The GitHub Graph: Open-Source Collaboration Patterns and Neo4j

on YouTube ↗28m

" summary

Uses Neo4j and Cypher to model GitHub as a PR–issue graph, scraping 56 projects and ~90,000 issues/PRs, then manually labeling a final sample of 64 connected clusters and running automated queries to find 1,000+ matches of nine prototypical workflows. Finds recurring, actionable patterns like competing PRs (many PRs fixing one issue with one merged), duplicate issue hubs (many later duplicates tied to a central report), and diversion PRs (one PR closing many issues). Provides reusable Cypher, a Neo4j Python pipeline, and an interactive visual explorer for maintainers to reproduce and inspect results.

AI-written summary · about

" chapters (13)

1. 0:00Introduction to research
2. 1:13GitHub collaboration basics
3. 3:00Link types in GitHub
4. 7:32Graph analysis with Neo4j
5. 10:41Workflow type definitions
6. 13:29Competing PR workflow
7. 17:51Duplicate issue workflow
8. 20:19Overdelivered PR workflow
9. 22:04Workflow type frequency
10. 23:29Developer validation insights
11. 25:16Practical applications
12. 26:40Automating analysis
13. 27:22Q&A

" related talks

• 32mCode More, Draw Less, and Debug Just a Little!Yash Saboo · PyCon Italia 2024
• 40mDiscovering Emerging Open Source Communities Through...Hema Veeradhi, Oindrilla Chatterjee · OSPOCon @ Open Source Summit North America 2023
• 47mPolicing Open-Source Projects at ScaleThomas Neidhart · Open Source Summit Europe 2024
• 14mGitHub's Data Story: Open Source at the Home for Open SourceAshley Wolf, Natalie D'Adamio · Open Source Summit North America 2022
• 42mEnhancing Open Source Collaboration: From Incentive Programs to Data-Driven Contribu...Xiaoya Xia · Open Source Summit Japan + AI_dev 2024
• 22mGetting Started with Open Source Project InsightsBrian Douglas · All Things Open 2022
• 01h 06mHow GitHub delivers GitHub using GitHubApril Yoho · NDC London 2024
• 01h 03mWhat can we learn from 750 billion GitHub events and 42 TB of codeFelipe Hoffa · NDC Sydney 2017