Dataclasses are an implementation of a code generator, similar to a Named Tuple. The idea of using a code generator is to save time and reduce wordiness.There are two views about the purpose of Dataclasses:<ol><li>It makes a mutable data holder, in the spirit of named tuples</li><li>It writes boiler-plate code for you, simplifying the process of writing the class</li></ol>These 2 world views are reflected in the name: “Dataclasses”Per the dataclasses PEP, they are roughly a “Mutable named tuple with defaults”They provide an elegant syntax for creating data holder objects

Dataclasses & What are they for

In the image example, the generated code includes:<ul><li>typed attributes</li><li>hidden attributes from __repr__ response</li><li>custom __hash__ dunder with the selected fields</li><li>attribute with default factory value in __init__ constructor</li></ul>

A Complex Example of using Dataclasses

<ul><li>Dunder methods like init, repr, eq, lt, le, etc.</li><li>Frozen nature by generating dunder methods for __setattr__ and __delattr__</li><li>Class variables for the provided attributes</li></ul>The code generation can be configured by providing keyword arguments to the @dataclass decoratorA complete list of methods and configurations can be found <a href="https://docs.python.org/3/library/dataclasses.html">here</a>

Generated Code Includes

If you have an immutable mind get out!

<ul><li class="ql-align-justify">Dataclasses are by default mutable</li><li class="ql-align-justify">Named Tuples have methods which start with an underscore which can be confusing. Ex: obj._replace(f=value)</li><li class="ql-align-justify">Named Tuples have legacy code. Ex: _asdict() returns an OrderedDict</li><li class="ql-align-justify">Named Tuples can be unpacked</li><li class="ql-align-justify">Dataclasses are faster at retrieving data but take up more space</li><li class="ql-align-justify">Dataclasses use functions instead of methods. Ex: asdict() vs nt._asdict()</li><li class="ql-align-justify">Dataclasses are underlying stored as instance dicts</li></ul>

Comparison with Named Tuples

In the beginning there where:<ul><li>dicts for look-ups</li><li>tuples for structures</li><li>hand-written classes for more functionality</li></ul>Later Named Tuples where added as a code generator with the goal of adding names to the fields of a tuple. This was limited by design, hence its nameORMs appeared and pioneered using class attributes to specify rich data structuresIn Python 3.6, it became possible to supply <a href="https://deepstash.com/article/137403/an-introduction-to-python-type-hints">type annotations</a>

A History Lesson

I started using Dataclasses and I think you should too

Speaker: Raymond Hettinger The PEP 557 dataclasses module is available in starting in Python 3.7. It will become an essential part of every Python programmer's toolkit. This talk shows what problem the module solves, explains its key design decisions, and provides practical examples of how to put it to work. Dataclasses are shown to be the next step in a progression of data aggregation tools: tuple, dict, simple class, bunch recipe, named tuples, records, attrs, and then dataclasses. Each builds upon the one that came before, adding expressiveness at the expense of complexity. Dataclasses are unique in that they let you selectively turn-on or turn-off its various capabilities and it lets the user choose the underlying data store (either instance dictionary, instance slots, or an inherited base class). Dataclasses and typing.NamedTuple both use variable annotations which were new in Python 3.6. Slides can be found at: https://speakerdeck.com/pycon2018 and https://github.com/PyCon/2018-slides

computerscience

Computer Science

Speaker: Raymond Hettinger

The PEP 557 dataclasses module is available in starting in Python 3.7.   It will become an essential part of every Python programmer's toolkit.  This talk shows what problem the module solves, explains its key design decisions, and provides practical examples of how to put it to work.

Dataclasses are shown to be the next step in a progression of data aggregation tools: tuple, dict, simple class, bunch recipe, named tuples, records, attrs, and then dataclasses. Each builds upon the one that came before, adding expressiveness at the expense of complexity.

Dataclasses are unique in that they let you selectively turn-on or turn-off its various capabilities and it lets the user choose the underlying data store (either instance dictionary, instance slots, or an inherited base class).

Dataclasses and typing.NamedTuple both use variable annotations which were new in Python 3.6.









Slides can be found at: https://speakerdeck.com/pycon2018 and https://github.com/PyCon/2018-slides