Improve oneOf / discriminator support

[Redmega]

Description

Hi folks! Love the project, the client feels much nicer to use than the "official" openapi typescript-fetch generator.

I'm not sure if this is a feature request or a bug report, but I'm marking it as a feature request just in case it was left out intentionally. It feels that way since I get a warning when generating the client:

❗️ Transformers warning: schema {"items":[{"type":"string"},{"type":"boolean"},{"type":"number"}],"logicalOperator":"or"} is too complex and won't be currently processed. This will likely produce an incomplete transformer which is not what you want. Please open an issue if you'd like this improved https://github.com/hey-api/openapi-ts/issues

I have an openapi schema that looks like this:

        field_type:
          type: string
          enum:
            - string
            - number
            - boolean
        setting_name:
          type: string
        setting_value:
          oneOf:
            - type: string
            - type: number
            - type: boolean
          discriminator:
            propertyName: field_type
            mapping:
              string: string
              number: number
              boolean: boolean

I would expect the type to be something like this:

export type Setting = { setting_name: string } & (
  | { field_type: "string"; setting_value: string }
  | { field_type: "number"; setting_value: number }
  | { field_type: "boolean"; setting_value: boolean }
);

so that we can perform type narrowing on switch(setting.field_type) for example.

Instead I get this:

export type Setting = {
  field_type: "string" | "number" | "boolean";
  setting_name: string;
  setting_value: string | number | boolean;
};

which forces me to do a lot of explicit type checking of the setting_value in practice.

Is there any way we can get better support for oneOf's with discriminators and/or with primitive types?

The client code ends up looking the same if I leave out the discriminator, so it seems like it might not even be getting read/applied?

[mrlubos]

Hi @Redmega, the first sentence is great to hear! The rest not so much 😀 there's an existing issue for discriminators, is yours the same? #1561

I see you've also added a log from transformers, is your issue about multiple problems?

[Redmega]

I saw that but I wasn't sure if it was the same issue as my code generates the same with or without the discriminators, and it's about primitives instead of refs. I do not get the union code from #1561 in any configuration.

If you think they're related I can post there and close this one :)