Python course / Advanced

Dataclasses and the data model

Let a dataclass write the boilerplate, and learn the methods Python calls on your behalf. A changeable object used as a dictionary key disappears from the dictionary the moment you change it.

fields
default_factory
__post_init__
frozen and replace
ordering
hashing
__call__
descriptors

Start this module

Runs in your browser. Start straight away; create an account when you want your progress kept.

frozen records, changed by replacement

from dataclasses import dataclass, replace


@dataclass(frozen=True, order=True)
class Item:
    name: str
    pence: int
    tags: tuple[str, ...] = ()

    def __post_init__(self):
        if self.pence < 0:
            raise ValueError("pence cannot be negative")


tea = Item("tea", 350, ("hot",))
cheaper = replace(tea, pence=300)

print(tea)
print(cheaper)
print("sorted:", sorted([tea, cheaper])[0].pence)

Item(name='tea', pence=350, tags=('hot',))
Item(name='tea', pence=300, tags=('hot',))
sorted: 300

What you will learn

Replace hand-written __init__, __repr__ and __eq__ with a dataclass, and know where it gets them from.
Give a dataclass field a list or dictionary default safely, with field(default_factory=...).
Validate fields and fill in derived ones after a dataclass's generated __init__ has run.
Make a dataclass that cannot be changed after creation, and make changed copies of it instead.
Choose between a dataclass, a NamedTuple, a TypedDict and a plain class by what the data has to do.
Write a descriptor that controls an attribute on every class that uses it, and see property and methods as descriptors.

The lessons in this module

What @dataclass writes for you
Defaults that are lists
__post_init__: checking and working things out
frozen dataclasses and replace
Equality, hashing and dataclasses
Ordering dataclasses
Dataclasses to JSON and back
Which shape for your data
Field order, defaults and kw_only
slots: catching the misspelt attribute
Objects you can call
__getattr__: answering for missing attributes
__setattr__: a rule for every assignment
Descriptors: how property works
__init_subclass__: classes that register themselves
__format__: your objects in f-strings
The hash that changed
When the magic is the problem
Capstone: one more field
Module review: the price list editor

Ready when you are

Start this module

Previous: Context managers. Next: functools and itertools. All modules are on the course page.