Why does .dropna().apply() on a dataframe actually work?

Question

Given the following dataframe:

df = pd.DataFrame({\'foo\': [1, np.nan, 1, 1]})

When having to deal with nan values, I\'m used to handle them with