Enum components are being generated as nullable when the property holding it is nullable

[renke0]

Description

I have the following scenario:

const Brand = z.enum(['Ford', 'Toyota', 'Tesla']).openapi('Brand');

const Car = z
    .object({
        brand: Brand.nullable(),
        model: z.string(),
    })
    .openapi('Car');

Expected result

I expected the property Car.brand to be a nullable Brand

components:
  schemas:
    Brand:
      type: string
      enum:
        - Ford
        - Toyota
        - Tesla
    Car:
      type: object
      properties:
        brand:
          oneOf:
            - type: 'null'
            - $ref: '#/components/schemas/Brand'
        model:
          type: string

Actual result

The spec generated declares Brand as nullable, instead of making the property Car.brand nullable

components:
  schemas:
    Brand:
      type:
        - string
        - 'null'
      enum:
        - Ford
        - Toyota
        - Tesla
    Car:
      type: object
      properties:
        brand:
          $ref: '#/components/schemas/Brand'
        model:
          type: string

More info

I understand the rationale behind the way this was generated. The problem is that Brand could be reusable, and in some places it might not be nullable. Is there a workaround this?

[AGalabov]

@renke0 I can see the problem. If you pass Brand before Car in the OpenApiGeneratorV3 constructor's array of definitions everything works fine. But if you switch their order or simply do not pass Brand at all then what you've described happens.

I am not sure how easy it would be for us to fix this and when we will have the time for it. In the mean time - you can make sure to pass it in order as a workaround.

[flipvh]

Having the same issue - registered models / $refs get an additional property null if I use them with an appended .nullable somewhere else. And due to a larger API, so far I have no luck changing the order to completely prevent it. @AGalabov would you be interested - and have time - to work on this as a sponsored improvement or are you open to a PR and if so, perhaps know someone who can work on this? thnks for this amazing lib!