Wes McKinney on Improving the Data Stack & Composable Systems

How do you avoid the bottlenecks of data processing systems? Is it possible to build tools that decouple storage and computation? This week on the show, creator of the pandas library Wes McKinney is here to discuss Apache Arrow, composable data systems, and community collaboration.

Wes briefly describes the humble beginnings of the pandas project in 2008 and moving the project to open source in 2011. Since then, he’s been thinking about improvements across the data processing ecosystem.

Wes collaborated with members of the broader data science community to build the in-memory analytics infrastructure of Apache Arrow. Arrow avoids the bottlenecks of repeated data serialization and format conversion. He shares examples of Arrow’s use across the spectrum in tools like Polars and DuckDB.

Wes advocates moving from vertically integrated tools toward composable data systems. We discuss his work on Ibis, a portable dataframe API for data manipulation and exploration in Python. Ibis supports multiple backends by decoupling the API from the execution engine.

This week’s episode is brought to you by Posit Connect.

Course Spotlight: Unleashing the Power of the Console With Rich

Rich is a powerful library for creating text-based user interfaces (TUIs) in Python. It enhances code readability by pretty-printing complex data structures and adds visual appeal with colored text, tables, animations, and more.

Topics:

• 00:00:00 – Introduction
• 00:02:26 – Dealing with limitations in early data science
• 00:04:53 – Making pandas open source
• 00:07:10 – Making changes to an existing platform
• 00:12:34 – Decoupling storage and computation
• 00:23:04 – Sponsor: Posit Connect
• 00:23:54 – Apache Arrow solving multiple issues
• 00:27:40 – DuckDB efficient analytic SQL database
• 00:30:24 – Polars dataframe library
• 00:31:04 – pandas 2.0 adding Arrow
• 00:35:56 – Video Course Spotlight
• 00:37:20 – Apache Software Foundation background
• 00:41:29 – Shifting from developer to organizer and collaborator
• 00:45:56 – Creating a portable query layer with Ibis
• 00:55:34 – Casualties of the language wars
• 00:57:57 – What’s your role at Posit?
• 01:01:23 – What are you excited about in the world of Python?
• 01:04:52 – What do you want to learn next?
• 01:06:21 – How can people follow your work online?
• 01:08:20 – Thanks and goodbye

Show Links:

• Wes McKinney - Personal Website
• Wes McKinney - The Road to Composable Data Systems: Thoughts on the Last 15 Years and the Future
• Wes McKinney - Leveling Up the Data Stack: Thoughts on the Last 15 Years - YouTube
• Apache Hadoop
• Cloudera - The hybrid data company
• Wes McKinney - Apache Arrow and the “10 Things I Hate About pandas”
• Voltron Data - The Leading Designer and Builder of Enterprise Data Systems
• Apache Arrow
• DuckDB - An in-process SQL OLAP database management system
• DuckDB-Wasm - Efficient Analytical SQL in the Browser
• Polars - Dataframes for the new era
• pandas 2.2.0 documentation
• Episode #167: Exploring pandas 2.0 & Targets for Apache Arrow – The Real Python Podcast
• ASF - Welcome to The Apache Software Foundation!
• Ursa Labs Blog
• Ibis - The Portable Python dataframe Library
• Python dataframe interchange protocol
• Hadley Wickham
• Rust Programming Language
• italki - Best language learning app with certificated tutors
• Wes McKinney - LinkedIn
• Wes McKinney (@wesmckinn) - X
• Posit - The Open-Source Data Science Company

Level up your Python skills with our expert-led courses:

• Data Cleaning With pandas and NumPy
• Unleashing the Power of the Console With Rich
• The pandas DataFrame: Working With Data Efficiently

Support the podcast & join our community of Pythonistas