mathewmarcus / marshmallow-pynamodb Goto Github PK

It appears that the ModelSchema is somehow unable to introspect the attributes of a model if it is a subclass of another model. Subclassing in this manner works fine in pynamodb, and it is expected that it would work here as well.

Test case:

from pynamodb.attributes import UnicodeAttribute
from pynamodb.models import Model
from marshmallow_pynamodb import ModelSchema


class Base(Model):
    class Meta:
        table_name = 'test-table'
    pk = UnicodeAttribute(hash_key=True)
    alpha = UnicodeAttribute(null=True)


class Sub(Base):
    class Meta(Base.Meta):
        pass


class BaseSchema(ModelSchema):
    class Meta:
        model = Base


class SubSchema(ModelSchema):
    class Meta:
        model = Sub


sub_schema = SubSchema()
base_schema = BaseSchema()

base_dumped = base_schema.dump(Base(pk='peekay', alpha='hi mom')).data
assert base_dumped['alpha'] == 'hi mom' # <-- PASSES OK

sub_dumped = sub_schema.dump(Sub(pk='peekay', alpha='hi mom')).data
assert sub_dumped['alpha'] == 'hi mom' # <-- FAILS

could you please add support for this field type

Hi.
I'm trying to use the package but I have a lot of fields in python mapped to fields named differently in Dynamo (using the attr_name).
Each of those fields results in a None value when serializing it, and in addition it uses the attr_name instead of the field name.
Example:

class Result(Model):
    app_id = NumberAttribute(attr_name="appId")
    domain = UnicodeAttribute()

class ResultSchema(ModelSchema):
    class Meta:
        model = Result

result = Result(app_id=5, domain='a.com')
ResultSchema().dump(result).data
>>> {'domain': 'a.com', 'appId': None}

Expected:
{'domain': 'a.com', 'app_id': 5}

Changing the model's attributes to the original attribute names is not an option as part of them have symbols that are not supported by python as valid names (appId#version for example)

Thanks

I updated the packages of my project and had problems with the use of this plugin.
I think it's related to the Breaking changes in the major relase 4.0 of pynamodb I installed.
To be honest I am not 100% sure it was this project but I think the changes to pynamodb could cause problems here

Currently, Binary types are not supported. Any chance of getting these added? I'm willing to submit a PR.

Hi!
There is a problem with ListAttribute():

from pynamodb.models import Model
from marshmallow_pynamodb import ModelSchema


class User(Model):
    nick_names = ListAttribute()


class UserSchema(ModelSchema):
    class Meta:
        model = User

/marshmallow_pynamodb/schema.py", line 39, in get_declared_fields
    element_type = type(attribute.element_type.__name__, (ModelSchema, ), {'Meta': Meta})
AttributeError: 'NoneType' object has no attribute '__name__'

It is understandable that is not so much sense for schema with list without type specification but it is valid model and possible use case.

So it seems two way to handle this issue. First option is forbid create list attribute without specifying type and second is handle this situation and create schema and validate only container type.