Requirements map and the next solid-fluid geometry contract
The first table records the earlier PR24/main and bounded static-obstacle checkpoint; the current reviewed runtime composition is described below. The new native wall/force sequence is qualified within its fixed-column model; the broader liquid capability is incomplete. Source links below freeze the inspected PR24 source, so a future implementation cannot silently change these claims.
| User requirement | Actual implementation | Research/teaching coverage | Remaining capability |
|---|---|---|---|
| Collision meshes | TriangleSurface provides immutable two-sided earliest segment queries/clipping and stamps; translation has its own prescribed-motion query scope. Open/disconnected/intersecting surfaces are admitted. | Chapter20 explains shared cut volumes, face openings and connectivity. BoundedPhysics proves infinite-plane crossing/clipping, not arbitrary triangle-solid assembly. | StaticObstacleGeometry now admits a single closed axis-aligned box and derives shared volumes, openings, labels and conservative flux from the retained surface. General mesh cuts remain absent. A bounded follow-on now adds sealed obstacle pressure and a separate reduced extruded-shear response from the same owner; general 3D coupling remains absent. A hit normal alone is not fluid topology. |
| Forces | BodyForce / ForceUnits supports explicit acceleration/force-density vectors and optional regions in the fixed box; LiquidStepInputs borrows them. PR24 adds uniform tangential forcing to the separate constrained slab. | Chapter19 explains external and moving-boundary work; the Poiseuille lab separates fixed-q and fixed-a density controls. | Mesh-based traction, moving-solid displaced volume/work and two-way solid impulse/torque are not supplied by these force APIs. |
| Viscosity | ViscosityWorkspace assembles symmetric Newtonian strain on a fully filled sealed MAC box with stationary free-slip walls. step_viscous retains that domain restriction. PR22–24 qualify a separate layer-uniform fixed slab with Navier/no-slip. | Chapter22 and finite-strain Lean identities distinguish dissipation from stencil correctness; fitted/coupled prototypes retain separate research scopes. | The same static owner now supplies a strictly admitted reduced shear channel, with actual cut-layer inertia, traction, forced response and work/impulse ledgers. No arbitrary embedded-wall stress, general evolving-free-surface traction, spatially varying/nonlinear viscosity or viscosity/mesh/interface composition in the general carrier. Passing an isolated slab does not remove this gap. |
| Stickiness / adhesion | Finite beta is mechanical slip friction. It has units Pa s/m and is not an adhesive material law. No implemented contact-angle/adhesion evolution is exposed by the slab. | Chapter23 defines wall/interface energies, Young angle and adhesion work; the cap lab selects analytic equilibrium records at fixed volume. | A declared wall energy/contact-angle reconstruction, balanced capillarity and moving contact-line regularization are needed. An attractive force or larger beta cannot substitute for these. |
| Density | LiquidTransportConfig uses a constant carrier density and represented density, with equality required in surface modes. Box/slab inertia and nu=mu/rho have explicit units. | Chapter21 includes hydrostatic layers and weighted pressure identities; the forced lab checks density scaling within a single constant material. | No general advected variable-density/two-phase inertia and matching variable pressure/stress coefficients, conservative interphase momentum or density-ratio liquid qualification. |
The liquid composition and fixed-box viscosity milestones already exist; they are not the remaining requirement or a new completion claim. Volume/pressure classifications, interval geometry and output publication are real bounded components. General three-dimensional fluid topology and coupled physical boundaries remain separate prerequisites. The unresolved research qualifications and failed numerical evidence are retained; this book integration executes no new E2 roster or reference integrations.
Geometry foundation and bounded operator follow-on
The bounded geometry owner retains an admitted closed box surface and supplies physical cell volumes, one shared opening per MAC face, component labels and a conservative flux helper. It refuses unsupported meshes and disconnected fluid pieces within one cell. The operator follow-on borrows this owner for sealed pressure and a separately bounded reduced shear mode. It does not add those operators to the existing carrier or compose their different boundary models. World-space vertices express translation and axis-preserving scaling; no moving-body work is implied.
Nine recorded native controls and independent rational checks cover wet/dry and partial cuts, three separator orientations, translations/scaling, container contact, collision source agreement and shared flux incidence. Three thin baffles are recorded topology refusals. The non-dyadic fixture has a declared binary64 assembly allowance; exact-real Lean identities do not prove implementation refinement. Managed buffer capacities and constructor queue lifetimes are reported with an explicit scope, rather than a whole-program cap claim.
The pressure follow-on checks component compatibility/gauges, full integrated and cellwise residuals, actual corrected fields and a face-area kinetic ledger. The reduced shear follow-on checks owner-derived inertia/traction and a forced momentum/work ledger. The next gate is general embedded symmetric-strain interaction geometry and a composed step under consistent boundaries. Material variation and balanced capillarity/contact-angle reconstruction come after their coefficient/interface contracts exist. Moving/two-way solids add swept geometry and reaction/torque obligations. This sequence connects the remaining mesh, force, viscosity, density and adhesion requirements without claiming capstone completeness.
Current reviewed runtime composition
The historical links above remain evidence for their recorded checkpoint. Runtime PR65 and the book integration PR67 now provide a reviewable combined source. Qualification/publication of that combined source is tracked separately; a capability being present is not a physical-validation claim.
| Capability in the combined source | Exact scope and remaining obligation |
|---|---|
| Native OBJ/glTF/GLB import and CollisionMesh | Finite bounded static geometry, convex watertight solid or explicitly two-sided surface barrier. Native filesystem import is excluded from the WASM browser host. Particle collision is available; general fluid cut-cell topology and swept moving-volume assembly are still absent. |
| Bounded obstacle and plate laboratories | Owner-derived sealed pressure, separate reduced shear, and finite-mass moving-plate coupling remain their explicitly reduced models. They do not compose arbitrary imported CAD into a general 3D FSI solver. |
| MaterialSession/SimulationSession/PressureSession | Version-two bounded APIs expose units, immutable accepted snapshots, cooperative cancellation and explicit capability flags. Constructor pressure is unavailable; the particle material model does not call the grid pressure provider. |
| Qualified pressure provider in a full step | Opt-in sealed full-box carrier path retains pressure and final-step acceptance before publication. Boundary flux, surface modes and viscosity composition are refused on this path. Provider/model/schema/resource provenance and real inference qualification are distinct from physical accuracy and speed. |
| Quantitative desktop views and VTK | Particle diagnostics, bounded obstacle/transport slices/probes and immutable snapshot exports identify their evaluation/publication stages. A pressure evaluated before integration is not a pressure recomputed at later accepted particle positions. |
| Vector/moving emitter controls | Jet velocity, source translation and seeded perturbations are independent prescribed controls. The perturbations are not a solved or validated SST turbulence model. |
Consequently, static collision import is no longer the next missing implementation in this composition. The remaining geometry gate is conservative fluid topology, shared face flux/stress assembly and interval geometry under consistent physical boundaries. These requirements differ from collision queries and inspection. The engineering studies supply bounded worked examples and scoped proofs without presenting that missing backend as done.