Missing type properties in some Entities

[Ndiritu]

We currently clean properties with the @odata. prefix by dropping the prefix e.g. @odata.type becomes type.

A derived type's type property will NOT be added to the model since it's parent already contains a property named type (odata type)

Example: microsoftgraph/msgraph-sdk-dotnet@00ffddc#diff-6a2934f9f1e41fcb43a8a28293569fd779095ce3bc9ff09b9f91a1f6ef22d0cdL99-L102
or check SubjectRightsRequest history

Proposed fix:
Doing away with stripping prefixes but instead only replace special characters & camel case?

[baywet]

Thanks for reporting this, already on top of it (for Go/Java) could you have a look at PHP with the fix present in #1739 please?

[Ndiritu]

I think the PR doesn't solve this:

Here's the openAPI spec snippet. There's already a type property here that is a microsoft.graph.subjectRightsRequestType and it's base class Entity has the "@odata.Type" which is cleaned to type. This results in the SubjectRightsRequest model not containing a type property because we find a property with the same name in the parent.

    microsoft.graph.subjectRightsRequest:
      allOf:
        - $ref: '#/components/schemas/microsoft.graph.entity'
        - title: subjectRightsRequest
          type: object
          properties:
            assignedTo:
              anyOf:
                - $ref: '#/components/schemas/microsoft.graph.identity'
              description: Identity that the request is assigned to.
              nullable: true
           ...
            type:
              anyOf:
                - $ref: '#/components/schemas/microsoft.graph.subjectRightsRequestType'
              description: 'The type of the request. Possible values are: export, delete, access, tagForAction, unknownFutureValue.'
              nullable: true
      x-ms-discriminator-value: '#microsoft.graph.subjectRightsRequest'

[baywet]

Ah thanks for the précision. I didn't get that. So would you argue that we should strip the prefix but only clean the special characters?

[Ndiritu]

I think we should only clean special characters and not strip the prefix. Stripping the prefix may mean we have to check if the property already exists and de-dupe names.

[Ndiritu]

I think such duplicates would only apply in OData to OpenAPI scenarios

Otherwise, I think "traditional" openAPI should take care of duplicate property names from the yml/json validation

[baywet]

Makes sense. I'll add that to my pull request tomorrow. @darrelmiller any objections? (currently @odata.type gets converted to type for the code symbol and that creates conflicts if the schema already describes a type property)