@Tom:
I don't think it's a matter of "better" or "worse"; just two different pieces of SAS functionality.
If you only want to know if the value to which an expression resolves is present an array A, then coding:
<expression> IN A
is less busy code-wise (and, IMO, more self-explanatory) than:
whichn (<expression>, of A[*]) > 0
Besides, IN is good for both numeric and character arrays, while WHICHN/C has to be type-specific. If you want to know the index of the first array item where <expression> is found, then of course IN doesn't tell while WHICHN/C does.
Performance-wise:
- Both use sequential search, so both execute in O(N) time.
- IN doesn't use any extra memory over the memory allocated for the array, no matter how many times the compiler sees it.
- WHICHN/C uses extra memory equal to that allocated for the array for every reference to WHICHN/C the compiler sees.
For example, run the following two pieces of code and compare their RAM footprints:
option fullstimer ;
data _null_ ;
array a [1000000] _temporary_ (1:1000000) ;
do x = 1, 500000, 1000000 ;
f1 = x in a ;
f2 = x + 1 in a ;
f3 = x - 1 in a ;
put (f1-f3) (=) ;
end ;
run ;
data _null_ ;
array a [1000000] _temporary_ (1:1000000) ;
do x = 1, 500000, 1000000 ;
f1 = whichn (x, of a[*]) > 0 ;
f2 = whichn (x + 1, of a[*]) > 0 ;
f3 = whichn (x - 1, of a[*]) > 0 ;
put (f1-f3) (=) ;
end ;
run ;
Then comment out one or two calls for IN and WHICHN and watch what FULLSTIMER reports compared to the uncommented version.
Kind regards
Paul D.
