Using middle

After installing middle, it is easy to use it in your project. Most of its functionalities are acessible within the middle module already:

import middle

middle consists basically in three parts:

  • middle.Model: the base class that needs to be inherited to declare all your models (except submodels);
  • middle.field: a function that is used to declare your field models; and
  • middle.asdict: a function required to convert your model instances to Python primitives.


The middle.Model class is the heart of middle. To have all middle functionality at your disposal, middle.Model needs to be subclassed when declaring your models:

class MyModel(middle.Model):
    name: str = middle.field()  # Python 3.6+ syntax
    name = middle.field(type=str)  # Python 3.5 syntax

In essence, middle.Model started as a syntactic sugar for the attr.s decorator but soon evolved to a more complex design, implementing its own metaclass to handle some aspects of its models and fields.


Since middle.Model already implements its own metaclass, it should be wise not to mix it with other classes that have a metaclass different than type.

To create an instance of your model, you can:

  • Use keyword arguments:

    >>> MyModel(name="foo")
  • Use a dict:

    >>> MyModel({"name": "foo"})
  • Use a dict as **kwargs:

    >>> MyModel(**{"name": "foo"})
  • Use any object instance that have acessible attributes with the same name as the required ones from your model:

    >>> MyModel(some_obj_with_name_accessible)


The middle.field function is used to declare your model’s fields, with support to the type definition and other options that can be used later to define your model behavior regarding converting input values, validating values and format values for Python primitives. middle.field makes heavy usage of attr.ib calls, specially to store information into the metadata dict.

There are three ways to declare your fields inside middle.Model, you don’t have to necessarily use middle.field, though it will be called under the hood to have a uniform model.

Declaring models, using middle.field and typing hints and annotations (PEP-526, for Python 3.6+):

class MyModel(middle.Model):
    id: int = middle.field()
    name: str = middle.field(min_length=5)
    active: bool = middle.field(default=False)
    created_on: datetime = middle.field(default=None)

Declaring models, using middle.field and type keyword (Python 3.5 compatible):

class MyModel(middle.Model):
    id = middle.field(type=int)
    name = middle.field(type=str, min_length=5)
    active = middle.field(type=bool, default=False)
    created_on = middle.field(type=datetime, default=None)

Declaring models, without middle.field, using typing hints, annotations (Python 3.6+ only) and a dict:

class MyModel(middle.Model):
    # id: int  # or ...
    id: int = {}
    name: str = {"min_length": 5}
    active: bool = {"default": False}
    created_on: datetime = {"default": None}

Declaring models, without middle.field, using only a dict (Python 3.5 compatible):

class MyModel(middle.Model):
    id = {"type": int}
    name = {"type": str, "min_length": 5}
    active = {"type": str, "default": False}
    created_on = {"type": datetime, "default": None}

Declaring models, without middle.field, using only typing hints and annotations (inspired by pydantic, works only with Python 3.6+):

class MyModel(middle.Model):
    id: int
    name: str
    active: str
    created_on: datetime


Declaring models using only typing hints annotations will not enable support for keyword embed validators.

Declaring models, the chaotic way (won’t work on Python 3.5):

class MyModel(middle.Model):
    id: int
    name = {"type": str, "min_length": 5}
    active: bool = middle.field(default=False)
    created_on = middle.field(type=datetime, default=None)


Developers are free to choose their preferred style (matching the Python version), although sticking to one can help readabilty.


This method, provided with an instance of a middle.Model class, will return a dict of key-values that will reflect the data of the instance against the model typing hints only.

>>> instance = MyModel(
...     id=42,
...     name="foo bar",
...     created_on=datetime.utcnow()
... )

>>> instance
MyModel(id=42, name='foo bar', active=False, created_on=datetime.datetime(2018, 7, 5, 14, 14, 12, 319270))

>>> middle.asdict(instance)
{'id': 42, 'name': 'foo bar', 'active': False, 'created_on': '2018-07-05T17:14:12.319270+00:00'}