Preventing Divide by Zero Errors

QUESTION: I'm dividing one array by another in IDL, but occasionally the divisor array contains zeros. This, of course, causes a divide by zero error and a positive or negative infinity (INF) to be inserted in the result array. Is there an elegant way to code this up to prevent this sort of thing?

ANSWER: Mais oui! Ed Meinel has pointed this out to me on more than a couple of occasions. Here is his solution for integers (represented as floating arrays, if you want to do floating point arithmetic, of course). The variable replacement_value represents whatever value you think is appropriate in the circumstances. Something like 1e-6 often works fine.

    IDL> c = a / (b + replacement_value * (b EQ 0))

For floating point values, where the number zero cannot be assumed to be exactly zero, Jean Hasban suggests this variation,

   IDL> c = a / (b + replacement_value * (ABS(b) LT epsilon)) 

where epsilon is usually taken to be something like the smallest number that can be represented on your computer, or:

   IDL> epsilon = (Machar()).eps    IDL> Print, epsilon         1.19209e-007

Other users just prefer to get the error. Then, at least they know what is going on.

Version of IDL used to prepare this article: IDL 7.0.1.

Google
 
Web Coyote's Guide to IDL Programming