Unexpected ibm-api-symmetry warning with dictionary

[vsandonis]

I am getting validation warning ibm-api-symmetry with an openapi spec that includes a dictionary following https://swagger.io/docs/specification/v3_0/data-models/dictionaries/ :

$ lint-openapi error_example.yaml --log-level ibm-api-symmetry=info
IBM OpenAPI Validator (validator: 1.20.0), @Copyright IBM Corporation 2017, 2024.

Validation Results for error_example.yaml:

[INFO] ibm-api-symmetry: checking variant schema DevicePrototype against canonical schema Device
[INFO] ibm-api-symmetry: schema defines 'additionalProperties' - it is not a proper graph fragment
[INFO] ibm-api-symmetry: nested object property input_parameters is not a graph fragment of canonical property input_parameters
[INFO] ibm-api-symmetry: nested object property device_config is not a graph fragment of canonical property device_config
[INFO] ibm-api-symmetry: nested object property details is not a graph fragment of canonical property details
[INFO] ibm-api-symmetry: variant schema DevicePrototype is not a graph fragment of canonical schema Device
Warnings:

  Message :   Variant schema should be a graph fragment of the canonical schema
  Rule    :   ibm-api-symmetry
  Path    :   components.schemas.DevicePrototype
  Line    :   120

Summary:

  Total number of errors   : 0
  Total number of warnings : 1

  Warnings:
   1 (100%) : Variant schema should be a graph fragment of the canonical schema

I am using this version: validator: 1.20.0; ruleset: 1.19.0 (default)

Here is the yaml file that generates the warning: error_example.txt

Is this a legitimate warning?

I guess it is not because if I define input_parameters in DeviceConfiguration as a string instead of a dictionary, no warning is generated, so I guess it has to do with the dictionary definition. In addition, if I comment the path /devices/{device_id} (lines 14 to 36 in the file) no warning is generated either.

Is there anyway to skip the validation for this input_parameters?

Thanks for your help.

[vsandonis]

Taking a look to the code in here, it seems that if the variant defines additionalProperties it is not considered a graph fragment.

[dpopp07]

Taking a look to the code in here, it seems that if the variant defines additionalProperties it not considered a graph fragment.

That's correct - if the variant schema is a dictionary, it may contain properties that are not (or even cannot) be represented on the canonical schema, which would violate the graph fragment pattern.

[vsandonis]

That's correct - if the variant schema is a dictionary, it may contain properties that are not (or even cannot) be represented on the canonical schema, which would violate the graph fragment pattern.

Thanks for the answer. Does it mean that the use of dictionaries is not allowed in variant schemas?

[dpopp07]

Does it mean that the use of dictionaries is not allowed in variant schemas?

Upon looking at your API in more detail, I actually think that this case may be allowed since the two schemas are using the same dictionary property. I will go over the code and confirm whether or not this is intended behavior or a bug. Thanks for opening the issue.

[dpopp07]

@vsandonis after review, this should not be a warning. I'll mark this as a bug and will update the code soon.