Taming A Monster With Git Driven Data
summary
Extracts raw git logs from a repo (demo uses TensorFlow) and generates a CSV of the top 50 files over the last six months, then visualizes change frequency (long tail: a small surface of files get most edits) and a simple language‑agnostic complexity proxy (lines/indentation) to identify refactor hotspots. Uses a codemap wrapper plus RawGraphs for charts, computes change coupling as files co-changing ≥X commits and ≥Y% of the time, and builds owner/knowledge maps; the method surfaces surprising hotspots and works on real repos.
AI-written summary · about
chapters (8)
- 0:00Introducing legacy systems
- 3:10Git as historical data
- 9:36Analyzing change frequency
- 16:23Identifying code hotspots
- 18:51Measuring change coupling
- 23:28Practical refactoring insights
- 26:37Ethics and empathy in analysis
- 28:45Further reading and Q&A
related talks
47mTreat Your Code as a Crime Scene
38mBeyond Code: Repository Mining with Clojure
49mTreat Your Code as a Crime Scene
31mA Picture is Worth a Thousand Words: Bring Data Visualization in Backstage
34mGit from the Ground Up
25mMachine Learning for source code analysis
31mFly, You Tools!
04mThink outside the box