Nan => UInt bug?

[10Things_Rob]

When I have a DBL with value NaN and I convert it into a U64 and flatten it to a string, I get x0000 0000 for a value.

When I covert it to a I64 first, I get x7FFF FFFF.

I believe that the U64 case is a bug (it's also true for U32s). It certainly is inconvenient for my application.

Any comment?

[GerdW]

Hi 10things,

my comment is copied from the LabView help on NaN:
*******
Floating-point operations propagate NaN and Inf faithfully. When you explicitly or implicitly convert NaN or Inf to integers or Boolean values, the values become meaningless. For example, dividing 1 by zero produces Inf. Converting Inf to a 16-bit integer produces the value 32,767, which appears to be a normal value.

Before you convert data to integer data types, use the Probe tool to check intermediate floating-point values for validity. Check for NaN by wiring the Comparison function Not A Number/Path/Refnum? to the value you suspect is invalid.

*******

The important parts are set to bold

Another note: How do you want to convert NaN? What should be the result? For integers something like NaN doesn't exist...

Message Edited by GerdW on 12-03-2007 09:40 AM