Fix missed response schema cases, add backup_name

[wazawoo]

I ran the cli on an API spec and it didn't generate anything. This spec had responses but no request bodies or other schemas. Also, the operations didn't have operation_ids.

I still wanted to generate models from these, so I made a couple changes:

• Fix models from inline responses not getting added to inline_models
• Add backup_name to uniquely specify operations with no operation_id

I can add tests and/or split into different PRs, but first I wanted to see of you were interested in either of these changes.

[denislituev]

Hello @wazawoo. Thank you for your contribution, I will check this PR.

I’d like to clarify the “it didn’t generate anything” part.
Even if the spec only contains inline response schemas and no request bodies or operation_ids, the generator should still produce models from response schemas.

If literally no output was generated, that would indicate a bug.

No need to split this for now.
I’ll review the changes and test them first.

[wazawoo]

@denislituev sounds good, thanks!

I added a test that I believe would fail before this PR (because the inline responses weren't added to inline_models)

[wazawoo]

I also noticed that objects in arrays in responses are not parsed. Started looking into it.

For example:

{
    "paths": {
        "example_path": {
            "get": {
                "summary": "...",
                "responses": {
                    "200": {
                        "description": "...",
                        "headers": {},
                        "content": {
                            "application/json": {
                                "schema": {
                                    "type": "array",
                                    "items": {
                                        "type": "object",
                                        "properties": {
                                            "exampleField": {
                                                "type": "string"
                                            },
                                            "anotherExampleField": {
                                                "type": "string"
                                            }
                                        }
                                    }
                                },
                                "example": []
                            }
                        }
                    }
                },
                "tags": [],
                "description": "..."
            },
            "parameters": []
        }
    }
}

[wazawoo]

Added more tests and fixed formatting issues. I don't love the long and verbose names I've given for the response objects.

I have one more case that's broken with the spec im using as input: A response object, with a parameter with an array of objects.

{
...
 "schema": {
    "type": "object",
    "properties": {
        "exampleProperty": {
            "type": "string"
        },
        "objectArray": {
            "type": "array",
            "items": {
                "type": "object",
                "properties": {
                    "propA": {
                        "type": "string"
                    },
                    "propB": {
                        "type": "string"
                    }
                }
            }
        }
    }
},
...
}

I think in general something might be missing with handling nested objects and arrays of objects. Not sure I'm approaching the fix in the cleanest way though.

[denislituev]

Minor suggestion: The operation_name logic is duplicated on lines 191 and 232. Consider extracting it to the top of process_operation to avoid repetition.

[wazawoo]

Thanks! Resolved here: 5bf7ccd