Python None vs Numpy Nan

[chenshuai_2004]

python原生的None和pandas, numpy中的numpy.NaN在功能上都是用来标示空缺数据

a)None是一个python特殊的数据类型， 但是NaN却是用一个特殊的float
print( type(None) ,type(np.nan)) --><class 'NoneType'> <class 'float'>

sum([1,2,None]) -->TypeError: unsupported operand type(s) for +: 'int' and 'NoneType'
sum([1,2,np.nan]) -->nan

np.nansum([1,2,np.nan]) -->3
np.nansum([1,2,None]) -->unsupported operand type(s) for +: ‘int’ and ‘NoneType’

b)All can be used as key in dict;pandas Series/dataframe. But the same & different always exist
None == np.nan -->False
{None:1, np.nan:2} -->None != NaN

s = pd.Series([None, np.nan, 'b']) -->None != NaN

pd.Series([1, None])
=== pd.Series([1, None]) -->True False
s.map({None:1,'b':'a'}) -->None == NaN
0 1
1 1
2 a
dtype: object

s.map({np.nan:2,'None':1,'b':'a'}) -->Nan take into effection 1st
0 2
1 2
2 a
dtype: object

c)np.nan & None impact on other type
np.array([1, None]).dtype -->dtype('O')
np.array([1, np.nan]).dtype -->dtype('float64')

How to convert from np.nan to None
s = pd.Series([None,np.nan,'a'])

s.replace([np.nan],None) -->Seriers（[None，None，'a'])
s.replace(np.nan,None) -->Seriers（[None，None，'a'])

s.where(s.notnull(), None) -->Seriers（[None，None，'a'])

[屎克螂]