When a surface has a RadialAperture with r_min > 0 (annular aperture), Surface2D._compute_sag() sets clipped coordinates to NaN. This works correctly for line plots (ax.plot() treats NaN as line breaks), but Lens2D._plot_single_lens() passes these NaN-contaminated vertices directly to matplotlib.patches.Polygon, which has undefined fill behavior with NaN vertices. The result is a spurious diagonal line across the lens body and, when annular and non-annular surfaces share a lens group, missing polygon chunks.

Fixes #538

Changes

• optiland/visualization/system/lens.py: In _plot_single_lens, detect NaN values in the vertex array. When present, split into contiguous non-NaN segments and create a separate Polygon for each valid segment. When no NaN values exist, behavior is unchanged.
• tests/visualization/test_visualization.py: Add TestAnnularApertureVisualization with two regression tests that build a lens matching the reproduction case from 2D layout: annular aperture (RadialAperture with r_min) causes polygon fill artifacts in Lens2D #538 and verify that (a) no polygon patch contains NaN vertices, and (b) polygon patches are still rendered.

Testing

Baseline (master, before patch):

47 passed (45 existing + 2 new regression tests would FAIL on unpatched code)

Patched:

47 passed in 13.63s

Ruff linting and formatting both pass on the changed files.

============================= test session starts =============================
platform win32 -- Python 3.13.5, pytest-9.0.3, pluggy-1.6.0
rootdir: D:\GITHUB_PR AUTO\optiland
configfile: pyproject.toml
collected 47 items

tests/visualization/test_visualization.py ... 45 existing PASSED
tests/visualization/test_visualization.py::TestAnnularApertureVisualization::test_draw_annular_aperture_no_nan_in_patches[backend=numpy] PASSED
tests/visualization/test_visualization.py::TestAnnularApertureVisualization::test_draw_annular_aperture_produces_patches[backend=numpy] PASSED

============================= 47 passed in 13.63s =============================