Instructions

The package needs Mojo 1.1.0 on linux-64. Pixi installs that compiler from the Modular channel and conda-forge.

Install the published package

pixi add --channel https://prefix.dev/leo-gan/leo-gan mojo-smile

The conda package name is mojo-smile. The Mojo import is smile. The generator binary is gld-smilegen-mojo.

Build this repository

git clone https://github.com/leo-gan/gld-smile.git
cd gld-smile
pixi install
pixi run test

If pixi install returns 401 from conda.modular.com, put PREFIX_API_KEY in a file named .env in the repository root and run scripts/ci-setup.sh. That file is gitignored. Do not commit it.

Generate structs

JSON Schema files live in testdata/schema/. The generator reads one file and writes one Mojo file per object or union.

pixi run mojo run -I src src/codegen/cli.mojo -- \
  --schema testdata/schema/point.json \
  --out tests/generated

pixi run generate regenerates every schema. pixi run check-generated regenerates into a temporary directory and diffs it against tests/generated.

A generated struct has encoded_len, encode_to, and decode_from. Integer fields are Int64. Number fields are Float64. binary is List[Byte]. biginteger and bigdecimal are decimal text, and the decimal text keeps scale (1.00 is not 1).

The generator accepts the same JSON Schema subset as the other gld JSON libraries: type, properties, required, items, $ref, $defs, definitions, enum, const, oneOf, and anyOf. Smile adds the type names binary, biginteger, and bigdecimal. A schema that refers to itself is rejected, because Mojo does not allow that struct shape.