[BUG][Python] Model template doesn't properly handle nested dicts and arrays.

[ottmar-zittlau]

Bug Report Checklist

• Have you provided a full/minimal spec to reproduce the issue?
• Have you validated the input using an OpenAPI validator?
• Have you tested with the latest master to confirm the issue still exists?
• Have you searched for related issues/PRs?
• What's the actual output vs expected output?
• [Optional] Sponsorship to speed up the bug fix or feature request (example)

Description

I have some components that contain dicts of dicts. With the default templates the generator doesn't generate valid python-code for these components.

Here is the generated code: https://gist.github.com/ottmar-zittlau/9d2d98ffd9b29c3bcb34b9e947228354

Some bugs:

Here the class on which from_dict is called is a type.

"array_of_dictionaries": [Dict[str, Vector3d].from_dict(_item) for _item in obj["array_of_dictionaries"]] if obj.get("array_of_dictionaries") is not None else None,

Here _item_array_of_dictionaries is a dictionary that doesn't have a to_dict method.

        if self.array_of_dictionaries:
            for _item_array_of_dictionaries in self.array_of_dictionaries:
                if _item_array_of_dictionaries:
                    _items.append(_item_array_of_dictionaries.to_dict())
            _dict['array_of_dictionaries'] = _items

openapi-generator version

7.16.0

OpenAPI declaration file content or url

https://gist.github.com/ottmar-zittlau/2e2a5357b1d1e60d41a2aef4ee6bde14

Generation Details

I generated code for python.

Steps to reproduce

Run

./openapi-generator-cli generate --generator-name python --input-spec openapi.yaml --output generated --package-name example

Related issues/PRs

Suggest a fix

I already have adapted the model_generic.mustache template file to deal with this, but before opening a merge request I wanted to know if this bug is known and whether there is already work going on to fix it.

[fa0311]

Can't reproduce this on current master (7.24.0-SNAPSHOT) anymore — array-of-map-of-model round-trips fine through to_dict/from_dict now. I think #23538 covered this. Please retest with the latest version and let me know if it's still broken for you.

[TypeError92]

Similar issue here on 7.24.0, and I don't see any changes in 7.25.0 that would suggest that this has been fixed.

Case 1:
Field: things: dict[str, dict[str, list[Thing]]]

to_dict override:

        _field_dict_of_dict = {}
        if self.things:
            for _key_things, _value_things in self.things.items():
                if _value_things is not None:
                    _field_dict_of_dict[_key_things] = {
                        _key: _value.to_dict() for _key, _value in _value_things.items()
                    }
            _dict["things"] = _field_dict_of_dict

-> _value.to_dict() errors because _value is not a model instance but a list of model instances.

Case 2:
Field: things: dict[str, dict[str, dict[str, Thing]]] | None

to_dict override:

        _field_dict_of_dict = {}
        if self.things:
            for _key_things, _value_things in self.things.items():
                if _value_things is not None:
                    _field_dict_of_dict[_key_things] = {_key: _value.to_dict() for _key, _value in _value_things.items()}
            _dict["things"] = _field_dict_of_dict

-> _value.to_dict() errors because _value is not a model instance but a dict of model instances.

I was able to patch this up locally by adding these cases to a custom template, but I'm wondering if playing Whack-A-Mole with nested fields in the to_dict method like this is the right approach? What if instead there was an independent, hard-coded to_dict function that can handle arbitrarily nested fields via recursion, and then the model method just fed self into that?