Issue #3942 - silx compare typo

[lesaintjerome]

Fix python set(3,4) command with set([3,4]).

[t20100]

Also, I guess this is to support RGB(A) images, and so checking the number of channels which is usually the last dimension:

Suggested change

	if data.ndim == 3 and data.shape[0] in set([3, 4]):
	if data.ndim == 3 and data.shape[2] in {3, 4}:

attn @vallsv

[lesaintjerome]

My use case was rather about comparing 2 (gray level) 3D volumes, possibly each of them with only one slice. Slice being the first index.

[t20100]

OK, I don't think this is supported: It is meant to support comparing 2 images..
Otherwise the limit to 3 or 4 slices makes no sense, while it does for RGB(A).

What could be added for stack with only one slice could be:

if data.ndim == 3 and data.shape[0] == 1:
    return data[0]

[t20100]

What's loaded here is then passed to:

silx/src/silx/gui/plot/CompareImages.py

Lines 473 to 482 in e912d5e

	def setImage1(self, image1, updateColormap="deprecated"):
	"""Set image1 to be compared.
	Images can contains floating-point or integer values, or RGB and RGBA
	values, but should have comparable intensities.
	RGB and RGBA images are provided as an array as `[width,height,channels]`
	of unsigned integer 8-bits or floating-points between 0.0 to 1.0.
	:param numpy.ndarray image1: The first image

[lesaintjerome]

OK. I probably need to be more precise on the issue. The error happened when with

silx compare "bambou_hercules_0001_rec_2040.800_00000.hdf5?path=/entry0000/reconstruction/results/data" "bambou_hercules_0001_rec_2058.800_00000.hdf5?path=/entry0000/reconstruction/results/data"

, i.e. with no slice number. But when the command

silx compare "bambou_hercules_0001_rec_2040.800_00000.hdf5?path=/entry0000/reconstruction/results/data&slice=0" "bambou_hercules_0001_rec_2058.800_00000.hdf5?path=/entry0000/reconstruction/results/data&slice=0"

is used (i.e. with slice number), then it works fine. That's why I thought stacks were possible....

[vallsv]

Yes, to support slicing with a slider for example, you better to create another dedicated issue.

I don't think we will have time to do it now.

Optionally you could try to see if slice=* works, but i am pretty sure it is not supported.

[vallsv]

@t20100 i agree there is something weird with the shape index which is supposed to be ordered y,x,rgb[a]. Ill check on my side.

[vallsv]

BTW, this shape[0] is maybe the layout provided by fabio or another lib when we load a RGB image. I have to check.

[vallsv]

I merge this PR for convenience.
I will rework the other proposals with the PR #3944