Subject: Macro clarification Summary: Classic macro vs DATA step array confusion is resolved, but formats provide the solution to real problem. Respondent: Ian Whitlock <whitloi1@westat.com>

Erik Godfrey <sg94x892@DUNX1.OCS.DREXEL.EDU> originally asked how do you bump macro variable I in the variable IN_&i during execution of the DATA step. This is a classic confusion about macro. The proper way to change the reference to a variable during execution of a DATA step is with arrays not macro code.

array in (2) $ 200 in_1 in_2 ; retain i 1 ;

/* when appropriate */ i = 2 ;

/* reference in ( i ) */

But Erik has now clarified his problem (see his message at the end) as opposed to telling one where he got stuck in his solution.

He has strings like 1-3,5,7-10 referring to a question ids and he wants to produce code (in pseudo form) like:

if q24 in (1-3,5,7-10) and q32 in (7,9,11-14,17) ;

Of course the IN operator does not understand the notation, hence he spells it out. The problem is that some strings spelled out require more than 200 bytes.

Personally I would say the IN operator is the wrong tool. I would use formats.

proc format ; value q24f 1-3,5,7-10 = 'GOOD' ; value q32f 7,9,11-14,17 = 'GOOD' ; run ;

Now the final form for the wanted code is

if put ( q24 , q24f4. ) = 'GOOD' and put ( q32 , q32f4. ) = 'GOOD' ;

Note that the need for most of the steps Erik spells out are no longer needed.

Ian Whitlock <whitloi1@westat.com> --------------------------------------------------------------------- Author: Erik Godfrey <sg94x892@DUNX1.OCS.DREXEL.EDU> at internet-e-mail Date: 4/11/96 2:36 PM Priority: Normal BCC: WHITLOI1 at REC TO: SAS-L@uga.cc.uga.edu at Internet-E-Mail Subject: Macro clarification ------------------------------- Message Contents ------------------------------- Thanks to everyone who responded to my posting (Need macro help, ver 2). Sorry the example I gave was rather vague; I will try to clarify what I need here. The situation is this: I have two strings, one called "question string", the other called "logic string". I am parsing both of these strings. Question string is being parsed as follows: the string will appear in a form similar to this: 1-3,5,7-10. There is also a variable called "qid" associated with this string. After it is parsed, I will have 8 observations, which will include the qid and question. So, for the preceding example, let's assume qid #24 is associated with question string. So, our dataset will appear as follows:

obs seq qid question (Seq is a sequence number. There may be --- --- --- -------- multiple question strings for a particular 1 1 24 1 group to process, so each question string 2 1 24 2 gets its own sequence number) 3 1 24 3 4 1 24 5 5 1 24 7 6 1 24 8 7 1 24 9 8 1 24 10

What I am then doing is this: I am creating a new dataset by setting the previously defined dataset. I am building "in" statements from the parsed question string, and am storing this "in" statement in a dataset variable called in_&i (where &i is an integer value, starting at 1). So, the value of in_1 could look as follows: in(1,2,3,5,7,8,9,10). The problem is with the 200 character limit of a variable - if the string I am building exceeds 200 characters, I want to increment the value of &i by 1 and store the remaining number of characters in in_2 (assuming the number of characters to be stored here does not exceed 200 characters). I would prefer not to write these individual in statements to external files, first of all, to avoid using an excessive number of %include statements later, and secondly, because I am then taking these individual in statements and building subsetting if statements in a succeeding dataset. This is being done as follows: I now have a dataset which may appear as follows:

obs seq qid in_1 --- --- --- ---- 1 1 24 in(1,2,3,5,7,8,9,10) 2 2 32 in(7,9,11,12,13,14,17)

The logic string may appear as: 1 & 2. What I am doing is replacing the numbers in the logic string with the contents of the in_1 statement, and writing this result to an external file to be included later with a %include. The file for this example will read as follows: if q24 in(1,2,3,5,7,8,9,10) and q32 in(7,9,11,12,13,14,17); All of this is working fine - my problem, as stated earlier, is generating multiple in_&i variables for a distinct seq number if the in statement being built exceeds 200 characters (if needed, when writing out the subsetting if statement to the file, I would append any subsequent in_&i statements, i.e., in_2, to the in_1 in statement). I hope this explanation clarifies what I need. Again, thanks to everyone who helped.

-- ============================================================================ == Erik Godfrey === == Programmer/Analyst === == Hay Group, Inc. === == Philadelphia, PA === ============================================================================