Reply to thread

Message: [QUOTE="Mark Dickinson, post: 4100503"] It's the Windows way of representing a NaN (not a number). A NaN is typically what you get when you try to perform an invalid floating-point operation, like taking the square root of a negative number, or dividing 0.0 by 0.0. Some people also use NaNs to represent uninitialized values, or as placeholders for missing data. Well, it's certainly not ideal. It shouldn't be a problem in Python 2.6 or 2.7, though; unfortunately, Python 2.5 is no longer receiving bugfixes, so it's not going to change there. Have you tried using pickle protocol 1 or 2, instead of pickle protocol 0? That may well solve your problem. (Those protocols write out the binary form of a float directly, instead of reading and writing a string representation.) [/QUOTE]