Swagger Resolution Errors

[Goldziher]

Expected Behaviour

Minimal Example Spec

Actual Behaviour

Steps to Reproduce

Environment

• OS: maxOf
• Python version: 3.9
• Swagger/OpenAPI version used: swagger 2.0
• Backend: (flex, swagger-spec-validator or openapi-spec-validator) using the resolver directly

Hi there, and first off thanks for this resolver!

I'm maintaining another library that relies on Prance as a derferencer. It works well in most cases, but I think I encountered a bug. I'm testing our library (https://github.com/snok/drf-openapi-tester) against various specs, and now I'm trying to test it against the kubernetes swgger 2.0 docs:, which can be found here: https://raw.githubusercontent.com/kubernetes/kubernetes/master/api/openapi-spec/swagger.json

It seems that this spec cannot be resolved at present because instead of a url what is passed to the following function is a dictionary {type: string}:

def absurl(url, relative_to = None):
....
# prance.util.url lines 49-59
  parsed = url
  if not isinstance(parsed, tuple):
    from .fs import is_pathname_valid
    if is_pathname_valid(url):
      from . import fs
      url = fs.to_posix(url)
    try:
      parsed = parse.urlparse(url)
    except Exception as ex:
      from .exceptions import raise_from
      raise_from(ResolutionError, ex, 'Unable to parse url: %s' % (url,))

I got to this point in debugging it.

According to the intellij debugger absurl was called with two parameters:

url={'type': 'string'},
relative_to=ParseResult(scheme='file', netloc='', path='/', params='', query='', fragment='/definitions/io.k8s.apiextensions-apiserver.pkg.apis.apiextensions.v1.JSONSchemaProps')

If you want I can add a json copy of the kubernetes spec to the repo and have it fail in a test if that would help :).

@jfinkhaeuser

[jfinkhaeuser]

Thank you for this!

It is true that validation and dereferencing do not have to happen at the same time. In fact, this is also how it's implemented - the validation function is just called internally.

I think that makes for a very easy change.

As to the bug: the spec could help, yes! Also, if/when I have time, #81 would help with locating the error. In the past, people passed malformed specs to get this kind of problem, but that doesn't mean it has to be the case here!

[Goldziher]

great, i will create the test branch and maybe add a PR if I figure it out 😉

[Goldziher]

@jfinkhaeuser, i added a PR with the Kubernetes swagger specs and a failing test.