Inconsistent "Inconsistent coordinate dimensionality" error

As found in #963, there is a bit of inconsistency in how mixed 2D and 3D coordinates/points are used to create LineString objects.

These are expected errors:

from shapely.geometry import LineString, Point
import pytest

with pytest.raises(ValueError, match="Inconsistent coordinate dimensionality"):
    LineString([(1, 2, 3), (4, 5)])

with pytest.raises(ValueError, match="Inconsistent coordinate dimensionality"):
    LineString([Point(1, 2, 3), Point(4, 5)])

However these should also raise the error too:

with pytest.raises(ValueError, match="Inconsistent coordinate dimensionality"):
    LineString([(1, 2), (3, 4, 5)])

with pytest.raises(ValueError, match="Inconsistent coordinate dimensionality"):
    LineString([Point(1, 2), Point(3, 4, 5)])

[jorisvandenbossche]

Ah, a first regression report! 😉 Can you open a new issue with that failing example? (we clearly were not testing this case)

[jorisvandenbossche]

Indeed, I think that’s a correct analysis.

Now for a solution, note that pygeos already checks for this (although the error message is not very clear / user friendly, not sure we can do something about that …):

In [3]: pygeos.linestrings([(1, 2), (4, 5)])  
Out[3]: <pygeos.Geometry LINESTRING (1 2, 4 5)>

In [4]: pygeos.linestrings([(1, 2, 3), (4, 5)]) 
...
ValueError: linestrings: Input operand 0 does not have enough dimensions (has 1, gufunc core with signature (i, d)->() requires 2)

In [5]: pygeos.linestrings([(1, 2), (3, 4, 5)])  
...
ValueError: linestrings: Input operand 0 does not have enough dimensions (has 1, gufunc core with signature (i, d)->() requires 2)

So this will be fixed in Shapely 2.0 (and not necessarily worth doing a separate fix in 1.8, or at least not priority, I would say)