Open source: Contributing to Pydantic

Improving the benchmark suite for pydantic, a Python library I use daily for data validation.

Pydantic is a popular Python library for data validation using Python type annotations. As a daily user of pydantic in my work, I've had the opportunity to contribute to its development, focusing on improving the benchmark suite used in their continuous integration (CI) pipeline.

Among many of the posible routes of contrbution, I decided to focus on the benchmark suite outlines on Issue #9711.

My Contributions

My main contributions to pydantic have been centered around enhancing the benchmark suite, particularly for schema generation during runtime. Here are some key areas I've worked on:

  1. Improved Benchmark Suite: I've created and improved benchmarks that measure the performance of schema generation, a critical aspect of pydantic's functionality.

  2. CI Integration: My work has helped to integrate these benchmarks into the project's CI pipeline, allowing for continuous performance monitoring.

  3. Performance Insights: The benchmarks I've developed provide valuable insights into pydantic's performance, helping the maintainers identify and address potential bottlenecks.

Notable Issues

Issue 9711: Expand breadth of benchmarks

I added the following benchmarks to contribute to Issue #9711:

  1. PR #10240: Add benchmarks for categories: serialization, validation and schema generations
  2. PR #10271: Add benchmarks for schema generation with custom validators
  3. PR #10290: Add schema generation benchmarks for models with custom serializers
  4. PR #10362: benchmark tagged/untagged unions and stdlib types in schema generation
  5. PR #10568: Add benchmarks for schema generation with pydantic custom types
  6. PR #10602: Add benchmarks for schema generation with recursive models

Impact and Learnings

Contributing to pydantic has been a rewarding experience. It has allowed me to:

  • Gain deeper insights into the internals of a widely-used Python library.
  • Improve my skills in performance optimization and benchmarking.
  • Collaborate with experienced developers in the open-source community
  • Contribute to the improvement of a tool I use daily in my professional work

I kind of knew all of these things before starting, but it's always nice to have concrete examples.

A note on hubris and ego

One things that always surpises me is how quickly your previous work is forgotten.

In some of my PR's, the onwers of the project ruthlessly requested changes to my code. I don't take it personally at all, it's all in good fun.

These experiences have taught me to remain humble and to always be willing to learn and improve.

print("Actually, you always learn more from criticism than you do from praise.")

Thanks pydantic team for your time and consideration!.

Future Work

I plan to continue contributing to pydantic, with a focus on:

  • Further refining the benchmark suite
  • Exploring opportunities for performance improvements in schema generation
  • Assisting in the development of new features related to data validation and serialization

If you're interested in contributing to pydantic or learning more about my work, feel free to check out the pydantic GitHub repository or reach out to me directly.