At 0.18, Modin is now an adult, and it’s feeling a strong sense of civic responsibility — a responsibility to vote, do community service, give back.
So after much introspection, Modin has decided to cast its vote for… the Pandas API! Concise, flexible, convenient — the Pandas API has a lot going for it. We hope it gets elected.
Modin has also done some self-reflection, and in conjunction with becoming 0.18, has made some changes to the way it works. Its 0.18 changes include:
- An exciting new feature: Support for an MPI backend through Unidist!
- Message Passing Interface, or MPI, is “used to send messages from one process (computer, workstation etc.) to another” and serves as an additional — very fast — backend. Like Ray, it uses a task-based parallel approach.
- You can get started with Modin on MPI/Unidist here.
- A major performance improvement to the shuffling mechanism, which makes sort much faster, and unlocks improvements to other functions, like groupby!
- Keep an eye out for a separate post about these massive shuffling / sort improvements.
- An upgrade to pandas 1.5.2
- Stability + bug fixes (30 of them since the 0.17 release), and performance + testing enhancements, and documentation improvements!
Modin is 4.5 years in the making, has been downloaded 4.8 million times, and has 8K GitHub Stars.
For more details, check out the release notes here.
Thank you to the 16 Modin developers who shepherded Modin through its 0.17 year.
To stay up-to-date on all things Pandas / Python / Data Science, and to watch Modin enter its final teenage year (0.19!), follow Ponder on Twitter and LinkedIn, and join the Modin Slack community!
