error with _utils.py (```pd.api.types.is_datetime64_dtype``` function not found)

[stephanieleroux]

Hi Guillaume,
Thanks for sharing your routines!
I have started using some of your fft functions.
Just in case it'd be useful to you and to others: when i installed the package on my laptop (OSX, with anacconda python 3.5.5) i ran your setup.py and everything went fine (no error message) but then, when i imported the package in my python script, i got an error at l.173 in _utils.py (function pd.api.types.is_datetime64_dtype not found).
I solved my problem by replacing :
l.173 if pd.api.types.is_datetime64_dtype(x):
by:
if pd.core.common.is_datetime64_dtype(x):
and then re-ran python setup.py install.

[serazing]

Hi Stephanie,

Are you using the latest version of xscale ? It seems that I have already corrected this error. See there the latest _utils.py.

Try git pull to update your local repository.

[stephanieleroux]

Ok, thanks. My mistake. i downloaded the package last week and true i haven't updated it since then.

[serazing]

Let me know if it works. I will close the issue if everything is right.

[stephanieleroux]

After updating xscale and testing again:

• If i use the latest version of xscale with if pd.core.common.is_datetime64_dtype(x): at l.173 in _utils.py, i get the following error when calling xfft.fft:
  AttributeError: module 'pandas.core.common' has no attribute 'is_datetime64_dtype'

• What DOES WORK for me is when i replace l.173 in _utils.py by if pd.api.types.is_datetime64_dtype(x):

• For info, my current pandas package is :

# Name                    Version                   Build  Channel
pandas                    0.23.0           py35h1702cab_0  

[serazing]

Ok, I understand better now. I have the same issue with the latest version of pandas. I think you have inverted your correction with the original code in your first message.

[serazing]

I just submitted a pull request, which should correct this issue. Feel free to submit PRs to fix future bugs.