Add support for loading MAT files

[cbrnr]

2D arrays from arbitrary MAT files can be loaded by selecting the variable and setting the sampling frequency in a dedicated dialog. Fixes #195.

[cbrnr]

@hofaflo if you have time please test this new feature.

[hofaflo]

This works really well, just two notes:

The entries in the "Name" column in the MatDialog are indented:

Would it make sense to import 3d-arrays as mne.Epochs?

[cbrnr]

The entries in the "Name" column in the MatDialog are indented:

I haven't noticed that with data I tested. Which file did you use that has this problem? Did you try a file where the entries are not indented?

Would it make sense to import 3d-arrays as mne.Epochs?

Definitely. But I think I'll do that at a later point, because this will require juggling more with the dimensions.

[hofaflo]

Happens with all files I found, even with a minimal example created like this:

import numpy as np
from scipy.io import savemat

a = np.zeros((100, 2))
savemat("test.mat", {"a": a})

[cbrnr]

Ah, I think that's because I'm displaying that as a QTreeWidget. When there's no nesting, this looks like an indentation where normally there would be an arrow. Try loading a .mat file from e.g. data set 1 from http://bnci-horizon-2020.eu/database/data-sets to see how a nested data structure (MATLAB struct) looks like.

Not sure it's worth special-casing when there is no nesting, only if it's not too much effort.

[cbrnr]

I'm now showing the file name as the root of the tree so there is no stray indentation. Better?

[cbrnr]

Off-topic: the default icon in the top left corner of a dialog on Windows looks really ugly. Can/should we disable this?

[hofaflo]

I'm now showing the file name as the root of the tree so there is no stray indentation. Better?

Yes that's perfect!

Off-topic: the default icon in the top left corner of a dialog on Windows looks really ugly. Can/should we disable this?

Not sure if that's worth the effort 😅

[cbrnr]

Not sure if that's worth the effort 😅

If it's easy to do (e.g. set a flag) then probably yes.