Idealized Commit Logs: Code Simplification via Program Slicing
summary
Claims a tool that produces an Idealized Commit Log by computing per-test program slicing from code‑coverage data, using AST transforms and type analysis to emit simplified code that still compiles and passes the originating tests. A Go prototype reconstructs tiny, human‑readable functions (examples: URL parser, scanner) by running coverage per test, differencing slices and cleaning dead branches. It works but requires tests, has limited empirical validation, and existing slicers mainly target Java/C so language/runtime integration is the main barrier.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:29Code reading neglect
- 5:05Software complexity barriers
- 8:42Code poor abstraction
- 10:09Simplifying code with slicing
- 14:45Program slicing history
- 17:15Program dependence graphs
- 20:54Slicing applications
- 24:29Slicing comprehension impact
- 27:00Slicing tools rarity
- 28:45Idealized commit log
- 31:55Live slicing demo
- 36:54Recap
- 37:59Q&A
related talks
17mConceptualizing Large Software Systems
23mAnalysis and Transformation Tools for Go Codebase Modernization
01h 00mBuild Your Own Tools to Refactor Go Safely
34m"Escaping The Tar Pit"
59mUnleashing the Go Toolchain
16mTools for working with Go Code
50mProgram Analysis
27mParsing, Compiling, and Static Metaprogramming