/black

@wyli PR is getting very big as I'm doing all padding and cropping. Apologies for that, it's a hard one to split.

@wyli Could you create unit tests for the _forward_meta methods? I added test_meta_update to test_pad.py (link). I wanted to check that for out = padder(im), the input metadata was unchanged, and out had been updated. I see that the output affine is only 2x2. Could you check this? Putting the same check for cropping would be good too.

Equally we would want to check that the affine isn't modified when we don't crop/pad from the origin.

@wyli Could you create unit tests for the _forward_meta methods?

sure let me add some basic tests to this PR now...

@wyli thanks for the help here. I'm off for a week, feel free to continue this without me if you want to. All transforms have been updated, just need to get all the linting happy. I suspect #4435 will need to be addressed first.

@wyli thanks for the help here. I'm off for a week, feel free to continue this without me if you want to. All transforms have been updated, just need to get all the linting happy. I suspect #4435 will need to be addressed first.

thanks! I'll try to merge this one soon.

cropping with multiple random samples (basic applied_operations collating/decollating work fine)

from monai.transforms import EnsureChannelFirst, LoadImage, RandSpatialCropSamples, SaveImage

filename = "~/Downloads/avg152T1_LR_nifti.nii.gz"
img = LoadImage()(filename)
img = EnsureChannelFirst()(img)
img = RandSpatialCropSamples(roi_size=(40, 50, 60), random_size=False, num_samples=3)(img)
for i in img:
    SaveImage(resample=False)(i)