type field is now placed inside of schema node for referenced objects

[AHBaber]

I have repeatedly gotten an error when validating objects passed into @doc.consumes(). The error being:

Failed validating 'oneOf' in schema['properties']['paths']['patternProperties']['^/']['properties']['post']['properties']['parameters']['items']:
{'oneOf': [{'$ref': '#/definitions/parameter'},
           {'$ref': '#/definitions/jsonReference'}]}

Specific error instance:

On instance['paths']['/endpointtest']['post']['parameters'][0]:
{'description': 'data object description',
 'in': 'body',
 'name': 'PassedInDataObject',
 'required': True,
 'schema': {'$ref': '#/definitions/PassedInDataObject', 'x-scope': ['']},
 'type': 'object'}

When we look at the documentation for the OpenAPI Specification we can find several examples that show that the "type" field should be inside the "schema" block. Whereas in the example above, the "type" field is at the same level as the "schema" block.

Here are some links to the examples listed in the OpenAPI Specification showing that "type" should be inside of "schema":

• Path Item Object Example
• Operation Object Example
• Parameter Object Examples
• Request Body Examples
• Response Object Examples

Solution:

In the sanic_openapi/swagger.py file, at line 213, we can see that if a field named "$ref" is dedicated, that a "schema" block is created, and the "$ref" is placed inside the "schema" block.

The new code I am submitting also checks to see if a "type" field exists, and then places it inside of the "schema" block as well.

[chenjr0719]

Hi @AHBaber

Could you provide an example of the issue you mentioned? I can not reproduce the error. And, please tell me which version of Sanic-OpenAPI you are using.

BTW, the examples you provide are using OpenAPI Specification 3.0. But, this project only supports OpenAPI Specification 2.0 and I'm not sure your issue is caused by the version compatibility or not.

[AHBaber]

@chenjr0719

versions of involved libraries:

• sanic_openapi: 0.6.0
• sanic: 19.9.0
• swagger-parser: 1.0.1
• swagger-spec-validator: 2.4.3

I was validating the swagger data using swagger parser.

That is using swagger-spec-validator for the actual validation.

Swagger-spec-validator is using what it lists as v2.0/schema.json

[chenjr0719]

After using swagger-spec-validator and reading the code base of this project, I think the problem is about this line:

https://github.com/huge-success/sanic-openapi/blob/3cdf8801836a3b31f19d8f9d19b7d27737a4257f/sanic_openapi/doc.py#L152

So, maybe just remove this line is a better solution. What do you think?

And this is the example I used to verify with swagger-spec-validator:

from sanic import Sanic
from sanic.response import json

from sanic_openapi import doc, swagger_blueprint

app = Sanic()
app.blueprint(swagger_blueprint)

app.config["API_HOST"] = "localhost:8000"
app.config["API_BASEPATH"] = "/"
app.config["API_CONTACT_EMAIL"] = "chenjr0719@gmail.com"
app.config["API_LICENSE_NAME"] = "MIT"
app.config[
    "API_LICENSE_URL"
] = "https://github.com/huge-success/sanic-openapi/blob/master/LICENSE"

app.config["API_SECURITY"] = []
app.config["API_SECURITY_DEFINITIONS"] = {}


class TestSchema:
    id = int


@app.post("/test")
@doc.consumes(
    TestSchema, location="body", required=True, content_type="application/json"
)
@doc.response(200, {"message": str}, description="Works")
async def test(request):
    return json({"message": "TEST"})


app.run(debug=True)

[AHBaber]

I agree that deleting line 152 eliminates the error.

If you could apply that fix, I would deeply appreciate.

Thank you for your time and assistance.

[chenjr0719]

Unfortunately, I can't fix it by myself. I can only review and merge PR which created by other guys, not including my PR. So, let's fix this issue through this PR. 😃

[AHBaber]

I apologize for the delay (holiday)

I pulled the the changes from sanic-openapi and merged them into the master branch of my fork.

I have then made the change we discussed above, and checked that into object_fix_2 so as to avoid cluttering the repos with unnecessary commits.

Do you need me to create a new pull request with the new branch?

[chenjr0719]

Sorry about no response for a couple of days.
I will close this PR first and review your new PR later if you don't mind.