Variance and mask as channel by default - 2

[YooSunYoung]

Fixes #25
Inherits #26

I just cleaned up everything...

[Copilot]

Pull Request Overview

Updates the I/O layer to make multi-channel concatenation the default, remove legacy warnings, and align tests and docs with the new behavior. Key changes include:

• Change default concat_stdevs_and_mask to True and resolve_channels to True in load/save functions.
• Remove _warn_about_multi_channel_images and related test contexts.
• Add _has_variances_or_masks helper, simplify channel validation/resolution logic, and strengthen fallback on read errors.
• Update test expectations and user-guide examples to reflect new defaults.

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
tests/io_test.py	Cleaned up warning contexts, adjusted indentation of asserts
src/scitiff/io.py	Changed default parameters, removed warning helper, adjusted channel logic, added fallback except clause
docs/user-guide/io.ipynb	Updated examples to include new default args and flags

Comments suppressed due to low confidence (2)

src/scitiff/io.py:787

• The generic type T is used in the function signature but never defined. Introduce a TypeVar declaration, e.g., T = TypeVar('T', sc.DataArray, sc.DataGroup), before using it.

        with values of 'intensities', 'stdevs', and 'mask' (see :class:`~.Channel`).

docs/user-guide/io.ipynb:332

• [nitpick] Add a space before (default) for readability: change True`(default) to True (default).

    "Or you can simply set the `resolve_channels` argument to be `True`(default) and the loader will try reassembling the data array with mask and variances."

[jokasimr]

There are a lot of changes to warnings and similar, and I'm not so in the loop on scitiff, so I found it hard to distinguish the changes to the logic. Can you point out the sections of the code where the most important changes are located? Or you can show me in person tomorrow maybe?

[jokasimr]

I think I got the change now, is it correct to say the main change is that the default value of the concat_stddevs_and_masks changes to True?

What does it mean in this context to concat stddevs and masks?
Do we need to concat stddevs and mask because tiff only supports having one big array containing all "channels" of the image?

[YooSunYoung]

@jokasimr

I think I got the change now, is it correct to say the main change is that the default value of the concat_stddevs_and_masks changes to True?

That was fast 😃 ...
Yes, we want to keep the channel handling by default.
So it's not just for concatenating, it's also about splitting them into intensities, stdevs and a mask.

What does it mean in this context to concat stddevs and masks?
Do we need to concat stddevs and mask because tiff only supports having one big array containing all "channels" of the image?

We decided to use channel dimension for stdevs and mask if they are really needed to be stored with the intensities and we don't want to stored them as a plain text.

And tiff only allows same sizes of images across all dimensions so we are concatenating intensities, stdevs and a mask into channel dimension, which is useless for neutron or x-ray images.

I also updated the description of the issue #25 to include more context.

[jokasimr]

Why did this change from being an error to now being ignored? 🤔 If we are asked to resolve_channels but are unable to because we're missing the intensity data or because we don't recognize the channel names, should we not raise then?

[YooSunYoung]

As we decided to resolve the channel by default, you can't tell if it was intentionally set or not.

And as a result of balancing between user experiences and robustness,
we don't want the loader to fail in any cases as much as possible,
and we want to keep the save method as strict as possible.

I thought about setting the default value as None so that we can tell if it's specifically set by user or not.
Do you think we should do that...?

[jokasimr]

Aha I see. I think it makes sense to do it like you did, just wanted to double check that it was intentional to remove the errors there.