Skip to main content

properties: an organizational aid and wrapper for validation and tab completion of class properties

Project description

Latest PyPI version MIT license ReadTheDocs Travis tests Code coverage

Overview Video

Python Properties

An overview of Properties, November 2016.


Properties provides structure to aid development in an interactive programming environment while allowing for an easy transition to production code. It emphasizes usability and reproducibility for developers and users at every stage of the code life cycle.


The properties package enables the creation of strongly typed objects in a consistent, declarative way. This allows validation of developer expectations and hooks into notifications and other libraries. It provides documentation with no extra work, and serialization for portability and reproducibility.


  • Keep a clean namespace for easy interactive programming

  • Prioritize documentation

  • Provide built-in serialization/deserialization

  • Connect to other libraries for GUIs and visualizations


API Documentation is available at ReadTheDocs.


  • attrs - “Python Classes Without Boilerplate” - This is a popular, actively developed library that aims to simplify class creation, especially around object protocols (i.e. dunder methods), with concise, declarative code.

    Similarities to Properties include type-checking, defaults, validation, and coercion. There are a number of differences:

    1. attrs acts somewhat like a namedtuple, whereas properties acts more like a dict or mutable object.

      • as a result, attrs is able to tackle hashing, comparison methods, string representation, etc.

      • attrs does not suffer runtime performance penalties as much as properties

      • on the other hand, properties focuses on interactivity, with notifications, serialization/deserialization, and mutable, possibly invalid states.

    2. properties has many built-in types with existing, complex validation already in place. This includes recursive validation of container and instance properties. attrs only allows attribute type to be specified.

    3. properties is more prescriptive and detailed around auto-generated class documentation, for better or worse.

  • traitlets (Jupyter project) and traits (Enthought) - These libraries are driven by GUI development (much of the Jupyter environment is built on traitlets; traits has automatic GUI generation) which leads to many similar features as properties such as strong typing, validation, and notifications. Also, some Properties features and aspects of the API take heavy inspiration from traitlets.

    However, There are a few key areas where properties differs:

    1. properties has a clean namespace - this (in addition to ? docstrings) allows for very easy discovery in an interactive programming environment.

    2. properties prioritizes documentation - this is not explicitly implemented yet in traits or traitlets, but works out-of-the-box in properties.

    3. properties prioritizes serialization - this is present in traits with pickling (but difficult to customize) and in traitlets with configuration files (which require extra work beyond standard class definition); in properties, serialization works out of the box but is also highly customizable.

    4. properties allows invalid object states - the GUI focus of traits/traitlets means an invalid object state at any time is never ok; without that constraint, properties allows interactive object building and experimentation. Validation then occurs when the user is ready and calls validate

    Significant advantages of traitlets and traits over properties are GUI interaction and larger suites of existing property types. Besides numerous types built-in to these libraries, some other examples are trait types that support unit conversion and NumPy/SciPy trait types (note: properties has a NumPy array property type).

  • param - This library also provides type-checking, validation, and notification. It has a few unique features and parameter types (possibly of note is the ability to provide dynamic values for parameters at any time, not just as the default). This was first introduced before built-in Python properties, and current development is very slow.

  • built-in Python dataclass decorator - provides “mutable named tuples with defaults” - this provides similar functionality to the attrs by adding several object protocol dunder methods to a class. Data Classes are clean, lightweight and included with Python 3.7. However, they don’t provide as much builtin functionality or customization as the above libraries.

  • built-in Python property - properties/traits-like behavior can be mostly recreated using @property. This requires significantly more work by the programmer, and results in not-declarative, difficult-to-read code.

  • mypy, PEP 484, and PEP 526 - This provides static typing for Python without coersion, notifications, etc. It has a very different scope and implementation than traits-like libraries.


  • casingSimulations - Research repository for electromagnetic simulations in the presence of well casing

  • OMF - Open Mining Format API and file serialization

  • SimPEG - Simulation and Parameter Estimation in Geophysics

  • Steno3D - Python client for building and uploading 3D models


To install the repository, ensure that you have pip installed and run:

pip install properties

For the development version:

git clone
cd properties
pip install -e .


Lets start by making a class to organize your coffee habits.

import properties
class CoffeeProfile(properties.HasProperties):
    name = properties.String('What should I call you?')
    count = properties.Integer(
        'How many coffees have you had today?',
    had_enough_coffee = properties.Bool(
        'Have you had enough coffee today?',
    caffeine_choice = properties.StringChoice(
        'How do you take your caffeine?' ,
        choices=['coffee', 'tea', 'latte', 'cappuccino', 'something fancy'],

The CoffeeProfile class has 4 properties, all of which are documented! These can be set on class instantiation:

profile = CoffeeProfile(name='Bob')

Out [1]: Bob

Since a default value was provided for had_enough_coffee, the response is (naturally)


Out [2]: False

We can set Bob’s caffeine_choice to one of the available choices; he likes coffee

profile.caffeine_choice = 'coffee'

Also, Bob is half way through his fourth cup of coffee today:

profile.count = 3.5

Out [3]: ValueError: The 'count' property of a CoffeeProfile instance must
         be an integer.

Ok, Bob, chug that coffee:

profile.count = 4

Now that Bob’s CoffeeProfile is established, properties can check that it is valid:


Out [4]: True

Property Classes are auto-documented in Sphinx-style reStructuredText! When you ask for the doc string of CoffeeProfile, you get

**Required Properties:**

* **count** (:class:`Integer <properties.basic.Integer>`): How many coffees have you had today?, an integer, Default: 0
* **had_enough_coffee** (:class:`Bool <properties.basic.Bool>`): Have you had enough coffee today?, a boolean, Default: False
* **name** (:class:`String <properties.basic.String>`): What should I call you?, a unicode string

**Optional Properties:**

* **caffeine_choice** (:class:`StringChoice <properties.basic.StringChoice>`): How do you take your caffeine?, any of "coffee", "tea", "latte", "cappuccino", "something fancy"

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

properties-0.6.1.tar.gz (52.0 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page