Delete the legacy _convertUnitValue helper on the domain and the
duplicate convertUnitValue export on curveNormalizer; both were
identical to UnitPolicy.convert. Callers in flowController, the
curve normalizer, and buildQHCurve now go through this.unitPolicy.
The contract in .claude/refactor/CONTRACTS.md §6 named these as the
target migration; this finishes the rollout for rotatingMachine.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The pre-existing efficiency formula `η = flow/power` produced tiny SI-unit
values (m³/J ≈ 1e-5), was monotonic in ctrl for centrifugal-pump curves
(no interior peak), and made NCog collapse to 0 — which cascaded into MGC
reporting BEP-position 0.0% always. Replaced with hydraulic efficiency
η = (Q·ΔP)/P_shaft, the dimensionless 0..1 ratio that has a real BEP and
matches the form MGC's group-level math uses.
- prediction/efficiencyMath.js:
* calcEfficiencyCurve takes pressureDiffPa; η = 0 when dP missing
* calcCog guards (yMax > yMin) before computing NCog (was unguarded /0)
* calcEfficiency falls back to predictFlow.currentF when measured ΔP is
missing, so predicted-variant calls still produce a meaningful η before
the differential measurement settles
- specificClass.js:
* Asset-registry lookup renamed: 'machine' → 'rotatingmachine' (matches
the datasets/assetData/ rename in generalFunctions). The error path
quotes the new filename so operators can find it.
* Two-call-site fix: with default-param stateConfig={}, the single-arg
constructor path (BaseNodeAdapter calls `new Machine(this.config)`
after pre-setting Machine._pendingExtras) was silently clobbering the
pre-set extras. Only overwrite when the caller explicitly passes them.
* Push port 0 deltas (notifyOutputChanged) after prediction updates so
dashboards see state + predicted-flow changes as they happen.
- pressure/pressureRouter.js: routing + fallback hardening (the trigger
for the bep-distance-cascade reproduction).
- display/workingCurves.js: Q-H curve generator extended.
- New tests:
* test/integration/qh-curve.integration.test.js — Q-H curve shape
* test/integration/bep-distance-cascade.integration.test.js — reproduces
the dashboard report (absDistFromPeak=0, NCog=0, efficiency=0 after a
setpoint move) at the unit level so future regressions fail loudly.
Full suite: 214/214 pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
specificClass._setupCurves now calls assetResolver.resolveAssetMetadata
to derive supplier/type/units from the model id, instead of trusting
denormalized fields on the node config. If the model isn't in the
registry, installs a null-predictor stub and logs a clear "pick a model
from the asset menu" error rather than crashing.
rotatingMachine.html: defaults block trimmed (supplier/category/assetType
were stale copies of registry data).
Tests:
- New test/basic/assetMetadata.basic.test.js covers the registry-resolve
path and the missing-model fallback.
- nodeClass-config / error-paths / nodeClass-routing / factories /
abort-deadlock fixtures updated to the trimmed asset shape.
- 209/209 tests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
For all 11 nodes with auto-gen markers: wiki/Home.md sections 5 (topic
contract) and 9 (data model) regenerated via npm run wiki:all. New
Unit column shows '<measure> (default <unit>)' for declared topics,
'—' otherwise. Effect column now uses descriptor.description (P11.2
field) overriding the generic per-prefix fallback.
For rotatingMachine + reactor: Phase 10 test rewrites — 3 + 8 files
moved off private nodeClass internals (_attachInputHandler, _commands,
_pendingExtras, _registerChild, _tick, etc.) to the public
BaseNodeAdapter surface (node.handlers.input, node.source.*).
+6 / +7 net new tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
