[petlenz]

Closes #125. Part of the uvwmaterial port (#94); stacked on #138.

• docs/uvwmaterial-mapping.md: every class of uvwmaterial with its numsim-materials counterpart, the renamed parameters (E, ν → K, G; aspec_ratio → aspect_ratio; vrh → hill; scs → self_consistent; the 9-component orientation tensor → 6 components), the status of each port with its PR or issue, and the nested card next to the flat graph document it becomes.
• tools/convert_uvwmaterial_card.py: converts a uvwmaterial card into a json_model document for the supported types — linear elasticity, incremental elasticity, rule of mixtures, mean-field composites with sphere / long-fibre / ellipsoid Eshelby tensors, short-fibre composites, and isotropic damage with its state function, yield function and propagation laws (damage_exponential2 becomes exponential_damage_law with β = 1/(ε_f − ε₀)). Composites come out in the producer/consumer form of rule_of_mixture: Voigt, Reuss and Hill effective stiffness #130–short_fibre_composite: Mori-Tanaka over the fibre orientation distribution #133: a linear phase material becomes constant K, G and an isotropic_tangent; every phase is a phase node (produced tangent, fraction, Eshelby tensor, the matrix phase with the fraction the inclusions leave); the composite is the tangent producer <name>_tangent and <name> its linear_stress; the point names the tangent producer. A self-consistent card wires the matrix shape and the effective-medium ν over a local edge and puts residual on the point. A rule of mixtures with a nonlinear phase (damage) is Voigt through mean_field_stress — Reuss and Hill with nonlinear phases need the local strains of Nonlinear rule of mixtures with per-phase history #117 and raise with that number, as a nonlinear matrix of a mean-field card raises with Nonlinear mean field by linear comparison: plasticity and damage in the phases #116. Types that are not ported yet raise with the issue number that covers them; the matrix Poisson ratio repeated in the old Eshelby blocks is dropped, since the graph reads it from the matrix tangent.
• Tests: tools/tests (8, ctest convert_uvwmaterial_card): the Mori–Tanaka, self-consistent, short-fibre, linear and damage-in-Voigt rule-of-mixture cards taken from uvwmaterial's examples, the propagation laws and state functions, the unported types; test_uvwmaterial_cards loads the converted Mori–Tanaka card through json_model and gets the analytic Mori–Tanaka bulk modulus from mt_tangent.

938/938 tests pass (GCC 14, reports on).

[petlenz]

Review findings (design, tmech expressions instead of component loops, modern C++, comments, tests, correctness). Severity: must = wrong result, hang, build break or unreachable feature; should = design, duplication, accuracy; nit = style.

mapping doc + converter — needs changes

• [must] convert_uvwmaterial_card.py:171 emits top-level point; json_model rejects it ("unrecognised top-level key"). The gtest pastes a hand-edited literal, never the converter's output. Emit {"model": {...}, "point": {...}} (the report layout) and let the gtest load converter output.
• [must] linear → linear_damage_law with K, but numsim's law takes kappa_0, kappa_f; uvw linear is D = K·κ → power_damage_law with m = 1 (mapping doc line 22 wrong too).
• [must] vector_strain_state_function emits m, n vectors; numsim wants m_direction, n_direction axis indices.
• [must] strain_energy_state_function needs tangent_source and youngs_modulus; converter emits only strain_source.
• [should] 4 of ~12 card kinds tested; one gtest loading every fixture through json_model would have caught all four musts. No card path in errors; if/elif chains next to the tables; handlers untyped.
• [nit] converter test under the numpy guard; literal 6000.000000000001; doc row 28 wrong name; Peerlings d_max 0.999 vs uvw 0.99.

[petlenz]

Review findings addressed in the last commit, together with the Eshelby-node refactor:

• must, output rejected by json_model: the converter writes the report layout {"model": {"materials": […]}, "point": {…}}; the C++ cards test now holds the converter's actual model output for the MORI_TANAKA card of the Python tests instead of a hand-edited literal.
• must, linear → linear_damage_law with K: it is power_damage_law with m = 1 (uvwmaterial's D = K·κ); the mapping row says so.
• must, vector_strain_state_function m/n vectors: converted to the axis indices m_direction/n_direction; a non-axis vector raises.
• must, strain_energy_state_function: tangent_source (the elastic material) and youngs_modulus (the card's E, or 9KG/(3K+G)).
• Eshelby nodes: one eshelby_sphere/_cylinder/_spheroid per uvwmaterial eshelby_tensor, named <prefix>_<key>_shape and wired to the composite's matrix; inclusion and short_fibre_composite name it. Mapping document rows and example regenerated.
• Tests: the Python suite covers the node wiring for Mori–Tanaka and short fibre, the linear law and both state functions' parameters, the axis-vector rejection.

Still open: a gtest that loads every Python fixture through json_model, card paths in error messages, the if/elif tables, typed handlers.

[petlenz]

Producer/consumer pass: the converter emits a constant_scalar per volume fraction and inclusion phases for the rule of mixtures, the Mori–Tanaka inclusions and the short-fibre phase; Eshelby nodes consume isotropic_moduli of the composite's matrix (<matrix>_moduli, made once). Mapping rows and the example are regenerated; the C++ cards test holds the converter's actual output.