Is a division by voxel size missing in sphere radius logic?

[jasper-tms]

It looks to me like sphere_annotation() is expecting radii in the units of the annotation layer (often voxels), and is then scaling z down so that the physical size (nm) of the sphere is the same in all axes:

nglui/src/nglui/easyviewer/ev_base/nglite/annotation.py

Lines 67 to 77 in 1ca1abf

	def sphere_annotation(
	center,
	radius,
	z_multiplier,
	id=None,
	description=None,
	linked_segmentation=None,
	tag_ids=None,
	):
	"""
	Assumes the z-axis is anistropic

That seems sensible enough, but that means sphere_annotation() must be called with radii already converted to the unit of the annotation layer. SphereMapper says it expects the radius_column to contain radii in nm, but I think those values are getting passed to sphere_annotation() as those nm values that the user provides, without the conversion to layer resolution that I think needs to happen before calling sphere_annotation():

nglui/src/nglui/statebuilder/mappers.py

Lines 759 to 772 in 1ca1abf

	rs = data[col_rad][relinds].values
	if viewer_resolution:
	z_multiplier = viewer_resolution[1] / viewer_resolution[2]
	else:
	z_multiplier = self._z_multiplier
	descriptions = self._descriptions(data[relinds])
	linked_segs = self._linked_segmentations(data[relinds])
	tags = self._assign_tags(data)
	annos = [
	viewer.sphere_annotation(
	pt,
	r,

If I'm seeing this correctly, then something like

        rs /= viewer_resolution[1]

could be stuck between the following two lines to do the conversion:

nglui/src/nglui/statebuilder/mappers.py

Lines 761 to 762 in 1ca1abf

	if viewer_resolution:
	z_multiplier = viewer_resolution[1] / viewer_resolution[2]

Having added that one line, my spheres now appear at the correct size when I'm passing in nm radii values in an annotation layer that has its units as voxels.

[ceesem]

Yeah, good catch. It looks like the current code only scales the points and not the radius. I'll fix this.

[ceesem]

Should be fixed by #65 in the current release