[Numpy-discussion] the difference of np.nan, np.NaN, np.NAN?

[David Cournapeau]

On Fri, Oct 14, 2011 at 11:53 AM, Olivier Delalleau <shish at keba.be> wrote:
> 2011/10/14 Matthew Brett <matthew.brett at gmail.com>
>>
>> Hi,
>>
>> On Fri, Oct 14, 2011 at 4:33 AM, Chao YUE <chaoyuejoy at gmail.com> wrote:
>> > Dear all,
>> >
>> > is there any difference between np.nan, np.NaN and np.NAN? they really
>> > confuse me....
>> > they are all Not a Number?
>> >
>> > In [75]: np.nan==np.NaN
>> > Out[75]: False
>> >
>> > In [77]: np.NaN==np.NAN
>> > Out[77]: False
>>
>> The nan value is not equal to itself:
>>
>> In [70]: np.nan == np.nan
>> Out[70]: False
>>
>> See:
>>
>> http://en.wikipedia.org/wiki/NaN
>>
>> But:
>>
>> In [71]: np.isnan(np.nan)
>> Out[71]: True
>>
>> In [72]: np.isnan(np.NAN)
>> Out[72]: True
>>
>> In [73]: np.isnan(np.NaN)
>> Out[73]: True
>>
>> Best,
>>
>> Matthew
>
> Also on my computer:
>>>> numpy.nan is numpy.NAN
> True
>>>> numpy.nan is numpy.NaN
> True
>
> So they really are the same. But you shouldn't rely on this (always use
> numpy.isnan to test for nan-ness).

They are the same, just not equal to each other  (or even to
themselves). As for the different NaN names in numpy namespace, I
think this is for historical reason (maybe because python itself used
to print nan differently on different platforms, although this should
not be the case anymore since python 2.6).

David