WIP: BUG: tzyxc bin shrink crash

Re: #157

Author: thewtex

ngff_zarr/methods/_itk.py

@@ -149,6 +149,11 @@ def _downsample_itk_bin_shrink(
         # For consistency for now, do not utilize direction until there is standardized support for
         # direction cosines / orientation in OME-NGFF
         # block_0.attrs.pop("direction", None)
+        if "c" in previous_image.dims:
+            raise ValueError(
+                "Downsampling with ITK BinShrinkImageFilter does not support channel dimension 'c'. "
+                "Use ITK Gaussian downsampling instead."
+            )
         block_input = itk.image_from_array(np.ones_like(block_0))
         spacing = [previous_image.scale[d] for d in spatial_dims]
         block_input.SetSpacing(spacing)

ngff_zarr/methods/_support.py

@@ -111,15 +111,25 @@ def _dim_scale_factors(dims, scale_factor, previous_dim_factors):
         result_scale_factors = {
             d: int(scale_factor[d] / previous_dim_factors[d]) for d in scale_factor
         }
+        # if a dim is not in the scale_factors, add it with a scale factor of 1
+        for d in dims:
+            if d not in result_scale_factors:
+                result_scale_factors[d] = 1
+
     return result_scale_factors
 
+
 def _update_previous_dim_factors(scale_factor, spatial_dims, previous_dim_factors):
+    previous_dim_factors = copy.copy(previous_dim_factors)
     if isinstance(scale_factor, int):
-        previous_dim_factors = { d : scale_factor for d in spatial_dims }
+        for d in spatial_dims:
+            previous_dim_factors[d] = scale_factor
     else:
-        previous_dim_factors = scale_factor
+        for d in scale_factor:
+            previous_dim_factors[d] = scale_factor[d]
     return previous_dim_factors
 
+
 def _align_chunks(previous_image, default_chunks, dim_factors):
     block_0_shape = [c[0] for c in previous_image.data.chunks]