Multi-canvas experiments

[aganders3]

This is an experiment in napari multi-canvas with an attempt to minimize breaking API changes. Tests are almost green!

There are some hacks and inefficiencies here, but in general it's working. The idea is you get a list of canvases (_MultiCanvas objects) as part of the ViewerModel. These canvases take ownership of the dims and camera, allowing you to have multiple views of the same data.

ViewerModel maintains dims and camera properties that simply delegate to a primary/active (right now just the first) canvas.

Since slice/view data currently lives on the layer model, the data must be resliced whenever the canvas changes. I'm going to start looking at this next.

Here's a brief demo:

Screen.Recording.2023-06-09.at.2.37.41.PM.mov

[aganders3]

Here's an updated screenshot of this WIP

import napari
import numpy as np

viewer = napari.Viewer()

viewer.open_sample('napari', 'cells3d')
viewer.add_canvas()
viewer.add_canvas()
viewer.add_canvas()

viewer.dims.ndisplay = 3
viewer._canvases[1].dims.order = (2, 1, 0)
viewer._canvases[3].dims.order = (2, 0, 1)

if __name__ == "__main__":
    napari.run()