Date:         Tue, 19 Dec 2000 15:06:11 -0500
Reply-To:     Ian Whitlock <WHITLOI1@WESTAT.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Ian Whitlock <WHITLOI1@WESTAT.COM>
Subject:      Re: Question on usage of data conversion functions in a Data Step
              ;
Comments: To: Vivek Annamalai <annamalaiv@HOTMAIL.COM>
Content-Type: text/plain; charset="iso-8859-1"

Vivek,

The line

   if data_mon = . then data_mon = input(put(dm,6.),yymmn6.);

contains a subtle error.  Consider the first value of DATA_MON in TEST2.  It
is missing.
Thus the value is replaced according to the above.  Now it is not missing.
Since it comes from a SAS data set it is not reset to missing.  Since it is
not on TEST2 the value cannot be clobbered by reading a new record.  Hence
it will not change on any subsequent records.

Your work around

   if dm ^= . then data_mon = input(put(dm,6.),yymmn6.);

corrects the problem.  However, I think most good SAS programmers would
prefer to test with an IN= variable, since it makes clear the intent - for
TEST2 get the value of DATA_MON from DM.  With your code one has to think -
ah DM is on TEST2; thus it shouldn't be missing when reading TEST2; DATA_MON
is not on TEST2; oh, so he wants to populate DATA_MON on TEST2 with values
from DM.

Good coding consists of more than getting it right.  It also means getting
the message to any reader of your code as efficiently as possible.  It pays
because most often you will be the reader and when not you may have to
explain the code to someone else.

Ian Whitlock <whitloi1@westat.com>


-----Original Message-----
From: Vivek Annamalai [mailto:annamalaiv@HOTMAIL.COM]
Sent: Tuesday, December 19, 2000 2:25 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Re: Question on usage of data conversion functions in a Data
Step;


Thanks to Shiling, John Goldberg, Patrick and Ian

I totally forgot about numric variables remain right justified while they
are converted to characters using the put function. That explained the
mystery behind the usage of best. as against explicitly best6. or just 6.


The usage of in= dataset option helps circumvent the problem. But I haven't
got an answer as to why my original code doesn't produce the correct
results.

Even replacing

if data_mon = . then data_mon = input(put(dm,6.),yymmn6.);

to

if dm ^= . then data_mon = input(put(dm,6.),yymmn6.);

solves the problem,which I found after getting your responses. There is no
need for the in= data set option

The code without any changes produces the following result

id    data_mon

11     199909
12     200010
15     200105
21     200009
22     200009
23     200009
24     200009

After making the above change the result looks like:

id    data_mon

11     199909
12     200010
15     200105
21     200009
22     200007
23     200008
24       .


This begs the question on how the SAS supervisor is parsing the statements
in both the cases.