[SciPy-user] nan puzzle

[Fernando Perez]

Alan G Isaac wrote:
> Is this expected behavior?

doesn't seem to happen under linux:

In [8]: from scipy import nan

In [9]: z=[0,1,nan]

In [10]: x=nan

In [11]: x is nan
Out[11]: True

In [12]: x = z[2]

In [13]: x is nan
Out[13]: True

In [14]: x
Out[14]: nan


If I recall correctly Tim Peter's posts on the subject, the IEEE-754 behavior 
of python is 100% platform dependent, and fully at the whim of the underlying 
C library implementation.  Under win32, this means whatever Microsoft was 
willing to do.  I wouldn't exactly hold my breath on that.

Cheers,

f