DevX: History of Code Search
summary
Argues that developers spend about 10x more time reading than writing code and therefore need code search to grok big code bases; traces concrete tools that worked: cscope for grep-based queries, web cross-referencing via LXR, Google’s distributed gsearch scaled by a trigram index, and Grok/kythe adding compiler-accurate find references and jump-to-definition. Shows practical wins: mass Python2→3 pattern fixes, discovering a PHP API break across versions, faster postmortems via shared search links, and adoption at Stripe, Debian, CERN and many firms.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 0:22Thesis
- 0:53What do developers actually do
- 1:32What is Grok
- 3:13Cscope
- 5:04Linux Kernel
- 6:07OS Wars
- 6:52Google Code Search
- 8:42The exploding world of code
- 10:35New code search engines
- 11:07DevX interface
- 11:46Why are interfaces so similar
- 12:09Learning by example
- 12:56Find references
- 13:22Identify patterns
- 13:52Ensure backwards compatibility
- 14:38Resolving production issues
- 15:50Patrick McKenzie
- 16:30Value of Code Search
- 17:14Debian Linux
- 17:54CERN
- 18:04Recap
- 19:57Outro
related talks
10mTo search or not to search: it depends on the question
08mSupercharging Developer Productivity With Advanced Code Search
38mCode reading
49mReading Code
32mPractical code archaeology
37mBeyond Grep: Semantic Code Search for Safer Change at Scale
04mGet Productive on Unfamiliar Source Code
21m10 Code Search Tricks for Open Source Contributors and Maintainers