A data scientist's guide to code reviews
summary
Argues that data-science reviews must shift from style and API design to reproducibility, chosen approach, evaluation metrics, data quirks and result plausibility; provides a GitHub checklist and recommends running the code to reproduce results, splitting large merge requests, and documenting methods and outputs. Names concrete tools: pre-commit hooks with nbconvert for notebooks and formatters, plus linters like black isort flake8, and proposes using GitLab/GitHub code-suggestion and mark-viewed features to make reviews practical.
AI-written summary · about
chapters (11)
- 0:00Start
- 2:46Doing code reviews on data science work
- 4:32What are code reviews for?
- 7:46Different focus
- 11:18What are code reviews for in data science?
- 14:45Code review checklist
- 15:37How to do code reviews?
- 16:35Useful Git functionality: pre-commit hooks
- 18:14Useful GitLab functionalities: Comment field
- 20:31Useful GitLab functionalities: Mark viewed
- 21:04Thank you!
related talks
41mCracking the Code Review from Guesswork to Automated Guidance
18mHow To Review Someone's Code While Not Being A Jerk!
40mCode Review - How to Do It Better?
33mCode review for Open Source
45mThe Science of Code Reviews
47mCode review beyond code style
11mTransforming Code Review into a Learning Tool for QA Teams
31mMay the Code Review Be with You