CHANGELOG.md (1)

121-121: Move this changelog item from Fixed to Added.

Line 121 describes new functionality, not a bug fix, so it should be listed under ### Added in [Unreleased].

Suggested changelog adjustment

 ### Added
@@
+- Import `UsdGeom.TetMesh` prims through `ModelBuilder.add_usd()` as soft meshes
@@
 ### Fixed
@@
-- Import `UsdGeom.TetMesh` prims through `ModelBuilder.add_usd()` as soft meshes

As per coding guidelines: user-facing changes must be placed under the correct [Unreleased] category in CHANGELOG.md.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@CHANGELOG.md` at line 121, The changelog entry "Import `UsdGeom.TetMesh`
prims through `ModelBuilder.add_usd()` as soft meshes" is under `Fixed` but
describes new functionality; move that line from the `### Fixed` section into
the `### Added` section under the `[Unreleased]` header in CHANGELOG.md so it
appears as an addition, not a bug fix, and ensure formatting/indentation matches
other `### Added` items.

newton/tests/test_import_usd.py (1)

7751-7792: Consider extending this transform test to include rotation and non-uniform scale.

Current coverage is translation-only. Adding one case with nested parent xforms + non-uniform scale would better lock in the TetMesh transform decomposition behavior introduced by this PR.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@newton/tests/test_import_usd.py` around lines 7751 - 7792, Update the
test_add_usd_imports_tetmesh_with_transforms case to cover rotation and
non-uniform scale by adding nested Xform hierarchy and non-uniform scale ops in
the USD payload and by calling builder.add_usd with a wp.transform that includes
a non-identity quaternion and non-uniform scale (instead of wp.quat_identity and
uniform translation only); specifically modify the
stage.GetRootLayer().ImportFromString USD to include a parent Xform with rotate
and scale ops (unique symbol: the TetMesh "SoftBody" under World) and change the
builder.add_usd(xform=...) invocation to pass a transform with rotation and
non-uniform scale, then update the expected positions array to the transformed
vertex coordinates and assert against builder.particle_q as before.

newton/_src/utils/import_usd.py (1)

2409-2414: Clarify precedence when both mesh and vertices are provided.

When scale is non-uniform, the code passes both mesh=tetmesh and vertices=tetmesh.vertices * scale. This relies on add_soft_mesh preferring the explicit vertices kwarg over mesh.vertices.

Consider adding a brief inline comment to clarify this expected behavior, or verify the contract with add_soft_mesh.

             if _is_uniform_scale(soft_mesh_scale):
                 add_soft_mesh_kwargs["scale"] = float(np.array(soft_mesh_scale, dtype=np.float32)[0])
             else:
+                # When non-uniform scale is present, pre-scale vertices.
+                # add_soft_mesh uses explicit `vertices` kwarg over mesh.vertices.
                 add_soft_mesh_kwargs["vertices"] = tetmesh.vertices * np.array(soft_mesh_scale, dtype=np.float32)

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@newton/_src/utils/import_usd.py` around lines 2409 - 2414, The code currently
passes both mesh=tetmesh and vertices=tetmesh.vertices * scale when
_is_uniform_scale(soft_mesh_scale) is False, relying on add_soft_mesh to prefer
the explicit vertices kwarg; add a short inline comment next to the
add_soft_mesh_kwargs construction (referencing _is_uniform_scale,
add_soft_mesh_kwargs, tetmesh, mesh, vertices, and builder.add_soft_mesh)
stating that add_soft_mesh must prefer vertices over mesh.vertices, and
optionally add a defensive check (e.g., assert or explicit override) to verify
the add_soft_mesh contract before calling builder.add_soft_mesh.

newton/_src/utils/import_usd.py (1)

2425-2429: Minor cleanup: avoid repeated scale conversion per TetMesh prim.

np.array(soft_mesh_scale, dtype=np.float32) is constructed multiple times in the same block; compute once for clarity and less allocation churn.

♻️ Suggested refactor

             add_soft_mesh_kwargs = {
                 "pos": soft_mesh_pos,
                 "rot": soft_mesh_rot,
                 "scale": 1.0,
                 "vel": wp.vec3(0.0, 0.0, 0.0),
                 "mesh": tetmesh_for_builder,
             }
+            soft_mesh_scale_np = np.asarray(soft_mesh_scale, dtype=np.float32)
             if _is_uniform_scale(soft_mesh_scale):
-                add_soft_mesh_kwargs["scale"] = float(np.array(soft_mesh_scale, dtype=np.float32)[0])
+                add_soft_mesh_kwargs["scale"] = float(soft_mesh_scale_np[0])
             else:
-                add_soft_mesh_kwargs["vertices"] = tetmesh_for_builder.vertices * np.array(soft_mesh_scale, dtype=np.float32)
+                add_soft_mesh_kwargs["vertices"] = tetmesh_for_builder.vertices * soft_mesh_scale_np

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@newton/_src/utils/import_usd.py` around lines 2425 - 2429, Compute the NumPy
array for soft_mesh_scale once and reuse it to avoid repeated allocations:
create a local variable (e.g., scale_arr = np.array(soft_mesh_scale,
dtype=np.float32)) before the if checking _is_uniform_scale(soft_mesh_scale),
then use scale_arr[0] when setting add_soft_mesh_kwargs["scale"] and multiply
tetmesh_for_builder.vertices by scale_arr when setting
add_soft_mesh_kwargs["vertices"]; update the logic around _is_uniform_scale,
add_soft_mesh_kwargs, and tetmesh_for_builder.vertices to reference this single
precomputed array.

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes