Delete ObjAdapter's copy of the mesh inertia integral; SymbolicAWEModels owns it

[1-Bort-1]

TL;DR

center_to_com!, calculate_inertia_tensor and calc_inertia_y_rotation are removed from ObjAdapter, with their export, their tests and their private_functions.md entries. SymbolicAWEModels computes mesh mass properties from its own copies of the same integral, so two packages held one piece of physics and nothing would ever have reported that they had drifted apart.

What changes for a caller

The three names no longer exist and are no longer exported; read_faces, which is all SymbolicAWEModels actually takes from ObjAdapter, is untouched. The changelog entry under ## Unreleased → ### Changed is BREAKING:.

An earlier round of this branch deprecated the three instead, on review; review then said the next release is breaking anyway, so that round is reverted and they go.

Who uses them

Nobody. The search was repeated on 2026-09-20, immediately before this push: gh search code for each of the three names across OpenSourceAWE and 1-Bart-1, then a shallow clone and rg of all 43 repositories in those two accounts, private ones included. No importer. SymbolicAWEModels (src/obj_adapter.jl:50, src/system_structure/rigid_body.jl:191) and StableAWEModels (src/system_structure/rigid_body.jl:180) define their own copies and call those; SymbolicAWEModels' only call into here is ObjAdapter.read_faces at src/obj_adapter.jl:75. BeyondTheSim examples/neuralfoil.jl:67 does a bare using VortexStepMethod.ObjAdapter, which would have pulled the three names into scope, and calls none of them.

Tests

Their testsets go with them: "Center of Mass Calculation", "Inertia Tensor Calculation" and "Alignment to principal frame" in test_kite_geometry.jl, and "center_to_com! rejects non-triangular faces" in test_obj_adapter.jl. The "Interpolation Creation" testset also serialized an _info.bin of inertia tensor, COM and interpolants that nothing ever read; that block is gone too, and with it Serialization from test/Project.toml and Interpolations from the test's imports — the file's last user of each. The .obj written there stays, because "OBJ File Reading" reads it back.

Found on the way, not changed here

• find_circle_center_and_radius and create_interpolations, the other two ObjWing leftovers in obj_geometry.jl, have no caller in src/ or ext/ either: only their own testsets call them. They are unexported, so deleting them is not breaking, but it wants its own importer search and belongs in its own diff.

Verification

• Reproduced first: n/a — a deletion. The evidence that it is safe is the code search above, not a failing test.
• test/ram_geometry/test_kite_geometry.jl: 17 pass, 1 broken (the pre-existing @test_skip), exit 0 in a fresh --project=test process
• test/obj_adapter/test_obj_adapter.jl: 51/51 pass · test/Aqua.jl: 10/10 pass
• Up to date with main: merged origin/main @ 0c94c6d (Give every untyped empty container its element type #338's typed read_faces, Build a Solver from VSMSettings or from panel and section counts, deprecate the body_aero constructors, and check the sizes in solve! #340, set_va! turns the body about a stored reference_point, and set_va!(body_aero, settings) applies yaw_rate #353), no conflict
• Local full suite (agent ci-local, Julia 1.13.0, one matrix cell): started 18:58Z, result not in when this was written
• Docs: no new public symbol; two private_functions.md entries removed with their docstrings, and every remaining bare name there still resolves. The build itself is the CI Documentation job — the docs/ environment has no manifest to build from locally.
• REUSE lint: n/a, the repo does not use REUSE
• Benchmark: n/a
• Risk: a caller outside these two GitHub accounts, which no search of mine can see.

Scope

+6 / −214 across 7 files against main, all of it deletion apart from the three-line changelog entry. #350 also edits test_obj_adapter.jl, and #320 and #346 also edit docs/src/private_functions.md; every open PR touches CHANGELOG.md. Whichever merges later may need a small textual merge.

Closes #324 · task VortexStepMethod.jl-324

[1-Bort-1]

Independent review (advisory)

Verdict: APPROVE · 0 inline, 0 off the diff

Good

• The diff matches the card: all three functions are removed along with their export, tests and docs entry, and read_faces stays. The commit stat is +11/−213 across 7 files, as the card says.
• No uses remain: grepping src/, test/, docs/ and *.toml finds no reference to center_to_com!, calculate_inertia_tensor or calc_inertia_y_rotation. Only the untracked .agent/ notes mention them.
• using NonlinearSolve is still needed because find_circle_center_and_radius calls NewtonRaphson at obj_geometry.jl:89. LinearAlgebra is still needed for norm at obj_geometry.jl:211 and in obj_slice.jl.
• It is safe to drop Serialization from test/Project.toml: no test uses serialize now, and the root Project.toml:31 still lists it, so the manifest does not change.
• The cleanup in the "Interpolation Creation" testset keeps behaviour: the faces, .obj write and _info.bin it removes are not used later in the testset. The read_aero_matrix and create_interpolations assertions are unchanged.
• No test was weakened: the only removed assertions tested the deleted functions, and the existing @test_skip is untouched.
• This is removal the rubric asks for: §2 wants one source per equation and nothing kept in case, and the other copy lives in SymbolicAWEModels, which takes only read_faces from here.
• The BREAKING: changelog entry goes under a new ## Unreleased, and the version in Project.toml is not bumped.

Not good

• The changelog line says the names are no longer exported and then "they are gone", which says the same thing twice. "ObjAdapter drops center_to_com!, …" would be enough.
• The docs build was not run locally, so the CI Documentation job is the only check. The risk is low: only two @docs entries were removed, and calc_inertia_y_rotation had no docstring or docs entry.
• test_obj_path is now written and read only by the "OBJ File Reading" testset. It could move inside that testset, but that is optional.

_{claude, rubric CLEAN_CODE.md. A different lab from the implementer
on purpose: a reviewer sharing its blind spots would not flag its mistakes.}

[1-Bort-1]

Local full suite: PASS (5 min, Julia 1.13.0, one cell of the matrix)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[1-Bart-1]

Keep the functions, mark them deprecated, and rm them in the next breaking release, because otherwise we have too many breaking changes.

[1-Bort-1]

Done in e344c8e: the three functions are back with their export, tests and docs, each emits a deprecation warning, and the changelog entry is a deprecation instead of BREAKING.

[1-Bart-1]

Ok fine we are going breaking already anyways. Just mark them breaking and not backwards compatible.

[1-Bart-1]

So remove them anyways.

[1-Bort-1]

Done in 8880556: the changelog entry is BREAKING again.

[1-Bort-1]

Removed in 8880556, reverting the deprecation round.