ANYmaterial 0.1.0__tar.gz

anymaterial-0.1.0/CHANGELOG.md

@@ -0,0 +1,191 @@
+# Changelog
+
+## Unreleased
+
+No changes yet.
+
+## 0.1.0 - 2026-08-20
+
+First public feature release. The material code is extracted from ANYsolver and
+ANYfem; see [MIGRATION.md](MIGRATION.md) for provenance.
+
+Added:
+
+- **Embeddable material selection.** `MaterialEditor(on_apply=...)` adds a
+  **Use material** button, and `open_material_editor` opens the same validated
+  editor inside an application's existing Tk event loop.
+- **A material library.** `library()` returns 33 named materials, including the
+  17 DNV-RP-C208 grade and thickness-class rows, curated cited entries, and
+  measured campaign means. `MaterialLibrary` supports `get`, `find` (by category, status,
+  text or whether it has a curve), `add`, `remove` and JSON round-tripping;
+  `LibraryEntry` carries the category, the standard, the source and free notes.
+- **Provenance as a first-class field.** Every entry has a `status`:
+  `tabulated` for a standard's own table, `indicative` for a typical published
+  figure that is **not a design value**, `user` for anything added at runtime.
+  `LibraryEntry.is_design_value` exposes the distinction, and the CLI and the
+  editor both surface it.
+- **User-extensible.** `add_to_user_library` writes to
+  `~/.anymaterial/materials.json`, or wherever `ANYMATERIAL_LIBRARY` points. A
+  user entry with the same name as a shipped one takes precedence, which is how a
+  grade gets corrected locally without editing the package. Anything added is
+  recorded as `status="user"` even if it claims otherwise, so adding cannot
+  launder provenance.
+- **Curve plots as SVG.** `anymaterial.plot` renders flow curves without a
+  plotting dependency — hand-written SVG, so the output goes into a report or a
+  commit as text. `sample_curve`, `curve_svg` and `write_curve_svg`.
+- **CLI** gained `library` (list and filter), `plot` (write an SVG) and `add`
+  (put a material in the user library).
+- **Editor** gained a library pane: browse and filter on the left, load an entry
+  into the form, see the selected entry's provenance, and plot the edited
+  material against the library selection. Buttons add the current material to the
+  library and export the plotted curves to SVG.
+
+**The shipped library is 33 materials.** It grew from 19 with a supplied
+candidate set, then shrank again when the citations were checked: seven entries
+were removed because their cited source does not support their numbers.
+
+**One composite is back, properly sourced.** `IM7/8552 UD lamina, RTD (measured
+mean)` uses room-temperature-dry campaign means from the
+[NCAMP/NIAR Hexcel 8552 IM7 unidirectional qualification programme](https://www.wichita.edu/industry_and_defense/NIAR/Research/hexcel8552.php),
+generated under FAA oversight — E1 23.51 Msi, E2 1.30 Msi, G12 0.68 Msi,
+nu12 0.316, read from data report CAM-RP-2009-015 Rev A sections 2.3.1/2.3.2/2.3.5
+with the environmental column resolved as RTD (70 °F, dry). The remaining five
+constants are **not measured**: E3, nu13 and G13 follow from assumed transverse
+isotropy, nu23 is assumed 0.45 outright, and G23 follows exactly from those. The
+`calculation` record says which is which. Internal check: nu21 by reciprocity is
+0.0175 against the programme's separately measured transverse-compression nu21 of
+0.024. It carries no yield stress and defines no failure criterion.
+
+Added `run_gui.py` at the repository root — an entry point for an IDE's Run
+button that puts `src` on `sys.path` first, so the editor opens from a bare
+checkout with nothing installed, and what runs is the working tree rather than an
+installed copy. There is a test that executes it.
+
+Added a new status, **`measured`**, and six entries under it from the
+[Zenodo coupon database](https://doi.org/10.5281/zenodo.6965147) (Hartloper,
+Ozden, de Castro e Sousa & Lignos 2022, v1.0.0, **CC-BY-4.0**): S235/275 JR+AR
+15 mm plate, S355 J2+N 15 mm plate, S355 J2+M IPE270 flange, S460 NL 25 mm plate,
+S690 QL 12 mm plate and A992 Gr.50 W14X82 flange. Each is the mean of a named
+campaign and carries its coupon count, the coefficient of variation on both yield
+and modulus, the product form, the campaign citekey, the licence and the full
+citation.
+
+A `measured` entry is **not** a design value and says so loudly: the measured
+mean yield of S355 is 412 MPa against 357 MPa tabulated, so using a mean as a
+design strength is unconservative. `LibraryEntry` refuses `status="measured"`
+without a measurement record — a measured value without its sample size and
+scatter is a number claiming an authority it cannot show.
+
+Removed, each after checking the cited source:
+
+| Entry | Why |
+| --- | --- |
+| T300/5208 UD lamina | cited NASA report is *Space Radiation Effects on Graphite-Epoxy* and covers T300/**934** |
+| E-glass/epoxy UD lamina | cited NASA report is *Thermal expansion properties of composite materials* |
+| IM7/8552 UD lamina | the entry's own notes admitted the citation does not source its elastic constants |
+| Norway spruce | Poisson ratios 0.37/0.47/0.44 are the Wood Handbook's **Sitka** spruce row, and the moduli match neither species |
+| Titanium Grade 2 | cited TIMET datasheet URL returns HTTP 404 |
+| Ti-6Al-4V Grade 5 | cited TIMET datasheet URL returns HTTP 404 |
+| EN AW-6082-T6 extrusion | cited hydro.com URL returns HTTP 404, and a verified 6082-T6 entry already ships |
+
+Re-sourced rather than removed:
+
+- **EN AW-5083-H116** — dead hydro.com link replaced; the 215 MPa proof is
+  confirmed as the EN 485-2 minimum for the temper.
+- **AA 6061-T6, AA 7075-T6** — no longer cite a licensed table as if it were the
+  source; the notes now say plainly that these are widely corroborated typical
+  values and that the authoritative minimums are elsewhere.
+- **The three nonlinear stainless/duplex curves** — the notes now separate what
+  the source supports (the Ramberg-Osgood form per EN 1993-1-4 Annex C, and the
+  fy/fu inputs) from what was computed here (the tabulated points).
+
+Everything that survived was validated before merging: it builds, its compliance
+is symmetric and positive definite, and every flow curve is finite, monotonic and
+has a non-negative hardening modulus.
+
+- `LibraryEntry` gained **`calculation`** (how a derived entry was derived — the
+  Ramberg-Osgood exponent behind a tabulated curve) and **`measurement`** (coupon
+  count, scatter, campaign, licence, citation). Without the first, the candidate
+  set's derivation records would have been silently dropped on import.
+- A test now fails the build if any shipped entry lacks a source, a standard or
+  notes.
+- `anymaterial library` prints only the warning matching the statuses actually
+  listed, so it stays worth reading, and takes `--status measured`. The choices
+  are derived from the library's own status list so the two cannot drift.
+- `anymaterial library` sizes its name column to the content, because a fixed
+  width truncated the longer alloy names into ambiguity.
+
+The first two curated materials were added from sources found online. Their
+numbers are attributed in `data/materials.json` and marked `indicative`:
+
+- **EN AW-6082-T6** — E 70 GPa, 0.2% proof 310 MPa, tensile 340 MPa, density
+  2700 kg/m³, from the thyssenkrupp datasheet. That datasheet does not separate
+  tempers, and EN 1999-1-1 characteristic values for design are lower and depend
+  on product form and thickness; the entry says so.
+- **EN 1.4404 (316L)** — E 200 GPa and density 8000 kg/m³ from the thyssenkrupp
+  datasheet, 0.2% proof 220 MPa and tensile 500 MPa as commonly quoted for the
+  grade under EN 10088-3.
+
+Neither carries a hardening curve, because no citable source for one was found.
+Nothing was invented to fill the gap: an aluminium entry (EN AW-5083-H111) was
+prepared and then dropped when both candidate sources returned HTTP 403 and the
+numbers could not be verified.
+
+### Initial package foundation
+
+Added:
+
+- **Contract** — `StructuralMaterial` protocol, `ENGINEERING_VOIGT_ORDER`,
+  symmetry resolution and `elastic_compliance_matrix`. Structural rather than
+  inheritance-based, so a material defined in another package satisfies it
+  without importing anything from here.
+- **Materials** — `IsotropicMaterial` and `OrthotropicMaterial`, both validated
+  on construction.
+- **Reductions** — `shell_material_matrices`, `beam_material_properties` and
+  `shell_characteristic_modulus`, all derived from the 6x6 compliance so
+  isotropic and orthotropic materials reach a formulation by the same path.
+- **Yield** — `Hill48Yield` plus free functions `hill48_strengths`,
+  `hill48_coefficients` and `hill48_equivalent_stress` that consume the
+  six-strength protocol by attribute lookup.
+- **Curves** — the `HardeningCurve` protocol and four implementations:
+  `LinearHardeningCurve`, `PiecewiseLinearCurve`, `PowerLawHardeningCurve` and
+  the ported `DNVC208MaterialCurve`. The first three are new.
+- **Library** — `steel()`, `available_grades()`, `thickness_classes()`,
+  `dnv_c208_steel_properties()` and `dnv_c208_steel_curve()`. The RP-C208 grade
+  table now lives in `data/dnv_rp_c208.json` instead of in code.
+- **Specifications** — `MaterialSpec`, which records *how to rebuild* a
+  hardening curve rather than a frozen copy, so a saved material does not reload
+  as elastic. `hardening_descriptor` raises for a curve it cannot describe rather
+  than serializing it away.
+- **Editor** — a tkinter form with live validation and a flow-curve plot on a
+  plain `Canvas`, entry point `anymaterial-gui`.
+- **CLI** — `anymaterial grades|properties|curve|show|validate`, each with
+  `--json`.
+
+Verified bit-exact against `anysolver` 0.1.3 at
+`8b4553cc680ff925df850e627165fc336615eaba` for the full RP-C208 table (17 rows),
+the flow stress and hardening modulus of every resulting curve, isotropic and
+orthotropic compliance, all three reductions, the Hill-48 coefficients,
+quadratic form, utilization and equivalent stress, and the exact wording of
+every validation message.
+
+Deliberate differences from the source, both to be confirmed when ANYsolver is
+stripped:
+
+- `IsotropicMaterial` validates on construction. `anysolver.fe_core.Material`
+  does not, so a caller that built a material with an inadmissible Poisson ratio
+  and never used it would now fail at construction instead of at first use.
+- Ramberg-Osgood is not implemented. It has no distinct yield point, so as a
+  flow curve its stress tends to zero as plastic strain does, which a return
+  mapping cannot use; the usual repairs produce a different curve while keeping
+  the name. `PowerLawHardeningCurve.from_yield` covers the same behaviour with an
+  explicit yield stress.
+
+Not included, and staying in ANYsolver: `FiberSectionPlasticityConfig` (a solver
+fiber grid, not a material) and all section properties.
+
+## 0.0.1
+
+- Repository scaffolding: packaging metadata, CI across Python 3.11-3.14 on
+  Windows and Linux, and the layering checks that keep the package a leaf of the
+  dependency graph.