Index integer_array_t in NumPy like in VUnit

[ru551n]

An integer_array_t is now indexed in Python like in VUnit: get(a, x, y, z) is a[x, y, z], with shape (width, height, depth). Before, the indices of the first two axes were swapped, a[y, x, z] with shape (height, width, depth).

This is a breaking change, with a release note entry.

Array	VHDL	Python before	Python now
1D	get(a, i)	a[i]	a[i] (unchanged)
2D	get(a, x, y)	a[y, x], shape (height, width)	a[x, y], shape (width, height)
3D	get(a, x, y, z)	a[y, x, z]	a[x, y, z], shape (width, height, depth)

How

• Arguments: VUnit stores an array by y, then x, then z. An argument is still a view of the transferred storage, with no copy, and that view has its first two axes swapped.
• Returned arrays: they get the reverse swap before they're copied out, so width and height now come from shape[0] and shape[1].
• Word size: a function returning its own argument still keeps that argument's word size.

Things users will notice

• Indexing: code that indexes a 2D or 3D array in Python must swap its first two indices.
• Flattening: NumPy now flattens a 2D argument x first. a.reshape(-1) of a 3×2 array gives 0, 10, 1, 11, 2, 12 rather than VUnit's storage order. The test of a returned array with a new shape checks this.
• Image libraries: Pillow, OpenCV and numpy.loadtxt of a CSV file written by VUnit's save_csv index rows first. Such an array needs a.swapaxes(0, 1) (or a.T in 2D), which the user guide now says.

Changes

• runtime.py: make_array swaps the axes of the view, and _array_result swaps them back.
• tests/tb_python_pkg_bridge.vhd: the 2D/3D argument and result tests are indexed like VUnit now, and the flatten test has the new order.
• examples/embedded_python: the image_model.pixel docs and code, and the comment in tb_example.vhd.
• docs/types.rst: the indexing table, plus a paragraph on row-first image libraries.
• release_notes/unreleased.md: a breaking change entry.

Tests

Locally:

• pytest: 111 passed.
• NVC and GHDL: 109/109, plus file_base 7/7.
• The embedded_python example on NVC and GHDL: 18/18, with the CI filters.
• Sphinx -W: builds cleanly.

🤖 Generated with Claude Code