Python: `obj.to_dict()` methods return wrong structures because they do not use `attribute_map`

Description

Generated model code does not make any use of the attribute_map table which holds the mapping between the OpenAPI property names and pythonic property names.

If you have an object with camelCaseProperty, then swagger will generate a model with the camel_case_property field and record the mapping in the attribute_map. Unfortunately, when calling obj.to_dict() this map is not used and the resulting object will be incorrect.

The proper conversion is be performed by the ApiClient.sanitize_for_serialization method which requires first creating the client object. I'm arguing that this conversion should be built-in the obj.to_dict() methods. These methods should never return incorrect objects.

Swagger-codegen version

2.3.1

Swagger declaration file content or url

Command line used for generation

java -jar swagger-codegen-cli.jar generate -l python -i experiment.swagger.json -o $DIR -c config.json

Suggest a fix/enhancement

The attribute_map conversion should be built-in the obj.to_dict() methods.

The issue concerns generated Python model obj.to_dict() methods and the ApiClient.sanitize_for_serialization path. Start by tracing the Python generation template and compare both conversion paths, using the supplied command to reproduce a model with a camelCaseProperty. Done means to_dict() applies attribute_map and returns the same property-key structure as the client serializer.