Finding NANs in Arrays
QUESTION: I'm having a problem testing for whether an entry in an array is not-a-number (NAN). I'm doing something like this:
IF(value EQ !VALUES.F_NAN) THEN BEGIN Print,'Not a number' ENDIF ELSE BEGIN Print,'Is a number!' ENDELSE
The problem is, this code always returns Is a number, even if I set value equal to !VALUES.F_NAN. What am I doing wrong?
ANSWER: I responded to this question with this sage advice.
The problem is that an NAN is ... well, not a number. Thus, you can't use it in expressions that require a number. (Think of it as a mathematical Catch-22, if you like.)
The proper way to write this code is like this:
IF(Finite(value) EQ 0 THEN BEGIN Print,'Not a number' ENDIF ELSE BEGIN Print,'Is a number!' ENDELSE
It is then that we all learned even simple answers aren't always so simple. Tom McGlynn immediately responded with this clarification.
That doesn't distinguish NaN from the infinities. The standard trick in any language for looking for NaN's isif x ne x then begin print,'This is a NaN' endif else ...
This can get optimized away if the compiler/interpreter is poorly designed. NaN's are not equal to anything --- even themselves.
To which Ken Bowman, who wisely was keeping his eye on the documentation, replied.
The FINITE function has three keywords: NAN, INFINITY, and SIGN to distinguish between NaNs, Infs, and to return the signs of arguments.
Note that you can only use NaNs in floating point (!Values.F_NaN) and double precision (!Values.D_NaN) arrays. There is no NaN bit pattern for byte, integer, or long data types.
Copyright © 1997-2003 David W. Fanning
Written 5 January 2003
Updated 17 October 2013