Incremental, zero-config Code Nav using stack graphs
summary
Claims an incremental, zero‑config code navigation system that runs at GitHub scale by generating per‑file stack graphs, extracting and storing partial paths at index time and stitching them at query time. Uses tree-sitter grammars and a graph DSL to support dynamic and static languages without repo config. Prototype indexes most commits in seconds (median ~5 s) and answers queries in ~50 ms median with p99 ≈ 100 ms and is open source.
AI-written summary · about
chapters (14)
- 0:00Start
- 1:14Code Navigation
- 3:59Why is this hard?
- 10:19Zero configuration
- 10:59SCALE
- 12:03When do we do the work?
- 14:17Incremental processing
- 16:24What would incremental results look like?
- 18:16The more complex example
- 22:19Stack graphs
- 33:53Concatenating partial paths
- 36:34Are we done?
- 38:16tree-sitter
- 43:05The really hard ones..
related talks
48mInstant Code Navigation for Any Commit ☠️
18mUnderstand Your Codebase to Innovate Faster
48mInstant code navigation for any commit ☠️
20mChronicler: Interactive Exploration of Source Code History
08mSupercharging Developer Productivity With Advanced Code Search
24mOnline Name-Based Navigation for Software Meta-languages
04mGet Productive on Unfamiliar Source Code
33mWRITING SCALA OUTSIDE IF INTELLIJ IDEA
