Date:         Fri, 30 Aug 2002 12:28:39 -0400
Reply-To:     Quentin McMullen <QuentinMcMullen@WESTAT.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Quentin McMullen <QuentinMcMullen@WESTAT.COM>
Subject:      Re: Is this %if syntax legal?
Comments: To: Jim Groeneveld <J.Groeneveld@ITGROUPS.COM>
Content-Type: text/plain; charset="iso-8859-1"

Jim Groeneveld [mailto:J.Groeneveld@ITGROUPS.COM] wrote in part:
> Quentin also very appropriately questioned the use of the
> %EVAL function. It
> could be a means to force an error indeed if your macro
> variables Rank#
> (#=1..9) would not be numeric (1..9). But it actually is superfluous,
> because the %EVAL function is implicit in macro conditions
> (%IF). And it
> would generate an error anyway if your macro variable would
> not be numeric.
> So, whatever your adapted code will be, you may omit the
> explicit %EVAL
> function.

Hi Jim,

I think what you wrote is only partly correct.  Yes, there is am implied
%eval() in the %IF when there are integer values being tested.  But if the
values being tested are non-integer or alpha, it will not run the %eval, it
will instead do a character comparison.  Thus below the macro language
happily evaluates 10<2.5 as true and 10<X as true.  The SAS language
evaluates 10<2.5 as false, and 10<X creates a note about the uninitialized
variable, then evaluates 10<. as false.

68   %macro comp(condition);
69     %if (&condition) %then %put In Macro Language: (&condition) is true;
70     %else %put In Macro Language: (&condition) is false;
71
72     data _null_;
73       if (&condition) then put "In SAS Language: (&condition) is true";
74       else put "In SAS Language: (&condition) is false";
75     run;
76   %mend comp;
77
78   %comp(10<2.5)
In Macro Language: (10<2.5) is true
In SAS Language: (10<2.5) is false
79   %comp(10<X)
In Macro Language: (10<X) is true
NOTE: Variable X is uninitialized.
In SAS Language: (10<X) is false

Given the willingness with which Macro language slips into a character
comparison for non-integers, I'm wondering if folks have made practice of
coding explicit %eval or %sysevalf in macro conditions that are designed to
compare numeric values?  Seems like a reasonable error-check to catch bad
parameters.

Kind Regards,
--Quentin