How we upstreamed our internal goals to JupyterLab 4
summary
Two Sigma describes concrete upstream contributions that land in JupyterLab 4: benchmarks show opening large notebooks falls from about four seconds to around one after virtual rendering and a CodeMirror 6 upgrade; protocol alignment removes server deserialize/serialize hops; Lumino2 brings component speedups; RTC gains include collaborator lists per-cell undo and a collaborator panel. Work happens via a partnership with QuantStack, runs against an internal ~500 distinct user deployment, and is built with about 0.5 FTE devoted in 2022.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:27Speaker background
- 2:28Reliance on JupyterLab
- 6:23QuantStack partnership
- 7:48Performance improvements
- 13:55CodeMirror 6 upgrade
- 15:19Faster outputs protocol
- 17:06Lumino 2 upgrade
- 18:10Real-time collaboration
- 21:17Two Sigma and QuantStack
- 26:52Staying up to date
- 29:01Extension management
- 30:47Key takeaways
- 33:11Volunteer at PyData
- 34:12Hiring at Two Sigma
- 34:34Q&A
related talks
29mJupyterLab+Real Time Collaboration
39mThe JupyterLab Extension Ecosystem- Trends from PyPI +GitHub
40mInstructLab: Applying Open Source Methods to Building and Trai...
20mKeynote: From Prototype to Innersource to Open Source: Lessons...
01h 39mOSS Keynotes - Day 2 Morning Sessions
30mApplying Open Source Methods to Building and Training Large Langu...
01h 36mLFX Mentorship Showcase
49mState of the Union