Talks - Samuel Colvin: How Pydantic V2 leverages Rust's Superpowers

preview_player
Показать описание
Pydantic is a data validation library for Python that has seen massive adoption over the last few years - it is estimated that Pydantic is now used by about 10% of professional web developers!

Over the last year I've been working full time to rebuild Pydantic from the ground up, using Rust for virtually all the validation and serialization logic. Pydantic V2, with these changes included, has recently been released.

In this talk I will give a brief introduction to Pydantic and the new features in Pydantic V2 before diving into how the use of Rust has allowed us to completely change the architecture of Pydantic to make it easier to extend and maintain while also improving performance significantly.

The majority of the talk will be devoted to using examples from the pydantic V2 code base to demonstrate the advantages (and disadvantages) of writing libraries like Pydantic in Rust. I'll cover the real life trade-offs and design decisions you might face while implementing logic in Rust rather than Python.

This talk should be interesting to any Python developer who's interested in combining Python and Rust - no knowledge of Rust or Pydantic is required. However if you'd like to get some context or learn more about the topics discussed, here are some useful resources:

Pydantic V2 Plan - blog post about the plan for Pydantic V2
pydantic-core - the python package that provides Rust logic in pydantic
PyO3 docs - the amazing library that allows Rust to be embedded in Python
Build your Python Extensions with Rust! by Paul Ganssle - good intro to building Python extensions in Rust
Рекомендации по теме
Комментарии
Автор

I'm enjoying using Pydantic for API schemas usually in FastApi. It's perfect. I am considering expanding its use into more general data structures.

demolazer
Автор

Turn audio all the way left or right and it works.

JOHNSMITH-verq
Автор

Instead of AnalyzedType, what about something like "to_model"? Then "Response= pydantic.to_model(Result | Error)" feels static, and (more or less) lives up to the name on the tin when "Result" and "Error" are dataclasses, or some other kind of type/object.

nicholaswillhite
Автор

Just set the audio to mono in YouTube edit please.

TreeLuvBurdpu