Date:         Wed, 5 Jun 2002 14:28:17 -0400
Reply-To:     Ian Whitlock <WHITLOI1@WESTAT.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Ian Whitlock <WHITLOI1@WESTAT.COM>
Subject:      Re: retaining the order of variables
Comments: To: William Kossack <kossackw@NJC.ORG>
Content-Type: text/plain; charset="iso-8859-1"

William,

Suppose at some time you have data set A in the order you want it.  Then use

   data lib.Aord ;
      stop ;
      set A ;
   run ;

To capture this order without wasting space.  Now some time further on you
have a copy of A say Acopy.  Then use

   set lib.Aord Acopy ;

to reorder Acopy in the manner of Aord.  Of course if you have the original
copy of A then you can use

   set A ( obs = 0 ) Acopy ;

and forget the first step.  One advantage of this solution is that it not
only saves the order, it also preserves labels, format information, and
variable type and length.  If this is perceived as disadvantage then use
a RETAIN statement at the top of your DATA step.

The above answers the direct question asked.  But it may not be the best
answer to the problem.  Remember the problem was that you have groups of
variables and you want to preserve the groups.  Order is one way, but not
the only one.  It is probably better to set up your groups one with code
like

   %let grpA = a b c ;
   %let grpB = r s t u v w ;
   ...

Store the code in a file and include it in any programs that need to know
about these groups.  Now you don't need to worry about order and can do
anything you could do before with order.  Moreover, anyone can know how you
see the groupings and therefore have a better understanding of how to
maintain your code.  Another big advantage is that the initial files do not
have to hold a whole group.  For example, A could come from one file while B
and C come from another.  After merging I can still think about using &GRPA.



For example, suppose the problem is to produce cross tabs of all variables
in grpA with those in grpB.  Then the tables statement could be

    table ( &grpA ) * ( &grpB ) ;

while using order one would have

    table ( a--c ) * ( r--w ) ;

Now that one has long names for variables another solution would be to name
the variables with a prefix indicating the group, GRPA_ or GRPB_.  In this
case there is no include file to consider and the table statement is

    table ( GRPA: ) * ( GRPB: ) ;

Note that with wise choices for the group names the first and third TABLE
statements make very clear what is going on, while there is no clue in
second TABLE statement unless the variable names happen to be more wisely
chosen than
usual.

IanWhitlock@westat.com

-----Original Message-----
From: William Kossack [mailto:kossackw@NJC.ORG]
Sent: Wednesday, June 05, 2002 12:30 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: retaining the order of variables


I am doing manipulations of a large number of variables using merge
statements and processing block of similar variables together.

I've discovered that I have reordered my variables which I don't like.

Is there a way of capturing the variable order of variables in a dataset
and forcing the dataset to maintain that order?