Codebase Research
summary
Runs Error Prone over Google’s monorepo (≈300 million Java LOC) and returns corpus-wide answers in hours so teams can mapreduce refactor at scale. Examples: 180,000 switch statements analyzed, 2.4% show nontrivial fall‑through, ~2–2.3% simulate case null, ~3% place default first; static analyses also detect patterns for API design and can auto-apply fixes across the repo, producing measurable evidence that informs feature tradeoffs.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 2:36Why we are here
- 3:30What is Google 3
- 5:20Design Decisions
- 10:25Infinite Insight
- 16:10Ericone
- 17:25Examples
- 20:55Switch Research
- 24:32Fall Through
related talks
48mJava's Hidden Gems: Tools and Libraries
47mJava's Hidden Gems: Tools and Libraries
27mLessons from a Large JVM Monorepo
38mBeyond Code: Repository Mining with Clojure
21mUnderstanding and Finding Java Decompiler Bugs
20mJava's Hidden Gems: Tools and Libraries
03h 35mJava ON 2024 Day 1
15mUsing AI effectively in large Java codebases
