Put a 2d Array into a Pandas Series

Question

I have a 2D Numpy array that I would like to put in a pandas Series (not a DataFrame):

>>> import pandas as pd
>>> import numpy as np
>>> a = np.zeros((5, 2))
>>> a
array([[ 0.,  0.],
       [ 0.,  0.],
       [ 0.,  0.],
       [ 0.,  0.],
       [ 0.,  0.]])

But this throws an error:

>>> s = pd.Series(a)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/miniconda/envs/pyspark/lib/python3.4/site-packages/pandas/core/series.py", line 227, in __init__
    raise_cast_failure=True)
  File "/miniconda/envs/pyspark/lib/python3.4/site-packages/pandas/core/series.py", line 2920, in _sanitize_array
    raise Exception('Data must be 1-dimensional')
Exception: Data must be 1-dimensional

It is possible with a hack:

>>> s = pd.Series(map(lambda x:[x], a)).apply(lambda x:x[0])
>>> s
0    [0.0, 0.0]
1    [0.0, 0.0]
2    [0.0, 0.0]
3    [0.0, 0.0]
4    [0.0, 0.0]

Is there a better way?

Answer 1

Well, you can use the numpy.ndarray.tolist function, like so:

>>> a = np.zeros((5,2))
>>> a
array([[ 0.,  0.],
       [ 0.,  0.],
       [ 0.,  0.],
       [ 0.,  0.],
       [ 0.,  0.]])
>>> a.tolist()
[[0.0, 0.0], [0.0, 0.0], [0.0, 0.0], [0.0, 0.0], [0.0, 0.0]]
>>> pd.Series(a.tolist())
0    [0.0, 0.0]
1    [0.0, 0.0]
2    [0.0, 0.0]
3    [0.0, 0.0]
4    [0.0, 0.0]
dtype: object

EDIT:

A faster way to accomplish a similar result is to simply do pd.Series(list(a)). This will make a Series of numpy arrays instead of Python lists, so should be faster than a.tolist which returns a list of Python lists.

Answer 2

 pd.Series(list(a))

is consistently slower than

pd.Series(a.tolist())

tested 20,000,000 -- 500,000 rows

a = np.ones((500000,2))

showing only 1,000,000 rows:

%timeit pd.Series(list(a))
1 loop, best of 3: 301 ms per loop

%timeit pd.Series(a.tolist())
1 loop, best of 3: 261 ms per loop