Schema Generation

This page covers the generation step of the workflow shown on the Tools page.

The schemas are written in LinkML because a single definition gives several outputs: the JSON Schemas in schemas/, the Pydantic models in mdstools/models/, and the schema reference pages of this documentation. The JSON Schemas and the Pydantic models are generated separately from the LinkML files and are never edited by hand. A schema change usually starts with an edited example in examples/, which shows how the metadata should look and is validated once the schema is updated.

Data packages and Frictionless

The package schemas echemdb_package and svgdigitizer_package build on the Frictionless Data Package standard. Each entry in resources must be valid against both the Frictionless data resource schema (fields such as name, path and format) and the echemdb resource definition, which adds the metadata property. The generated JSON Schemas express this with allOf and refer to the Frictionless schema by its URL (https://datapackage.org/profiles/2.0/dataresource.json), so any JSON Schema validator can resolve it. Within this repository, pixi run validate and the schema enrichment use a copy of the Frictionless schemas that is downloaded on first use into schemas/frictionless/ (not part of the repository).

Commands

Schemas are generated from LinkML YAML definitions:

# Generate JSON Schemas
pixi run generate-schemas

# Generate Pydantic models
pixi run generate-models

# Generate both
pixi run generate-all

Validate the example files against the schemas:

pixi run validate