Support sequences

[hanjinliu]

Trying to add typing supports of such as list[str] and tuple[int, int].
These examples are how it is intended to be used.

@magicgui
def random_2d_image(shape: tuple[int, int] = (100, 100), max: float = 1) -> ImageData:
    return np.random.random(shape) * max

@magicgui
def gaussian_filter(img: ImageData, sigma: list[float] = [1.,]) -> ImageData:
    if len(sigma) == 1:
        sigma = sigma[0]
    return skimage.filters.gaussian(img, sigma)

Still need more strict typing to pass the tests.

[tlambert03]

i like it! thanks! let me know when you're ready for a full review

[codecov]

Codecov Report

Merging #342 (9c72f7c) into main (1890143) will increase coverage by 0.31%.
The diff coverage is 88.17%.

@@            Coverage Diff             @@
##             main     #342      +/-   ##
==========================================
+ Coverage   88.33%   88.65%   +0.31%     
==========================================
  Files          30       30              
  Lines        3609     3790     +181     
==========================================
+ Hits         3188     3360     +172     
- Misses        421      430       +9     

Impacted Files	Coverage Δ
magicgui/widgets/__init__.py	100.00% <ø> (ø)
magicgui/widgets/_concrete.py	85.65% <87.70%> (+1.01%)	⬆️
magicgui/type_map.py	95.59% <100.00%> (+1.44%)	⬆️
magicgui/backends/_qtpy/widgets.py	86.33% <0.00%> (+0.26%)	⬆️
magicgui/widgets/_bases/widget.py	87.84% <0.00%> (+0.55%)	⬆️
magicgui/_type_wrapper.py	67.52% <0.00%> (+2.95%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1890143...9c72f7c. Read the comment docs.

[hanjinliu]

Hi @tlambert03,
I think things are almost done. Could you check my PR?

A few things that I cannot decide by myself are:

• Should we also support typing like Sequence[int] or Iterable[int]? The new type matcher only detects list[...] and tuple[...] (and their subclasses).
• For ListEdit, widget options of the child widgets can be set using "options" argument. For instance, options of the int in list[int] is editable by ListEdit(value=[0, 1, 2], options={"max": 10}). But there is no elegant way for TupleEdit since its child widgets could be different (like tuple[int, str]).

[hanjinliu]

Sorry, @tlambert03, the last commit is a separated one... (concerning #361)
Can I split them into another PR in some way?

[tlambert03]

Sorry, @tlambert03, the last commit is a separated one... (concerning https://github.com/napari/magicgui/issues/361)
Can I split them into another PR in some way?

Hey @hanjinliu, sorry for the delay here. I'm ready to take a look at this.

Note that https://github.com/napari/magicgui/pull/362 is solving #361, and a lot of other issues with forward refs. But will likely create conflicts here. Perhaps we should merge that one before coming back to the type_map updates in this PR? I can help with that merge if you'd like.

I'll do a more thorough review once that is resolved, and we can get this in for the next release, which I'd like to do soon.

[hanjinliu]

Hi @tlambert03, I've read the upcoming changes in type mapping.
I have no idea how I can pass the "segmentation fault" thing... could you help me solve this?

[hanjinliu]

I was misreading a little bit few days ago... but I updated things for the new TypeWrapper.
I hope I'm not doing anything wrong.

[gselzer]

@hanjinliu this is super awesome! I'd love to use this in my plugin, what is the status of this PR?

[tlambert03]

what is the status of this PR?

waiting on me/review at this point

[tlambert03]

so sorry for the delay on this @hanjinliu. been busy and haven't found the time to give this the close look I wanted to. Since it's pretty much just a new feature, (and at first glance looks to be nicely implemented!) I'm going to just merge this so people like @gselzer can take a crack at it. We can iterate as needed if anything arises. Thanks again for your contribution and your patience!

[hanjinliu]

Hi, @tlambert03
Thank you for the full review! Yes, I totally agree with that. Let's see how it goes well with plugin development.

[gselzer]

@hanjinliu @tlambert03 this seems to work great! Thanks for working on this!

This allows us to finally set up variable length arrays and lists in napari-imagej!

Here's an early look at a N-Dimensional gaussian blur with separate dimensional sigmas: