SCRIP 0/360 seam duplicates nodes, giving euler_characteristic = 1 on a closed sphere

[rajeeja]

Version

2026.9.0 (and main)

How did you install UXarray?

Source

What happened?

A closed global SCRIP mesh reports euler_characteristic = 1 instead of 2.

SCRIP files store corner longitudes in 0..360 and include both endpoints of the seam: outCSne8 has 30 corners at lon=0 and 30 at lon=360. Those are the same meridian, but the reader's dedup compares (lon, lat) pairs literally, so the seam gets two sets of nodes and the mesh is topologically split along it.

mesh	duplicate node coords	V	E	F	χ
outCSne8	15	407	790	384	1
ne30pg2	119	21,727	43,326	21,600	1
HEALPix z3 (control)	0	770	1,536	768	2

The HEALPix control has no duplicates and gives the right answer, which locates the problem in the SCRIP path rather than in the Euler computation.

Both meshes are reported as closed: true, so a caller gets a closure verdict that disagrees with the characteristic printed beside it.

What did you expect to happen?

χ = 2 for a closed sphere, and nodes on the 0/360 seam merged.

Can you provide a MCVE to reproduce the bug?

import numpy as np, uxarray as ux

for name, path in [
    ("outCSne8", "test/meshfiles/scrip/outCSne8/outCSne8.nc"),
    ("ne30pg2", "test/meshfiles/scrip/ne30pg2/grid.nc"),
]:
    g = ux.open_grid(path)
    V, E, F = int(g.n_node), int(g.n_edge), int(g.n_face)
    nodes = np.stack([np.round(g.node_lon.values, 9),
                      np.round(g.node_lat.values, 9)], 1)
    dupes = V - len(np.unique(nodes, axis=0))
    print(f"{name}: V={V} E={E} F={F} chi={V-E+F} duplicate_nodes={dupes}")

g = ux.Grid.from_healpix(3)          # control: no duplicates
V, E, F = int(g.n_node), int(g.n_edge), int(g.n_face)
print(f"healpix3: V={V} E={E} F={F} chi={V-E+F}")

Note

Same class as #1692 (coincident nodes), but the trigger here is specifically the 0/360 seam in SCRIP rather than general coincidence, and it reproduces on the meshes already in test/meshfiles. Unchanged by #1775 — that PR swaps the dedup algorithm but compares the same (lon, lat) pairs, so it inherits the behaviour rather than causing it.