Check Truthiness of Values in Pandas Series

Question

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?

Answer 1

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>