Unit Testing Data with Marbles
summary
Argues that unit testing is a practical framework for validating data assumptions and shows a working Python extension that does it: Two Sigma's Marbles installs with pip and plugs into unittest, providing semantically rich assertions, captured local variables, author notes (format strings), and assertion logging so failures carry the context needed to act. Uses concrete checks (e.g. longest confirmed lifespan 122 years 163 days) and reports that Marbles produces human‑readable failure messages that help diagnose and triage data breaks.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:48Data assumptions pitfalls
- 5:21Unit testing needs data
- 10:21Python data validation tests
- 12:30Unit test for data assumptions
- 15:23Data unit testing challenges
- 18:58Recovering test failure context
- 20:27Marbles for failure messages
- 23:44Customizing Marbles for data
- 27:59Test failures as documentation
- 29:04Marbles open source next steps
- 31:43Test severity and scheduling
- 33:53Temporal data logging
- 35:17Q&A fairness and bias
- 37:11Q&A scaling failure management
related talks
44mUnit Testing Data with Marbles
01h 30mAdvanced Software Testing for Data Scientists
39mUnit Testing for Data Scientists
44mJust Another Marble Monday – Unit Testing NGRX/RxJS w/Marbles
35mTesting: Choose Your Weapons
44mThey are not unit tests: a survey of unit-testing anti-patterns
55mThe Science of Unit Tests
33mUnit Testing with Pytest