MoziThermoCalcDB-NASA


Performance-oriented notes for working with the NASA polynomial toolchain in src/.

📥 Data ingestion

- Use the standalone DataLoader/loadModelSource script in scripts/ to parse CSVs once into an in-memory ModelSource; invalid or incomplete rows are skipped early via transformRow validation.
- Each row is indexed under multiple ids (Name-State, Formula-State, Name-Formula) so lookups avoid string recomputation in hot paths.
- CSV parsing is synchronous and per-file; feed an array of { path, range } objects to batch all ranges in a single pass.

🔎 Lookup pipeline

- Source resolves a component's equation set by id and validates required coefficients before returning data.
- Range selection prefers the requested NASA window but falls back to adjacent ranges (buildRangePreference) to keep calculations moving when partial data exists.
- validateRangeData filters out records with missing or non-finite coefficients to prevent downstream math errors.

♻️ Coefficient access and reuse

- HSG lazily extracts and caches NASA7/NASA9 coefficients on construction, avoiding repeated Source lookups per call.
- Required coefficients are asserted once (requireCoeffs); failures return null instead of throwing inside the compute path to keep execution cheap.
- Molecular weight is cached (props) so mass-basis conversions reuse the same value instead of re-reading coefficient blobs.

🚀 Batch computation path

- HSGs builds and caches per-component HSG instances up front and reuses them for all property calculations.
- A single NASA range decision (selectNasaType) is shared across the batch in calc_components_hsg, reducing branching inside per-component loops.
- Reaction helpers (RXNAdapter) consume these batch results directly, so reaction properties reuse the already-computed component thermodynamics.

🧮 Thermo kernels

- Polynomial evaluators in src/thermo are pure functions: they convert to Kelvin once, operate on numbers, and return CustomProp structs with units intact.
- Mass-basis conversions (toMassBasis) are applied only when explicitly requested via basis: 'mass'; default is molar to avoid extra work.

⚙️ Usage tips for better throughput

- Load all CSVs once into a modelSource, then reuse a single Source/HSG or HSGs instance for repeated queries.
- Prefer the batch API when possible:
- Component: H_T, S_T, G_T, Cp_T for single values.
- Reaction: dH_rxn_STD, dS_rxn_STD, dG_rxn_STD, Keq, Keq_vh_shortcut for multi-component calculations in one pass.
- Stick to 'nasa9' unless you need NASA7 data; NASA9 has fuller validation (b1/b2) and better range coverage in the fallback order.

🧪 Examples

- Run examples/appUsage.ts end-to-end with npx ts-node --esm --experimental-specifier-resolution=node examples/appUsage.ts to load the bundled NASA9 CSVs and print species/reaction properties.
- If you adjust CSV paths or ranges, tweak the loadExampleModelSource call inside examples/appUsage.ts to match your files.

📘 Minimal snippet

``ts
import { loadModelSource } from './scripts/DataLoader';
import { H_T, dG_rxn_STD } from './src/app';

const modelSource = await loadModelSource([
{ path: 'data/nasa9_range1.csv', range: 'nasa9_200_1000_K' },
{ path: 'data/nasa9_range2.csv', range: 'nasa9_1000_6000_K' }
]);

const water = { name: 'Water', formula: 'H2O', state: 'g' } as const;
const H = H_T({ component: water, temperature: { value: 900, unit: 'K' }, model_source: modelSource });

const reaction = {
name: 'water-formation-gas',
reaction: '2 H2(g) + O2(g) => 2 H2O(g)',
components: [
{ name: 'Hydrogen', formula: 'H2', state: 'g' },
{ name: 'Oxygen', formula: 'O2', state: 'g' },
water
]
} as const;

const dG = dG_rxn_STD({ reaction, temperature: { value: 900, unit: 'K' }, model_source: modelSource });
``