Support for multiple content types in error response

Utilisez la réponse 403 OpenAPI fournie avec application/json et text/plain pour reproduire le comportement actuel du client généré. Inspectez le chemin de réponse d’erreur du client généré et exécutez la génération pertinente du client ainsi que les tests de réponses, s’ils sont disponibles. C’est terminé lorsque le client généré sélectionne le traitement à partir du Content-Type de la réponse, afin que les détails JSON restent disponibles sans tenter d’analyser text/plain comme du JSON.

Describe the bug
Similar to previous issues about multiple content types support like https://github.com/openapi-generators/openapi-python-client/issues/453 or https://github.com/openapi-generators/openapi-python-client/issues/1004

For a single endpoint, the same error code can be returned by different piece of software. For example, an application can return error details as application/json but have some infra elements in front (cloud gateway/load balancer/reverse proxy/firewall) that can block request and also return their own errors in their own format text/json/xml (usually for network errors 401,403,502,503,504).

In my use case, I realized a firewall can sometimes block requests and returns a text/plain with a string Access Forbidden. So I'm trying to represent that in my OpenAPI spec so the generated client can handle these errors properly without crashing, and this works for some of the other openapi generators. But it doesn't work well with this python client, because the content type is not validated in the generated code. It only depends on the first content declared in the schema and seems to ignore the other ones.

So I'm getting the following implementation:

if response.status_code == 403:
  response_403 = APIError.from_dict(response.json())

Which works for json error, but json parsing throw an exception if it's something else (like text/plain)

And if I declare text/plain before application/json, I get:

if response.status_code == 403:
  response_403 = response.text

But in that case, I'm losing any detail about the json error.

What I expect here, is something similar to the oapi-codegen golang implementation.
If there are multiple content types defined for an error code, check that the response content type match before handling it.
So each content type can coexist and be handled separately.

case strings.Contains(rsp.Header.Get("Content-Type"), "json") && rsp.StatusCode == 403:
    // this is a json error
    var dest APIError
    if err := json.Unmarshal(bodyBytes, &dest); err != nil {
      return nil, err
    }
    response.JSON403 = &dest
case rsp.StatusCode == 403:    
    ... // default case for text/plain error

OpenAPI Spec File

403:
  description: Forbidden
  content:
    application/json:
      schema:
        $ref: '#/components/schemas/APIError'
    text/plain:
      schema:
        type: string

Desktop (please complete the following information):

• OS: macOS 14.7.4
• Python Version: 3.11.88
• openapi-python-client version: 0.24.3