In pandas, is it possible to construct a boolean series for indexing that use custom objects?
i.e.
class Test():
def __init__(self, num):
self.num = num
def __bool__(self):
return self.num == 3
x = Test(2)
y = Test(3)
df = pd.DataFrame({'A':[x,y]})
print(bool(df['A'].iloc[1]))
print(df.where(df['A'] == True))
returns
True
A
0 NaN
1 NaN
What I'd like would be something like
True
A
0 False
1 True
Or similar so that I can use .first_valid_index() to grab the first occurrence in a different function.
Is there any way to check the "Truthiness" of an object to construct the new Series?
Don't use ==. map bool instead
df.where(df['A'].map(bool))
A
0 NaN
1 <__main__.Test object at 0x000002A70187E6D0>
Or astype(bool)
df.where(df.astype(bool))
A
0 NaN
1 <__main__.Test object at 0x000002A70187E6D0>
However, if you define an __eq__
class Test():
def __init__(self, num):
self.num = num
def __bool__(self):
return self.num == 3
def __eq__(self, other):
if isinstance(other, type(self)):
return bool(other) == bool(self)
else:
try:
return type(other)(self) == other
except:
return False
x = Test(2)
y = Test(3)
df = pd.DataFrame({'A':[x,y]})
print(bool(df['A'].iloc[1]))
print(df.where(df['A'] == True))
True
A
0 NaN
1 <__main__.Test object at 0x000002A701897520>