Code as Data: How data insights on legacy codebases fill the knowledge gap in modernization projects
summary
Code becomes queriable metadata instead of a mystery: parsing a 4M‑line legacy codebase down to a 22k‑line entry set and 100k tokens, filtering to ~5k tokens and 128 referenced tables exposes forgotten rules and reduces the field perimeter to 36% of fields. The team wires an automated pipeline (built in Clojure using ANTLR grammars) as a 300-line pipeline, exports hyperlinked Excel for SMEs and VS Code deep‑links, and prototypes whole‑system exploration with Pharo/Rasa and 11GB JSON graphs.
AI-written summary · about
chapters (6)
- 0:00Start
- 13:17Hyperlink to Visual Studio Code
- 16:27Outcomes
- 26:44Build the Pipeline
- 28:07Exploratory Navigation
- 35:39Moldable Data Analysis
related talks
38mBeyond Code: Repository Mining with Clojure
47mKnowledge in code is gold — how to extract it and power your team and AI
55mSo, you inherited a large code base...
39mGenAI-assisted reverse engineering for incremental Legacy modernisation
28mI've Fallen Into a Large Codebase and I Can't Get Up!
51mSoftware Archeology and The Code Of Doom
47mTreat Your Code as a Crime Scene
51mWhere to Start?