openapi-schema-pydantic

OpenAPI (v3) specification schema as Pydantic classes.

The naming of the classes follows the schema in OpenAPI specification.

Installation

pip install openapi-schema-pydantic

Try me

from openapi_schema_pydantic import OpenAPI, Info, PathItem, Operation, Response

# Construct OpenAPI by pydantic objects
open_api = OpenAPI(
    info=Info(
        title="My own API",
        version="v0.0.1",
    ),
    paths={
        "/ping": PathItem(
            get=Operation(
                responses={
                    "200": Response(
                        description="pong"
                    )
                }
            )
        )
    },
)
print(open_api.json(by_alias=True, exclude_none=True, indent=2))

Result:

{
  "openapi": "3.1.0",
  "info": {
    "title": "My own API",
    "version": "v0.0.1"
  },
  "servers": [
    {
      "url": "/"
    }
  ],
  "paths": {
    "/ping": {
      "get": {
        "responses": {
          "200": {
            "description": "pong"
          }
        },
        "deprecated": false
      }
    }
  }
}

Take advantage of Pydantic

Pydantic is a great tool, allow you to use object / dict / mixed data for for input.

The following examples give the same OpenAPI result as above:

from openapi_schema_pydantic import OpenAPI, PathItem, Response

# Construct OpenAPI from dict
open_api = OpenAPI.parse_obj({
    "info": {"title": "My own API", "version": "v0.0.1"},
    "paths": {
        "/ping": {
            "get": {"responses": {"200": {"description": "pong"}}}
        }
    },
})

# Construct OpenAPI with mix of dict/object
open_api = OpenAPI.parse_obj({
    "info": {"title": "My own API", "version": "v0.0.1"},
    "paths": {
        "/ping": PathItem(
            get={"responses": {"200": Response(description="pong")}}
        )
    },
})

Use Pydantic classes as schema

• The Schema Object in OpenAPI has definitions and tweaks in JSON Schema, which is hard to comprehend and define a good data class
• Pydantic already has a good way to create JSON schema, let's not re-invent the wheel

The approach to deal with this:

1. Use PydanticSchema objects to represent the Schema in OpenAPI object
2. Invoke construct_open_api_with_schema_class to resolve the JSON schemas and references

from pydantic import BaseModel, Field

from openapi_schema_pydantic import OpenAPI
from openapi_schema_pydantic.util import PydanticSchema, construct_open_api_with_schema_class

def construct_base_open_api() -> OpenAPI:
    return OpenAPI.parse_obj({
        "info": {"title": "My own API", "version": "v0.0.1"},
        "paths": {
            "/ping": {
                "post": {
                    "requestBody": {"content": {"application/json": {
                        "schema": PydanticSchema(schema_class=PingRequest)
                    }}},
                    "responses": {"200": {
                        "description": "pong",
                        "content": {"application/json": {
                            "schema": PydanticSchema(schema_class=PingResponse)
                        }},
                    }},
                }
            }
        },
    })

class PingRequest(BaseModel):
    """Ping Request"""
    req_foo: str = Field(description="foo value of the request")
    req_bar: str = Field(description="bar value of the request")

class PingResponse(BaseModel):
    """Ping response"""
    resp_foo: str = Field(description="foo value of the response")
    resp_bar: str = Field(description="bar value of the response")

open_api = construct_base_open_api()
open_api = construct_open_api_with_schema_class(open_api)

# print the result openapi.json
print(open_api.json(by_alias=True, exclude_none=True, indent=2))

Result:

{
  "openapi": "3.1.0",
  "info": {
    "title": "My own API",
    "version": "v0.0.1"
  },
  "servers": [
    {
      "url": "/"
    }
  ],
  "paths": {
    "/ping": {
      "post": {
        "requestBody": {
          "content": {
            "application/json": {
              "schema": {
                "$ref": "#/components/schemas/PingRequest"
              }
            }
          },
          "required": false
        },
        "responses": {
          "200": {
            "description": "pong",
            "content": {
              "application/json": {
                "schema": {
                  "$ref": "#/components/schemas/PingResponse"
                }
              }
            }
          }
        },
        "deprecated": false
      }
    }
  },
  "components": {
    "schemas": {
      "PingRequest": {
        "title": "PingRequest",
        "required": [
          "req_foo",
          "req_bar"
        ],
        "type": "object",
        "properties": {
          "req_foo": {
            "title": "Req Foo",
            "type": "string",
            "description": "foo value of the request"
          },
          "req_bar": {
            "title": "Req Bar",
            "type": "string",
            "description": "bar value of the request"
          }
        },
        "description": "Ping Request"
      },
      "PingResponse": {
        "title": "PingResponse",
        "required": [
          "resp_foo",
          "resp_bar"
        ],
        "type": "object",
        "properties": {
          "resp_foo": {
            "title": "Resp Foo",
            "type": "string",
            "description": "foo value of the response"
          },
          "resp_bar": {
            "title": "Resp Bar",
            "type": "string",
            "description": "bar value of the response"
          }
        },
        "description": "Ping response"
      }
    }
  }
}

Notes

Use of OpenAPI.json() / OpenAPI.dict()

When using OpenAPI.json() / OpenAPI.dict() function, arguments by_alias=True, exclude_none=True has to be in place. Otherwise the result json will not fit the OpenAPI standard.

# OK
open_api.json(by_alias=True, exclude_none=True, indent=2)

# Not good
open_api.json(indent=2)

More info about field alias:

Non-pydantic schema types

Some schema types are not implemented as pydantic classes. Please refer to the following for more info:

Use OpenAPI 3.0.3 instead of 3.1.0

Some UI renderings (e.g. Swagger) still do not support OpenAPI 3.1.0. It is allowed to use the old 3.0.3 version by importing from different paths:

from openapi_schema_pydantic.v3.v3_0_3 import OpenAPI, ...
from openapi_schema_pydantic.v3.v3_0_3.util import PydanticSchema, construct_open_api_with_schema_class

License

MIT License