Can Haskell Become a Great Language for Data Science?
summary
Argues Haskell can span quick exploration to reliable pipelines by combining a notebook UI with typed DataFrame semantics, demos loading Iris (150 rows), interactive lasso selection, feature engineering and model export, and a markdown-first reactive notebook prototype called Sabella. Shows Immediate Haskell progressive typing, supports lazy DataFrames with a DataFusion backend for out-of-memory work (1 billion rows on a 3 GB machine in ~10 minutes), uses a GHCi subprocess model, and notes ongoing Apache Arrow / Parquet integration.
AI-written summary · about
chapters (16)
- 0:00Introduction And Motivation
- 5:14Exploring Data With Interactive Notebooks
- 8:49Why DataFrames Matter
- 12:33Formalizing DataFrames And Types
- 15:43Type-Level DataFrames In Haskell
- 18:33Ergonomic Data Processing APIs
- 20:43Performance And Apache Arrow
- 23:05Building A Haskell Data Ecosystem
- 24:13Literate Programming And Notebooks
- 25:50Reactive Notebooks Explained
- 28:25Notebook Architecture And Package Management
- 31:22Notebook Demonstration And Reporting
- 38:53Exporting Notebooks To Haskell Modules
- 42:43DataFrames, Grouping And Query Execution
- 47:58The Future Of Data Haskell
- 48:43Q&A And Ecosystem Discussion
related talks
47mScala: The Unpredicted Lingua Franca for Data Science
46mPainless software development with Haskell
01h 04mHaskell - A Peek Inside the Ivory Tower
01h 22mHaskell for Everyday Programmers
26mDoing data science with Clojure: the ugly, the sad, the joyful
01h 02mHaskell for Java Programmers
47mIndustrial Strength Laziness: What's Next?
40mWhat Haskell taught us when we were not looking