Project Immortality: Using GitHub To Make Your Work Live Forever
summary
Shows how to turn R scripts into a minimal package (DESCRIPTION + wrapped functions) so dependencies and a license are reproducible, use a ~20-line GitHub Actions workflow to run nightly model predictions, and store large binary outputs in GitHub Releases instead of committing RDS/parquet files. Demonstrates pitfalls: a 60 MB RDS can bloat git history to ~6 GB, GitHub file limits (~75 MB file / 2 GB repo) break workflows, and piggyback’s PB_release_create / PB_upload / PB_download automate releases.
AI-written summary · about
chapters (7)
- 0:00Project Immortality intro
- 2:25Why projects die
- 4:34Who this talk is for
- 6:43Reproducible with R packages
- 9:35Automating predictions
- 11:01Storing data efficiently
- 16:26Summary
related talks
17mProject Immortality: Using GitHub To Make Your Work Live Forever
53mGit, GitHub, and gaming
01h 06mHow GitHub delivers GitHub using GitHub
27mTaking code to next level with Open Source (Git & Github)
43mTop Ten Worst Repositories to host on GitHub
53mThe Power of Modern Version Control
27mBuilding a Secure Open Source Project on GitHub
28mBreaking out of the cage: move complex development to GitHub