In Pandas for np.nan, why is "mul()" not equal to "mul()"?

Question

On a simple Pandas Series, when I am comparing the results of mul() and __mul__(), there is a 'False' boolean that is being returned for NaN value.

Why?

    x = pd.Series([0,1,2,np.nan,5])

    x.mul(1) == x.__mul__(1)

Result:
    0     True
    1     True
    2     True
    3    False
    4     True
    dtype: bool

`NaN` never compares equal to anything, not even itself. That's part of its definition. — Nate Eldredge, May 21 '20 at 19:27

score 1 · Answer 1 · answered May 21 '20 at 19:36

One of the defining properties of NaN is that it is not equal to itself. See What is the rationale for all comparisons returning false for IEEE754 NaN values?

So when x equals np.nan, presumably x.mul(1) and x.__mul__(1) are both np.nan again, and the result of testing this value for equality with itself is false.

In Pandas for np.nan, why is "mul()" not equal to "__mul__()"?

1 Answers1

In Pandas for np.nan, why is "mul()" not equal to "mul()"?