contours() emits NaN coordinates for rasters containing +inf/-inf

[brendancol]

Description

contours() puts NaN coordinates into its output when the input raster has a +inf or -inf cell and you ask for a finite level near that cell.

The marching-squares kernel skips quads with a NaN corner using if tl != tl or tr != tr or bl != bl or br != br: continue (xrspatial/contour.py:67). That catches NaN but not infinity, since inf != inf is False. So a quad with an infinite corner is processed: the corner classifies as above any finite level via >=, and the edge interpolation t = (level - tl) / (tr - tl) either divides by an infinite denominator or computes inf - inf, which produces NaN in the emitted segment.

Reproduce

import numpy as np
from xrspatial.contour import contours
from xrspatial.tests.general_checks import create_test_raster

data = np.array([
    [0., 0., 0., 0., 0.],
    [0., 1., 1., 1., 0.],
    [0., 1., np.inf, 1., 0.],
    [0., 1., 1., 1., 0.],
    [0., 0., 0., 0., 0.],
], dtype=np.float64)
agg = create_test_raster(data, backend='numpy')
result = contours(agg, levels=[1.5])
for level, coords in result:
    print(coords)

The output contains NaN rows such as [nan, 1.0] and [1.0, nan].

Expected

A quad touching an infinite corner should be skipped the same way a NaN quad is, so the output never contains non-finite coordinates. The skip guard at line 67 needs to reject inf as well as NaN (e.g. a finiteness check on the four corners).

Notes

Found during a test-coverage audit of the contour module. A separate test-only PR adds an xfail test documenting this behavior; this issue tracks the source fix.

[Melissari1997]

Hi @brendancol, I came across xarray-spatial and wanted to see if I could contribute :)
I tried to fix this here: #2749

[brendancol]

@Melissari1997 Hey thanks so much and excited for your contribution!