BUG: Fix image_from_array shape handling for F-contiguous arrays

[Copilot]

Description

itk.image_from_array() incorrectly preserved the original array shape when given F-contiguous inputs (e.g., array.T), while itk.image_view_from_array() correctly used the transposed shape. This inconsistency broke user expectations and made the two functions behave differently for the same input.

Root cause: Double reversal in shape handling

• Python code unconditionally reversed shape: ndarr.shape[::-1]
• C++ code reversed again for F-contiguous buffers
• Net effect: original shape preserved instead of transposed

Fix: Conditional shape reversal based on contiguity

• C-contiguous: reverse in Python (C++ expects ITK order)
• F-contiguous: pass as-is (C++ handles reversal)

This matches the existing vector image logic and aligns both functions.

Changes

• Modules/Bridge/NumPy/wrapping/PyBuffer.i.in: Added contiguity check before shape reversal in GetImageViewFromArray()
• Wrapping/Generators/Python/Tests/extras.py: Added tests for transposed array handling

Example

import itk
import numpy as np

array = np.empty((1, 2, 3))

# Before fix:
itk.image_from_array(array.T).shape    # (1, 2, 3) - WRONG
itk.image_view_from_array(array.T).shape  # (3, 2, 1) - correct

# After fix:
itk.image_from_array(array.T).shape    # (3, 2, 1) - correct
itk.image_view_from_array(array.T).shape  # (3, 2, 1) - correct

PR Checklist

• No API changes were made (or the changes have been approved)
• No major design changes were made (or the changes have been approved)
• Added test (or behavior not changed)
• Updated API documentation (or API not changed)
• Added license to new files (if any)
• Added Python wrapping to new files (if any) as described in ITK Software Guide Section 9.5
• Added ITK examples for all new major features (if any)

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Unexpected behavior of itk.image_from_array when passing array.T</issue_title>
<issue_description>

Hi, and thanks again for making ITK!

Description

The shape of an image created with itk.image_from_array is unexpected when the input is not a contiguous array. I discovered this when passing array.T and was surprised when the shape was the same as for array. The issue doesn't happen with itk. image_view_from_array.

Steps to Reproduce

import itk
import numpy as np

array = np.empty((1, 2, 3))

print(f"{array.shape = }")
print()
print(f"{itk.image_from_array(array).shape = }")
print(f"{itk.image_from_array(array.T).shape = }")
print(f"{itk.image_from_array(array.T.copy()).shape = }")
print()
print(f"{itk.image_view_from_array(array).shape = }")
print(f"{itk.image_view_from_array(array.T).shape = }")
print(f"{itk.image_view_from_array(array.T.copy()).shape = }")
print()
print(f"{array.flags['C_CONTIGUOUS'] = }")
print(f"{array.T.flags['C_CONTIGUOUS'] = }")
print(f"{array.T.copy().flags['C_CONTIGUOUS'] = }")
print()
print(f"{array.flags['F_CONTIGUOUS'] = }")
print(f"{array.T.flags['F_CONTIGUOUS'] = }")
print(f"{array.T.copy().flags['F_CONTIGUOUS'] = }")

Expected behavior

I expected reversed shapes for itk.image_from_array(array) and itk.image_from_array(array.T).

Actual behavior (output of script above)

array.shape = (1, 2, 3)

itk.image_from_array(array).shape = (1, 2, 3)
itk.image_from_array(array.T).shape = (1, 2, 3)  # I expected (3, 2, 1) here!
itk.image_from_array(array.T.copy()).shape = (3, 2, 1)

itk.image_view_from_array(array).shape = (1, 2, 3)
itk.image_view_from_array(array.T).shape = (3, 2, 1)
itk.image_view_from_array(array.T.copy()).shape = (3, 2, 1)

array.flags['C_CONTIGUOUS'] = True
array.T.flags['C_CONTIGUOUS'] = False
array.T.copy().flags['C_CONTIGUOUS'] = True

array.flags['F_CONTIGUOUS'] = False
array.T.flags['F_CONTIGUOUS'] = True
array.T.copy().flags['F_CONTIGUOUS'] = False

Reproducibility

Always.

Versions

$ python -c "import numpy; print(numpy.__version__)"
1.26.2

$ python -c "import itk; print(itk.Version.GetITKVersion())"
5.3.0

Environment

$ python <(curl -s https://raw.githubusercontent.com/fepegar/torchio/main/print_system.py)
Platform:   macOS-14.0-arm64-arm-64bit
[...]
Python:     3.11.5 (main, Sep 11 2023, 08:31:25) [Clang 14.0.6 ]

</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes Unexpected behavior of itk.image_from_array when passing array.T #4329

---

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