resource_hyperlink.py

class ResourceHyperlink(ma.Field):
    """
    Serializes as a hyperlink to a flask_restful Resource
    """

    def __init__(self, endpoint, url_args=None, *args, **kwargs):
        super().__init__(*args, **kwargs)

        # If the user passes in a list of args, we assume a field with the exact same name appears on the relation
        if isinstance(url_args, list):
            self.url_args = {key: None for key in url_args}
        else:
            self.url_args = url_args

        self.endpoint = endpoint

    def convert_url(self, object):
        args = {}
        for key, value in self.url_args.items():
            if value is None:
                args[key] = getattr(object, key)
            else:
                args[key] = getattr(object, value)

        return url_for(self.endpoint, **args)

    def _serialize(self, value, attr, obj):
        """
        :param value: The current value of this attribute (the SQLAlchemy model)
        :param attr: The name of the attribute
        :param obj: The current object we're serializing
        """
        if isinstance(value, InstrumentedList):
            return [self.convert_url(relation) for relation in value]
        else:
            return self.convert_url(value)

    def _deserialize(self, value, attr, data):
        raise NotImplementedError()

Copied from marshmallow-code/flask-marshmallow#144

You use it like this:

class ReportSchema(ModelSchema):
    samples = ResourceHyperlink(endpoint='rest_api.sample', url_args=[
        'report_id',
        'sample_id'
    ])

Basically you can either provide url_args as a list of keys that we take directly from the related model to apply to the URL (in this case, report_id and sample_id), or you can make that a dictionary that maps URL segments to model properties as I explained above.

I also require that you pass in an endpoint rather than a resource object, just because there will be issues with circular dependencies if you import the resources from the schema and then import the schema in the resources in order to dump/load. The endpoint string for a given resource will be module.path.resource_name, although you can override it with the restful.add_resource() function.

I also recommend that you prefetch the related objects for this field, using query.options(joinedload(user_models.User.roles)), since my class makes use of the related object. If you don't do this, the efficiency will be much worse