pip install marshmallow-models
Feature requests, feedback, issues, and pull requests are welcome and appreciated.
Please follow the Marshmallow Contributing Guidelines.
Inspired by Schematics, powered by Marshmallow.
Whereas Marshmallow is an excellent serialization/deserialization and validation library, it wasn't intended to be a class or type definition library, which Schematics was.
This library provides a Schematics-like Model but using Marshmallow's Fields and validation. This library also intentionally maintains the usage style of marshmallow so that users of Marshmallow Schemas will be able to use these Models easily.
Models are defined like Schemas, but whereas a Schema is instantiated with parameters and then used to schema.dump(data) or schema.load(data), or schema.validate(data), Models are instantiated, attributes may be assigned to them, and then they can be .dump()'d, .dumps()'d or .validate()'d.
from marshmallow_models import Model
from marshmallow.fields import String, Integer
class PersonModel(Model):
name = String(required=True)
age = Integer(required=True)
person = PersonModel()
person.name = 'Tester'
person.age = 100
# or equivalently:
person = PersonModel({'name': 'Tester', 'age': 100})
# or equivalently:
person = PersonModel(name='Tester', age=100)
# throws marshmallow.exceptions.ValidationError if invalid
person.validate()
person.dump().data # {'name': 'Tester', 'age': 100}
class PersonModel(Model):
name = String(missing='Anonymous')
age = Integer(default=0)
person = PersonModel()
person.name # 'Anonymous'
person.age # 0
Default and missing parameters may be provided as they are to Marshmallow Schemas.
Constructing a model is treated like
"loading" data (as in, schema.load(data)). If attributes are
missing and a missing
configuration was provided, those values
will be assigned to the missing attributes.
Reading attributes is treated like "dumping" data (as in, schema.dump(data)), as are calls to model.dump() and dumps(). If a value doesn't exist when read or dumped, the default value will be substituted for that attribute.
In many cases default
and missing
can be used interchangeably
in the context of Models, but there may be cases where their
different treatment is meaningful.
Nested models are also supported.
class ParentModel(PersonModel):
child = NestedModel(PersonModel)
parent = ParentModel(name='Tester', age=40, child=dict(name='Child', age=10))
self.assertEqual(parent.child.name, 'Child')
parent.child.name = 'Kid'
self.assertEqual(parent.child.name, 'Kid')
Marshmallow Models support the "class Meta" configuration method.
An additional Meta attribute is supported: strict_constructor
.
In Marshmallow Schemas, transformation of input data to output data was a single step process. In Marshmallow Models, it might be reasonable for users to instantiate a model with incomplete attributes and then fill in the attributes before attempting to validate() or dump() the data.
By default, even for Models with strict = True
the constructor does not raise exceptions for incomplete attributes.
If exceptions are wanted in this case, set strict_constructor = True
.
class PersonWithStrictConstructorModel(Model):
class Meta:
strict_constructor = True
name = String(required=True)
age = Integer(required=True)
with self.assertRaises(ValidationError):
person = PersonWithStrictConstructorModel()