Environment install lacks dependencies

[cversteeg]

Import statements used in the demo are not in the general environment installation. (xarray, at least). I might also suggest that a docker container be made, if it's not a huge hassle.

[cversteeg]

mat73 also on the list.

[raeedcho]

Docker seems like overkill for what's basically a pandas wrapper. Perhaps some indication of which Python version to use would be good though...I didn't have any issues with using the code when I used Python 3.7.

Also mat73 is an optional dependency (i.e. to be installed by the user), which is only imported in cases where the mat file was saved with the 'v7.3' flag. See discussion here for more info.

[raeedcho]

Oh also, I highly recommend using pipenv instead of Docker for managing simple python environments (though I suppose conda works too, if you're using Anaconda).

[bagibence]

Thanks for raising this! Yes, we'll definitely need to document the dependencies. I didn't include xarray because it's only used in the demo, the internals of pyaldata don't rely on it, and -- even though it's one of my favorite packages -- I didn't want to force it on the user.

I agree with @raeedcho in that I don't see why Docker would be needed.

I'm not sure about the Python version. My guess now would be anything >=3.6 because we use f-strings quite a lot. I'm on 3.8 and it works fine, so we'll just have to test 3.6 and 3.9.

That said, @catiafortunato discovered that the trial_average function doesn't work with the newest version of pandas, so that will have to be fixed and indicated somewhere.

[bagibence]

Fixed by #141 and #143

• 141 updates trial_average to work with pandas v2

• 143 specifies some dependencies that work