Date:         Tue, 18 May 1999 12:37:42 -0400
Reply-To:     "Paul M. Dorfman" <sashole@EARTHLINK.NET>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         "Paul M. Dorfman" <sashole@EARTHLINK.NET>
Subject:      Re: [Warning] Retain statement to control variable's order  in
              dataset
Comments: To: JP <cruguel@dms.umontreal.ca>
Content-Type: text/plain; charset=us-ascii

JP wrote:
 
> I read once here that you could control the variable's order in the dataset
> by using a RETAIN statement.
> I tried it. It seems to work fine, as it orders variables the way I
> specifies it.
> But it creates errors, with missing values. From the RETAIN synthax, it
> seems that: If an obs has a missing value for a given var, it puts the
> value of the previous observation, for the given var.
> I remember someone proposing SQL to do this, with SELECT.
> This should work properly.
 
JP,
 
Could you give an example? Judging from this code snippet, the RETAIN statement
fails to exhibit the behavior you described:
 
DATA XY; INPUT X Y; LINES;
1 9
2 .
3 4
. 3
5 2
6 .
7 1
. 2
4 3
;
RUN;
DATA YX; RETAIN Y X; SET XY; RUN;
PROC PRINT; RUN;
 
OBS    Y    X
 
 1     9    1
 2     .    2
 3     4    3
 4     3    .
 5     2    5
 6     .    6
 7     1    7
 8     2    .
 9     3    4
 
And this:
 
PROC SQL; SELECT Y, X FROM XY; QUIT;
 
produces exactly the same output.
 
Kind regards,
 
++++++++++++++++
Paul M. Dorfman
Jacksonville, FL
++++++++++++++++