NumPy (hyper)cube view

[s-celles]

Hello Yuri,

I'm trying to view a NumPy 3D array with gtabview

a = np.array(np.random.random((3,4,5)))

Compare
gtabview

with

Spyder variables editor

With an hyper-cube (4 dimensions)

a = np.array(np.random.random((3,4,5,2)))

gtabview

Spyder variables editor

(which means that arrays with more than 3 dimensions can't be displayed)

Being able to view a (3D) cube will help to view a Pandas Panel as slice (and not only a flatten Panel as it's done actually).

An Excel file is also a kind of (3D) cube.

First dimension is sheetname
Second is row
Third is column

Kind regards

[wavexx]

On 25/07/15 09:16, scls19fr wrote:

Being able to view a (3D) cube will help to view a Pandas Panel as slice
(and not a flatten Panel)

This is something I'd love to support.

Where would you choose/handle the third dimension?

The usual approach is another slider.

Maybe we can re-use that empty space in the top-left corner with a spinner.

[s-celles]

I like the UI approach of Spyder about this slicing problem.

[s-celles]

If you have time constraints I'd prefer Blaze support before cube and hyper-cube and matrix because I'm pretty sure that I will use database connection with my students in September.

[wavexx]

On 25/07/15 11:18, scls19fr wrote:

If you have time constraints I'd prefer Blaze support before cube and
hyper-cube and matrix.

I'll check out blaze definitely before.
I'm particularly interested in browsing large DB tables downloading them
in full.

3d is "cool to have", but since it's going to be some sort of slicing
anyway, there's nothing you cannot already do (albeit manually).

[s-celles]

That's exactly how I consider things too.