[peterdsharpe]

Mesh.with_cells rescales each effective measure by the new cell area over the old one. When a row only reorders its own vertices, as in a winding flip, the cell is the same simplex. Its measure should not change, but float32 areas differ between vertex orders.

import torch
from physicsnemo.mesh import Mesh
from physicsnemo.mesh.calculus.measure import cell_measures, set_cell_measures

mesh = Mesh(
    points=torch.tensor([[0.0, 0.0, 0.0], [1.0, 0.0, 0.0], [2.0, 0.0, 0.0]]),
    cells=torch.tensor([[0, 1, 2]]),
)
set_cell_measures(mesh, torch.tensor([1.5]))
flipped = mesh.with_cells(mesh.cells[:, [0, 2, 1]])
# main:    ValueError: Cannot transform nonzero effective measure on a zero-measure cell
# this PR: cell_measures(flipped) is tensor([1.5000])

A winding flip, with each measure set to twice its cell's area (1.5 for the zero-area cell):

Cells	main	This PR
10,000 random triangles	26% of the measures change, by up to 2.7e-6 relative	unchanged, bit for bit
100,000 near-degenerate triangles	measures scale by 0.022 to 26	unchanged, bit for bit
One zero-area cell with a nonzero measure	ValueError	unchanged

fix_orientation flips cells through with_cells, so it raised on such a cell too.

with_cells now marks each row whose sorted vertex indices equal the old row's. Those rows keep their effective measure and cached cell area bitwise, so a later transform's area ratio starts from the same pair. They also keep a cached centroid. Rows whose vertices change keep the area-ratio transfer and its error. The selection uses torch.where, adds no host sync, and writes no cache while tracing or when points require gradients.

New tests cover flips of subsampled meshes and slivers, flips and cyclic shifts of a zero-area cell, changed rows, cached centroids, unsigned and empty replacements, keep, gradients, torch.compile and fix_orientation.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.
• The CHANGELOG.md is up to date with these changes.
• An issue is linked to this pull request.
• If I am implementing a new model or modifying any existing model, I have followed the Models Implementation Coding Standards.

Dependencies

None.

Review Process

All PRs are reviewed by the PhysicsNeMo team before merging.

Depending on which files are changed, GitHub may automatically assign a maintainer for review.

We are also testing AI-based code review tools (e.g., Greptile), which may add automated comments with a confidence score.
This score reflects the AI’s assessment of merge readiness and is not a qualitative judgment of your work, nor is
it an indication that the PR will be accepted / rejected.

AI-generated feedback should be reviewed critically for usefulness.
You are not required to respond to every AI comment, but they are intended to help both authors and reviewers.
Please react to Greptile comments with 👍 or 👎 to provide feedback on their accuracy.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

CODEOWNERS review map

Current for commit b0e979a7411e. An approval covers every file listed for that owner; one owner is sufficient for shared files.

⏳ @peterdsharpe — 5 file(s)

• physicsnemo/mesh/calculus/measure.py
• physicsnemo/mesh/mesh.py
• test/mesh/calculus/test_measure.py
• test/mesh/repair/test_orientation.py
• test/mesh/utilities/test_cache.py

No CODEOWNER

• CHANGELOG.md

Comment /codeowners-info to refresh.

[greptile-apps]

[Medium impact] The change affects mesh connectivity replacement and orientation repair.

Fix the unsigned-index comparison before merging.

Findings

1.  Unsigned replacement cells fail ▶

Summary

This PR preserves effective measures when a cell’s vertices are reordered, including zero-area cells.

• Reordered cell vertices keep the same measure and cached geometry.

_{Reviews (1) · Last reviewed commit: "Keep measures of reordered cells in Mesh..." · Reviewed by Greptile}

[mnabian]

LGTM. Left a few minor comments.

[peterdsharpe]

/ok to test 2f72cd5