Treat Your Code as a Crime Scene
summary
Argues that version‑control data plus behavioral analysis (the CodeScene tool) reveal where to spend effort: hotspots are typically 1–3% of code but drive most work, e.g. a React profile of ~400k LOC and a .NET CLR profile of 8.5M LOC; hotspots can hide 14k LOC files, 500‑line functions, a function with cyclomatic complexity 112 and 42 commits/year. Uses a hotspot x‑ray to find function‑level hotspots and change coupling (or shin coupling) to expose costly cross‑package edits; shows actionable prioritization and high truck factors (React ~2, Vue ~1, 70% knowledge loss).
AI-written summary · about
chapters (14)
- 0:00Treating code as crime scene
- 1:19Time spent understanding code
- 4:28Forensic psychology meets dev
- 9:54Behavioral code analysis
- 13:45Analyzing .NET CLR codebases
- 16:29Hotspot X-ray for refactoring
- 20:44Marauders vs commuters
- 23:16Change coupling explained
- 26:51Change coupling patterns
- 29:59Social side of code risks
- 33:10Business impact of code quality
- 35:17Truck factor risks
- 37:45Truck factor case studies
- 38:27Summary
related talks
49mTreat Your Code as a Crime Scene
47mTreat Your Code as a Crime Scene
56mCode & Homicide: What Software Devs (..) from Offender Profiling
39m7 Secrets of Maintainable Codebases
31mMaking sense of frontend code with forensic techniques
39mArch. App: Prioritizing Technical Debt as if Time and Money Matters
50mGuide Refactorings With Behavioral Code Analysis
32mPrioritizing Technical Debt as if Time and Money Matters