Remove NaN value from a set

Question

Is it possible to easily remove NaN values for a Python Set object? Given that NaN values do not equal anything (and float('nan') is float('nan') is also False), you can end up with many NaN values in a Set.

>>> a = set( (float('nan'), float('nan'), 'a') )
>>> a
{nan, nan, 'a'}

The best I can come up with it to define a function like math.isnan, but that is tolerant of non-float types like:

def my_isnan(x):
    try:
        return math.isnan(x)
    except TypeError:
        return False

Then you can use a set comprehension like this:

>>> {x for x in a if not my_isnan(x)}
{'a'}

Answer 1

In practice, you could look at the fact that nan != nan as a feature, not a bug:

>>> a = {float('nan'), float('nan'), 'a'}
>>> a
{nan, nan, 'a'}
>>> {x for x in a if x==x}
{'a'}

On the positive side, no need for a helper function. On the negative side, if you have a non-nan object which is also not equal to itself, you'll remove that too.

Answer 2

Use pd.notna() from pandas, e.g.:

In [219]: import pandas as pd

In [220]: a = set((float('nan'), float('nan'), 'a'))

In [221]: a = {x for x in a if pd.notna(x)}

In [222]: a
Out[222]: {'a'}