Thread by @rmcomplexity

Do you know what Python descriptors are?

Descriptor is a protocol in python that enables data to have a saying about what happens on lookup, storage and deletion.

A great use case for descriptors are data validation.

#100DaysOfCode #freeCodeCamp

🧵👇

When a class defines `__get__`, `__set__` or `__delete__` it becomes a descriptor and is called with the `.` (dot) operator.

By implementing validators as descriptors we push everything validation-related outside the class that holds data.
👇

Let's define two validators as descriptors.

They both store the actual value assigned in a private variable and run a validation function when a value is assigned.

Let's go over the special methods implemented 👇

A descriptor that ONLY implements `__get__` is called a non-data descriptor.

The method is called by the `.` dot operator when a value is accessed.

For example: logging.info(item.sku)

👇

A descriptor that implements `__get__` AND `__set__` is called a data descriptor.

The `__set__` method is called by the `.` dot operator when a value is assigned.

For example: item.sku = 'item-sku'

👇

A descriptor could also implement `__delete__`.

This method gets called when an attribute is deleted.

In most cases we don't need this. For simplicity I'm going to leave it out of this example.

👇

A descriptor can optionally implement `__set_name__`. This is not part of the descriptor protocol.

The method is called when the parent class is created.

Useful to store the attribute name the descriptor is assigned to.

Descriptors only work when assigned to class variables.

A non-data descriptor can be overridden by an instance variable.

A data descriptor will NOT be overridden by instance variables.
👇

We can use the previously defined descriptors to validate the data in a dataclass like this.

👇

A dataclass using descriptors can be instantiated as normal.

The main downside of using descriptors together with dataclasses is that we have to use typing.cast to avoid mypy errors.

👇

If we initialize an instance of a dataclass with validator descriptors in its fields with invalid values we get an `AttributeError`

👇

Using dataclasses together with descriptors means apart from validation we can also get serialization out of the box.

👇

You can play around with the code in this thread using this notebook:
mybinder.org/v2/gh/rmcomple…

Read more on descriptors: docs.python.org/3/howto/descri…