w1n5t0n 129e28b207 Add codegen tool: schemas -> C++ headers + TS modules

bun-runnable TypeScript script (codegen/generate.ts) that:
- Validates each schemas/modes/*.json against the meta-schema via
  ajv (Draft 2020-12).
- Cross-checks params.length == ml.output_size and
  ml.input_channels.length == ml.input_size.
- Emits constexpr C++ data into nisps/modes/generated/ (one
  schema_types.hpp + one <mode_id>_schema.hpp per mode). Uses
  std::string_view + std::array; no std::vector, no heap, .f
  suffixed float literals (perf contract §3.3).
- Emits TS modules into playground/src/modes/generated/ (one
  types.ts + one <mode_id>_schema.ts + index.ts barrel).
- Idempotent: re-running yields byte-identical output.
- Exits non-zero on schema validation failure.

Reference templates live in codegen/templates/ (not consumed at
codegen time -- for human reviewers).

Golden test (codegen/tests/golden_test.ts) snapshots
paf_synth_schema.{hpp,ts} and verifies regeneration matches the
golden + that a second run is idempotent.

Until stream 1 lands nisps/core/math.hpp, schema_types.hpp ships
its own minimal Curve enum with a TODO marker pointing at the
eventual include.

2026-04-29 15:29:08 +03:00

2.8 KiB

Raw Blame History

MEMLNaut Mode-Schema Codegen

Bun + TypeScript tool that turns schemas/modes/*.json into:

C++ headers under nisps/modes/generated/<mode_id>_schema.hpp (constexpr data, no runtime cost).
TypeScript modules under playground/src/modes/generated/<mode_id>_schema.ts (typed ModeSchema objects).

The schemas are validated against schemas/schema.json (JSON Schema Draft 2020-12) on every run. Codegen exits non-zero if any schema fails validation.

Run

cd codegen
bun install        # one-shot, fetches ajv + types
bun run generate.ts   # or: bun run generate

The script writes everything into the two output dirs in one shot. Re-running with no schema changes is a no-op (byte-identical output).

Tests

tests/golden_test.ts regenerates from the live schemas into a temp dir and diffs against tests/golden/. The golden directory contains a snapshot of paf_synth_schema.{hpp,ts}. To refresh after intentional codegen changes:

bun run generate.ts
cp ../nisps/modes/generated/paf_synth_schema.hpp tests/golden/
cp ../playground/src/modes/generated/paf_synth_schema.ts tests/golden/

Run the test:

bun run test

Adding a new mode

Drop a new <mode_id>.json into schemas/modes/ (must validate against schemas/schema.json).
Run bun run generate.ts.
Commit the JSON + the regenerated C++/TS pair.

Layout

codegen/
├── package.json         # ajv (+ formats) + bun-types
├── tsconfig.json
├── generate.ts          # ~400 lines, single entrypoint
├── README.md            # this file
├── templates/           # reference templates (NOT consumed — for review)
│   ├── cpp_schema.hpp.template
│   └── ts_schema.ts.template
└── tests/
    ├── golden_test.ts   # diffs latest output against tests/golden/
    └── golden/
        ├── paf_synth_schema.hpp
        └── paf_synth_schema.ts

Conventions

C++ namespace: nisps::modes::generated. All generated symbols are inline constexpr so #include-ing the header in multiple TUs is safe.
C++ types: std::array, std::string_view, std::size_t. No std::vector, no heap.
Curve enum: declared in nisps/modes/generated/schema_types.hpp as a temporary local copy. Once stream 1 lands nisps/core/math.hpp, replace the local enum with an #include (search for TODO(stream-1) in the generated header).
Float literals: emitted with explicit .f suffix and decimal point, per the perf contract (architecture §3.3).
Order: schemas are processed in alphabetical order of mode_id so output is stable.
Naming: mode_id is snake_case in JSON; the generated C++ const prefix is k + PascalCase (e.g. kPafSynthParams); TS const is PascalCase + Schema (e.g. PafSynthSchema).

2.8 KiB Raw Blame History