Let's focus on IDE Indexing
summary
Uber Focus implements on‑demand lazy indexing for IntelliJ/Android Studio in a Buck monorepo, raising IDE heap to 4GB and setting caches.indexer.thread count=8. A Buck wrapper auto‑installs the plugin and VM options, cutting initial indexed files from 250,000 to ~1,400 and lowering first‑file indexing to ~1.5s from multi‑minute waits. Find/rename operations prompt a module load on demand, crash reports ship to Jira/Fabricator, and rollouts hit ~90% adoption.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:06IDE performance challenges
- 7:46Root causes of slow indexing
- 12:59Uber Focus on-demand indexing
- 17:59Limitations and workarounds
- 22:55Plugin rollout strategy
- 25:00Crash reporting and monitoring
- 27:27Future plans and open source
- 30:00Q&A
related talks
01h 00mIndexing, or How We Made Indexes Shared and Fast
46mHow we’re improving performance of IntelliJ IDEA Kotlin plugin
54mHow IntelliJ IDEA Performance Improved: Past, Present, and Future
42mLook Ma – No Indexing!
23mKotlin 1.4 in IntelliJ-based IDEs: Boosting Quality and Performance
31mTaming the Beast - Better Bazel in IntelliJ
32mSupercharging IDE and Sync Performance
35mIntelliJ IDEA Tips & Tricks