pseudo_gravity_packing: Not adding all possible spheres and input radius not correctly used

[heinsimon]

pseudo_gravity_packing: Not adding all possible spheres and input radius not correctly used

[jgostick]

I have fixed a lot of bugs in my recent PR (#922), so caught one of the ones you mentioned: the new version will add +1 to the radius if smooth=True and leave it as is if smooth=False. However, the code you have above still still shows a bug, but I can explain why. If you reduce the r to 14, then you get 2 spheres. The reason is that to create the 'contained' edge condition I put a layer of False around the outside, but this technically makes the box 28 wide instead of 30, so spheres with r=15 do not fit. And the 3rd sphere is missing because the image is too short, only 83 instead of 90. So if you do imbox=np.ones([30, 30, 90]) and r=14 it all fits, of do imbox=np.ones([31, 31, 91]) with r=15.

Anyway, your bug report is valid, and I think I can fix it by padding the input im with False instead of inserting False along the edges of the existing image. This is probably doable, so I will add it to the PR.

[jgostick]

I was mistaken, the code works fine. I do 'pad' the image, not overwrite the boundaries. The reason that the spheres don't fit is because a sphere with a radius of 15 voxels is actually 31 voxels wide. The following circle has a radius of 1:

[[0 1, 0],
 [1, 1, 1],
 [0, 1, 0]]

So with a radius of 1, the sphere is actually 3 voxels wide. The 'radius extends from the center of the middle pixel, to the center of the outer pixel, but the digitization requires us to draw an entire pixel.