Martin Schorb

> @martinschorb Did you install with `pip install dash[testing]` ? `dash` was installed from `conda-forge`

is the `[testing]` also required when using conda?

OK, that makes sense. What confuses me are the brackets. When using conda: `conda install "dash[testing]" -c conda-forge` I get: ``` Looking for: ['dash'] ... All requested packages already installed...

I have some weird behaviour that I don't fully understand (this is my first time using pytest). - when I do not pass `output_chunks` on to `affine_transform` the test run...

OK, it seems there is also a test failure with the `[1,1,1]` chunk size when the timeouts actually work.

> @martinschorb Were you having anything specific in mind regarding how to best implement these functions for `dask_image`? I was playing around with `rotate` and found that it accepts dask...

So, one issue I discovered when looking at `rotate` and also with the other high-level functions is the `output` parameter. https://github.com/scipy/scipy/blob/0a8e3d226f3c6bc24451e9cd4b00543ddf673c31/scipy/ndimage/interpolation.py#L937 This calls a constructor resulting in a `np.array` of...

Hi Marvin, > * Stick to the `scipy` API and don't support differently shaped output arrays (and exclude the above test) or > > * Include an `output_shape` argument to...

I am still getting slightly wrong results when comparing an image that is transformed with the `output_shape` set and the respective sub-area of a transformed image without. It looks like...

> @gcaria I think that's a good idea! In the case of `dask_image.ndinterp.affine_transform` the shapes of the input and output images can strongly vary, and there's no reason to assume...