Identifying bad design in codebases, objectively…
summary
A CLI + web app called Arc Stats analyzes Java repos with 42 metrics and finds concrete problems: a file in bytecode generation with max indentation 22, an overly complex DefaultRestChannel creating a REST↔HTTP cycle, an action package involved in ~4,000 cycles and Elasticsearch listing ~4,800 short cycles overall. Visualizations include a distance to main sequence plot and a component Walker to trace cycles. Outputs combine complexity, coupling, centrality and git-change data to pinpoint refactor targets with surprisingly good accuracy.
AI-written summary · about
chapters (11)
- 0:00Introduction and personal story
- 4:37Why codebases decay over time
- 8:32Measuring code complexity
- 11:44Demo of Arc Stats tool
- 16:02Understanding coupling and stability
- 23:53The dangers of component cycles
- 26:22Demo of cycle detection
- 32:02Graph theory in code analysis
- 36:22Developer activity metrics
- 41:21Future of Arc Stats project
- 42:45Q&A session
related talks
51mSoftware Archeology and The Code Of Doom
31mMaking sense of frontend code with forensic techniques
47mTreat Your Code as a Crime Scene
49mTreat Your Code as a Crime Scene
39mArch. App: Prioritizing Technical Debt as if Time and Money Matters
42marchitect's Guide to Managing a Code Base
20mFind Out If Your Design System Is Better Than Nothing
40mTreat Your Code as a Crime Scene