Add reverse_coords option to point conversion functions

[Copilot]

ITK uses x,y,z coordinate order while napari uses z,y,x (row-major NumPy). The point conversion functions now handle this automatically via a reverse_coords kwarg, enabled by default.

Changes

• points_layer_from_point_set: Added reverse_coords=True kwarg to reverse ITK (x,y,z) → napari (z,y,x)
• point_set_from_points_layer: Added reverse_coords=True kwarg to reverse napari (z,y,x) → ITK (x,y,z)
• Updated existing tests to expect reversed coordinates by default
• Added tests for reverse_coords=False, roundtrip conversions, and 2D points

Usage

from itk_napari_conversion import points_layer_from_point_set, point_set_from_points_layer

# Default behavior: coordinates are reversed automatically
points_layer = points_layer_from_point_set(itk_point_set)  # x,y,z → z,y,x
point_set = point_set_from_points_layer(napari_layer)      # z,y,x → x,y,z

# Opt out if needed
points_layer = points_layer_from_point_set(itk_point_set, reverse_coords=False)

Original prompt

Per:

• point set conversion #13 (comment)
• ENH: Support napari layers of corresponding points SuperElastix/elastix-napari#65 (comment)

Create a pull request that changes the order of points from x,y,z to z,y,z or the inverse, as appropriate, in points_layer_from_point_set, point_set_from_points_layer. Make the behavior a boolean kwarg, but enable by default. Add tests.

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

[N-Dekker]

Thanks very much for fixing the order of point coordinates, @thewtex

I find some parts of this pull request (adjusting ".gitignore" and adding/removing binaries for Python 3.12) hard to review. Are they essential for the "reverse_coords" fix? If so, can you place them in a separate commit? If they are unrelated, please move them into a separate pull request.

[thewtex]

@N-Dekker I'll get this published soon to ease your application in elastix-napari.

I've cleaned up the history and added the .gitignore changes to a new PR.

[N-Dekker]

This is the main part of the fix, right? Cool! I can't wait to try it out, with SuperElastix/elastix-napari#65 😃

[N-Dekker]

Ah, I see, the other "main" part of the fix 😃

[N-Dekker]

@thewtex For my understanding, is this proposed new reverse_coords argument intended to be a permanent feature, or is it meant as a workaround?

In practice, shouldn't the coordinates always be reverted anyway, when converting between a napari Points layer and an itk.PointSet?

[thewtex]

@thewtex For my understanding, is this proposed new reverse_coords argument intended to be a permanent feature, or is it meant as a workaround?

In practice, shouldn't the coordinates always be reverted anyway, when converting between a napari Points layer and an itk.PointSet?

@N-Dekker yes, it will be permanent. It is left as an option in case the client calling the function already made the change (using with a different library, etc.) to avoid the need and performance hit for doing it again.

[thewtex]

@N-Dekker published to PyPI as version 0.5.1