Error for ImageLayer creation for an image with channels, scale, transform

🐛 Bug

After upgrading napari, a traceback is now obtained for an ImageLayer created with a channel dimension, scale, and translate.

To Reproduce

Test script:

import napari
import numpy as np

data = np.random.rand(20, 20, 2).astype(np.float32)
image_layer = napari.layers.Image(data, scale=[1.0, 1.0], translate=[3.0, 4.0])

Results in the traceback:

Traceback (most recent call last):
  File "/tmp/napari_test.py", line 5, in <module>
    image_layer = napari.layers.Image(data, scale=[1.0, 1.0], translate=[3.0, 4.0])
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/napari/layers/image/image.py", line 290, in __init__
    self._update_dims()
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/napari/layers/base/base.py", line 544, in _update_dims
    self.refresh()
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/napari/layers/base/base.py", line 946, in refresh
    self.set_view_slice()
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/napari/layers/base/base.py", line 810, in set_view_slice
    self._set_view_slice()
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/napari/layers/image/image.py", line 519, in _set_view_slice
    indices = np.array(self._slice_indices)
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/napari/layers/base/base.py", line 612, in _slice_indices
    inv_transform = self._data_to_world.inverse
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/napari/layers/base/base.py", line 996, in _data_to_world
    return self._transforms[1:3].simplified
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/napari/utils/transforms/transforms.py", line 125, in simplified
    return tz.pipe(self[0], *[tf.compose for tf in self[1:]])
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/toolz/functoolz.py", line 627, in pipe
    data = func(data)
  File "/home/matt/bin/miniconda/envs/napari/lib/python3.9/site-packages/napari/utils/transforms/transforms.py", line 460, in compose
    affine_matrix = self.affine_matrix @ transform.affine_matrix
ValueError: matmul: Input operand 1 has a mismatch in its core dimension 0, with gufunc signature (n?,k),(k,m?)->(n?,m?) (size 3 is different from 4)

Expected behavior

No traceback. Identified in CI: https://github.com/InsightSoftwareConsortium/napari-itk-io/pull/3

Environment

• Please copy and paste the information at napari info option in help menubar here:

napari: 0.4.10
Platform: Linux-5.9.8-050908-generic-x86_64-with-glibc2.31
System: Ubuntu 20.04.2 LTS
Python: 3.9.2 | packaged by conda-forge | (default, Feb 21 2021, 05:02:46) [GCC 9.3.0]
Qt: 5.15.2
PyQt5: 5.15.4
NumPy: 1.20.2
SciPy: 1.6.2
Dask: 2021.04.0
VisPy: 0.6.6

OpenGL:
- GL version: 4.6.0 NVIDIA 460.80
- MAX_TEXTURE_SIZE: 32768

Screens:
- screen 1: resolution 3840x2160, scale 1.0
- screen 2: resolution 3840x2160, scale 1.0

Plugins:
- console: 0.0.3
- napari-itk-io: 0.1.1.dev4+g2874092
- svg: 0.1.4

[jni]

Yeah I hadn’t read the above correctly — indeed @thewtex for that image you will have to provide channel_axis=-1 as a kwarg to add_image. There is otherwise no way for us to know that the last axis is channels. (yes yes, ITK and typing is so good, we know. 😂)

[thewtex]

@andy-sweet @jni @sofroniewn thank you! 💯