Pb converting a list of pandas.Series into a numpy array of pandas.Series

Question

I would like to convert a list of pandas.Series into a numpy array of pandas.Series. But when I call the array constructor, it also converting my Series.

>>> l = [Series([1,2,3]),Series([4,5,6])]
>>> np.array(l)
array([[1, 2, 3],
       [4, 5, 6]], dtype=int64)

My list is small (~10 elements), so for performances issues I would like to avoid to create a pandas.DataFrame. Is there an easy workaround?

Thanks in advance

Solution

You should set the dtype of the array when you assign it:

l = [pd.Series([1,2,3]),pd.Series([4,5,6])]
np.array(l, dtype=pd.Series)

Though it is raises the question: why do you want an ndarray of series, and not an ndarray of the contents of the series?

Comments

[deubNippon]

Thanks a lot! I need to use later on,on some of those Series, a rolling_moment function, which to my knowledge, exist only in Pandas.

[jme]

Don't thank me too fast! On second inspection, it may not work. But you might want to just keep the series objects in a python list, in that case.

[deubNippon]

Well..thanks for trying, but it's not working indeed. Unfortunately, I really need a np.array

[MANU]

If the aim is to convert to Numpy-array representation the you can see pandas.Series.as_matrix or pandas.DataFrame.values