python-lapidary / lapidary-render Goto Github PK

https://github.com/python-jsonschema/referencing

Name of any schema, parameter or property can ba any valid string, excluding Unicode control characters, while python accepts only alphanumerics + underscore (_), which has a special meaning if placed at the beginning.

Possible ways of handling names that are not valid python identifiers:

• fail
• use x-lapidary-names or x-lapidary-name
• employ default name mangling
• in case of object schema properties - skip from model definition, but support as an object variable + validate

Composite schemas are those using oneOf, anyOf, allOf and not.

• each schema will be represented as a single class (no inheritance)
• only schemas that are either defined inline or referenced under the paths, responses or x-apiResponses elements will be source for model classes
• schemas will be processed recursively
• all properties from all schemas will be present, except from not schema, which properties should be excluded
• each fields attributes (type, required, nullable) will reflect all the schemas, using the least restrictive one.
• processing contradictory schemas will fail

module:class

It would be great if I could offload rendering to an external project like cookiecutter or copier. lapidary-render could focus on developing template and generating model.

Unfortunately theses projects are designed to generate project stubs, with single file being generated from each template.

https://redocly.com/docs/api-reference-docs/spec-extensions/

add .lapidary_cache and standard python dirs

use

• title
• description
• example
• default

Params and attributes change order even though schema doesn't change.
Makes a mess in git commit.

Poetry has an undocumented feature allowing to run a custom script while building the project.

Use this feature to generate sources on build, instead committing the code to the VCS.

readOnly and writeOnly properties really mean response- and request - only. it's really a direction and not access.

1. Add field Annotation Only(direction:Literal['read',''write], required: bool)
2. add custom parser to base model class to support that annotation
3. pass direction in the model.dump_python call
4. render all directed properties as non-required, with added Only annotation

class MyModel(BaseModel)
    directed_field: Annotated[MyType|None, pydantic.Field(...), Only('read', required=True)] = None