Beyond Grep: Semantic Code Search for Safer Change at Scale
summary
Searches moving from grep to IDEs to Tree-sitter and trigram indexes leave gaps at portfolio scale, so Moderne builds a Lossless Semantic Tree that preserves full type info, whitespace and compiler context across repos and a Trigrep index that layers trigram speed over LST semantics. On demos it narrows a 200-repo scan down to ~50 relevant repos, supports structural Comby/Tree-sitter queries and runs a find types recipe and a Javax→Jakarta transform only on matches, avoiding agents reading whole files.
AI-written summary · about
chapters (10)
- 0:00Introduction and motivation
- 3:34Evolution of code search tools
- 9:53Abstract syntax trees and structural search
- 13:29Triagram indexing for large-scale search
- 17:17Lossless semantic trees explained
- 20:07Demo of semantic search with LSTs
- 23:18Introducing Tripp and its relation to LSTs
- 28:15Demo of Tripp from the CLI
- 34:22Agents and code search challenges
- 35:49Summary and future of code search
related talks
17mSyntactic Code Search with Sequence-to-Tree Matching: Supporting Syntactic Search with(…)
08mSupercharging Developer Productivity With Advanced Code Search
20mDevX: History of Code Search
01h 22mGlean: C++ Code Indexing at Meta
22mAutomate your inner scout
39mTree-sitter - a new parsing system for programming tools
36mCode as Data: How data insights on legacy codebases fill the knowledge gap in modernization projects
21m10 Code Search Tricks for Open Source Contributors and Maintainers