Update minimum versions in dependencies, remove scikit-image & slicerator dependency

[jakirkham]

It looks like a few dependencies had higher minimum versions when using setup.py

dask-image/setup.py

Lines 31 to 39 in 243fb0a

	requirements = [
	"dask[array] >=2024.4.1",
	"dask[dataframe] >=2024.4.1",
	"numpy >=1.18",
	"scipy >=1.7.0",
	"pims >=0.4.1",
	"tifffile >=2018.10.18",
	"pandas >=2.0.0",
	]

Though we wound up with older versions after switching to pyproject.toml

dask-image/pyproject.toml

Lines 26 to 35 in 49d9c33

	dependencies = [
	"dask[array,dataframe] >=2024.4.1",
	"numpy >=1.11.3",
	"scikit-image >=0.19.3",
	"scipy >=0.19.1",
	"pandas >=2.0.0",
	"pims >=0.4.1",
	"slicerator >= 1.1.0",
	"tifffile >=2018.10.18",
	]

Guessing this is just because the pyproject.toml PR has fallen out of date in this area and we missed it

So went ahead and bumped these. Though please let me know if I'm missing something

[jakirkham]

Also assuming scikit-image and slicerator were added intentionally. Please let me know if not and can update accordingly

[jakirkham]

@GenevieveBuckley could you please review?

[GenevieveBuckley]

Guessing this is just because the pyproject.toml PR has fallen out of date in this area and we missed it
So went ahead and bumped these. Though please let me know if I'm missing something

Yes, I think the PR just fell out of date between it being opened and merged.

Also assuming scikit-image and slicerator were added intentionally. Please let me know if not and can update accordingly

I added them because they are included in the environment yaml files. I was trying to make sure it wouldn't matter which method someone used to install it, you'd still get something that matches. That might not have been the best thing to do though.

I also thought for some reason that scikit-image was required for a new-ish function, but I can't actually see it being imported anywhere (including the tests). So maybe it shouldn't be there at all.

I think slicerator is a dependency of one of our other dependencies, so it might not be a thing we need to worry about in the pyproject.toml at all. Like I said, I was just trying to match the environment files, but that's maybe not the best strategy.

[GenevieveBuckley]

I also saw this issue from Mark about switching from dask to dask-core: conda-forge/dask-image-feedstock#16

[jakirkham]

No worries. It's better to move forward with changes (like pyproject.toml) and fix a few things afterwards than let it linger indefinitely. So think you did the right thing 🙂

Think scikit-image was added for testing originally. If that is all we are using it for still, it should be safe to drop. So went ahead and did that. If we want to use things from scikit-image in the future, we can always add it back then

Regarding slicerator that was added as a workaround ( #17 ). Think we are safe to drop it at this point. So went ahead and did that

The dask-core bit is only relevant for the conda-forge feedstock. So let's save that discussion for there

Maybe there is a better way to manage the dependencies here and call out what is needed for testing to make it clearer. Open to thoughts on that, but maybe we can discuss in a new issue?

[GenevieveBuckley]

Thank you for doing this @jakirkham

[jakirkham]

Happy to help 🙂