Minimum Integer Values

• To: rakesh@xxxxxxxxxxxxxxxxxxxxxxxx
• Subject: Minimum Integer Values
• From: Russ Rew <russ@xxxxxxxxxxxxxxxx>
• Date: Thu, 5 Mar 1992 15:45:23 -0700

Rakesh Mithal (rakesh@xxxxxxxxxxxxxxxxxxxxxxxx) asks:

> Can the use of -2147483648 for missing value in netCDF file create
> problems on some machine?

Yes, and the situation for floating-point values is even worse, since there
is no suitable value near the edge of the floating point range that maps
into the same IEEE floating-point external representation for all platforms.

That's one of the reasons we added the special treatment of the "_FillValue"
attribute.  We now recommend not using the default fill values when you need
a special value to represent missing-values.  It's better to use
application-specific and application-appropriate fill values, by defining
the attribute "_FillValue" for a variable before writing it.

--Russ

• References:
  • Minimum Integer Values
    • From: rakesh