shape/make_valid consistency

[gmakarowski-yara]

Expected behavior and actual behavior.

Consider the following examples:

from shapely.geometry import mapping, shape
from shapely.validation import explain_validity, make_valid
import numpy as np
g = {
    'type': 'Polygon',
    'coordinates': [[[30.0, 10.0, np.NaN], [40.0, 40.0, np.Inf], [20.0, 40.0], [10.0, 20.0],
                     [30.0, 10.0]]]
}
s = shape(g)

Result:

ValueError: Inconsistent coordinate dimensionality

g = {
    'type': 'LineString',
    'coordinates': [[30.0, 10.0, np.NaN], [40.0, 40.0, np.Inf]]
}
s = shape(g)
make_valid(s)

Result:

<LINESTRING (30 10, 40 40)>

I would expect either shape() to raise an error in both cases or to silently accept the invalid coordinate component and then fix in in the make_valid().

Right now geometries that allow the user initialize with a NaN are: Point, LineString, MultiPoint
make_valid() fixes the error in all of these cases.

Geometries that raise a ValueError: Polygon, MultiLineString, MultiPolygon.

Steps to reproduce the problem.

See the example from the previous section.

Operating system

macOS Monteray 12.6.2 (21G320)

Shapely version and provenance

Shapely 2.0.4, installed from PyPI using pup