Your series contains other pd.Series
objects. This is bad practice. In general, you should ensure your series is of a fixed type to enable you to perform manipulations without having to check for type
explicitly.
Your error is due to pd.Series
objects not being hashable. One workaround is to use a function to convert pd.Series
objects to a hashable type such as tuple
:
s = pd.Series(['one string', 'another string', pd.Series([1, 2, 3])]) def converter(x): if isinstance(x, pd.Series): return tuple(x.values) else: return x res = s.apply(converter).unique() print(res) ['one string' 'another string' (1, 2, 3)]