Doing Data Science with Clojure
summary
Argues Clojure wins for exploratory data work by manipulating nested structures instead of forcing 2D dataframes, shows a small library that composes with core collections, and demonstrates a pipeline using Kafka ingestion, precomputed materialized views and REPL-server + notebook integration that keeps many analyses runnable on a laptop. Emphasizes spec for runtime validation and test generation that catches numerical instabilities. Sets a concrete productivity target — answers under 2 minutes (small queries) vs ~20–30 minutes for ad hoc tasks — and reports the approach scales in production.
AI-written summary · about
chapters (21)
- 0:00Intro
- 0:33Data Science vs Programming
- 1:59Paradigms
- 4:09Data Frame
- 6:17Time Spent
- 7:51Perlis
- 8:15Speed of answers
- 11:01Design decisions
- 16:31Iterating
- 17:12Decomposition
- 21:16Clojure Lisp
- 21:30Clojure Spec
- 23:17Test Generation
- 27:19R Instance
- 30:34Notebooks
- 35:45Notebooks as tools
- 36:18The most nagging problem
- 36:45Machine Learning Visualization
- 36:55Mini Compilers
- 38:44Conclusion
- 39:14Philosophical Points
related talks
26mDoing data science with Clojure: the ugly, the sad, the joyful
24mThe Return of Clojure Data Science
27mData and Clojure: Using Clojure for more than software development
35m"Clojure for Data Science in the Real World"
18mNavigating Clojure: A Fresh Look From a Data Analysis Background
38mScientific Clojure, a bird's eye view
34mClojure by neophyte
44mData Science in Clojure