⚠️ I'm willing to work on this change myself. But before starting, I'd like to discuss it.

Is your feature request related to a problem? Please describe.

Here's a repro: https://github.com/goodoldneon/datamodel-code-generator-Repro

OpenAPI allows you to specify properties as:

• Required
• Required with default
• Required nullable
• Unrequired

However, mashalling the pydantic models (e.g. BaseModel.dict()) will not differentiate them.

For example, let's say we have the following OpenAPI schema:

Foo:
  type: object
  properties:
    required:
      type: integer
    required_default:
      type: integer
      default: 2
    required_nullable:
      type: integer
      nullable: true
    unrequired:
      type: integer
  required:
    - required
    - required_default
    - required_nullable

If I marshall the generated model into JSON, I expect that all of the required* properties will always exist but the unrequired property may not.

If I instantiate Foo like this:

Foo(required=1, required_nullable=None)

Then I'd expect the marshalled JSON to be:

{
  "required": 1,
  "required_default": 2,
  "required_nullable": null,
}

But that doesn't seem to be possible.

Describe the solution you'd like

I thought I could do this with a new BaseModel.dict() argument, but the maintainers of pydantic rejected my PR:

https://github.com/samuelcolvin/pydantic/discussions/2837

They provided a (seemingly) good solution to the problem on their end. This makes me think the solution for datamodel-code-generator would be a new CLI arg that creates models like @PrettyWood's example.

Describe alternatives you've considered

I considered making a change to pydantic but my PR was rejected.