 10 May 2007 20:13:14 +0000
Reply-To: toby dunn <tobydunn@HOTMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: toby dunn <tobydunn@HOTMAIL.COM>
Subject: Re: 'Local' datasets in Macro?
In-Reply-To: <051020071959.6431.464379B6000614140000191F220076106405029A06CE9907@comcast.net>
Content-Type: text/plain; format=flowed
Ian ,
Nope not screaming yet the number is large enough to avoid a name collision
problem unless one is doing something ungodly that they would need to stop
and rethink their abilities as a programmer.
I did have one minor thingy you missed a semi colon when you assiged Macro
variable r lol....
%macro q (len=14) ;
%local r ;
%let r = %sysevalf(10**14 * %sysfunc(ranuni(444))) ;
%let len = %sysfunc(min(&len,%length(&r))) ;
%substr(&r,1,&len)
%mend ;
Toby Dunn
On the other hand, you have different fingers. ~ LCG
The early bird may get the worm, but the second mouse gets the cheese in the
trap. ~ LCG
What happens if you get scared half to death, twice? ~ LCG
From: Ian Whitlock <iw1junk@COMCAST.NET>
Reply-To: iw1junk@COMCAST.NET
To: SAS-L@LISTSERV.UGA.EDU
Subject: Re: 'Local' datasets in Macro?
Date: Thu, 10 May 2007 19:59:50 +0000
Summary: DATAnn convention does not guarantee unique names.
#iw-value=1
In response to the question how to guarantee that work data set
names are unique. Don Henderson and others have suggested that
the names be left to the default SAS names DATAnn. SAS guarantees
that it will not override it's own default named data sets, not that it
won't clobber your data sets.
Consider the macro
%macro q ;
data ;
y = 1 ;
run ;
%mend q ;
And a consumer who sets up some data
data data2 ;
x = "my data" ;
run ;
And has the misfortune to call for Q's services twice.
%q
%q
DATA2 with the variable X is gone!
The problem is really a more general one, SAS does not have a concept of
name spaces. Hence most named things can be clobbered by the careless
use of names. In fact the whole problem has nothing to do with macro, but
rather large programs which often contain macros.
Long ago on SAS-L Paul Dorfman suggest a random tag for macro variable
names based on RANUNI. Consider
%macro q (len=14) ;
%local r ;
%let r = %sysevalf(10**14 * %sysfunc(ranuni(444)))
%let len = %sysfunc(min(&len,%length(&r))) ;
%substr(&r,1,&len)
%mend ;
The first time Q is called a seed based on the clock time is created.
There after SAS manages the seed so one is guaranteed a "random" stream of
numbers with successive calls to Q. Hence one might write
%let dsid = %q() ;
data w&dsid ;
...
run ;
and PROBABLY have a unique name. (I can hear Toby screaming, "NO that is
not perfect." If only he would write the program to test how many data set
names must be created on average to get a collision of names, and then run
it until a collision is realized, we could get on with some greater
worries.)
Here is the log for creating a subdirectory of the work directory and then
putting a simple data set in it. Unfortunately the work will *NOT* be
wiped out when the SAS session is closed, probably due to containing a
subdirectory.
179 %let unq = %q() ;
180 %let path = %sysfunc(pathname(work))\&unq ;
181 %***put path=&path ;
182 %let rc = %sysfunc(system( md "&path" )) ;
183 %let ulib = %q(len=7) ;
184 libname q&ulib "&path" ;
NOTE: Libref Q2671566 was successfully assigned as follows:
Engine: V9
Physical Name: C:\DOCUME~1\IANWHI~1\LOCALS~1\Temp\SAS Temporary
Files\_TD1028\57702832556191
185
186 data q&ulib..w ; x = 1 ; run ;
NOTE: The data set Q2671566.W has 1 observations and 1 variables.
Ian Whitlock
_________________________________________________________________
Catch suspicious messages before you open them—with Windows Live Hotmail.
http://imagine-windowslive.com/hotmail/?locale=en-us&ocid=TXT_TAGHM_migration_HM_mini_protection_0507
