General best practices

  1. Write good log messages
  • first line: “what this commit does” is <= 80 characters (to be compatible with log analysis tools)
  • blank line
  • multiline: “why this commit is necessary or desired”
  1. Don’t track derived files
  • don’t track: binaries, object files, logs, .pdf, .aux, .bbl, .avi, etc.
  1. Don’t track large files
  • git doesn’t handle them well - makes repo large and harder to work with
  • can use other tools: git annex or LFS (large file storage) module
  1. Don’t alter public history
  • no –amend, rebase, reset on commits that have been pushed to a public server
  • similarly, never push to any non-bare repository
  1. Public repos (including forked) should be bare
  2. Every repo needs 4 files
  1. No merge commits in pull requests
  • makes code review easier
  • keeps project history clean
  1. Use amend/rebase (squash) for small changes when iterating on pull requests
  2. Use topic (e.g. “foo”) or fix (e.g. “fix/bar”) for pull requests
  3. Make code cite-able with a DOI:
  4. Use semantic versioning ( for release branches and tags

And some best practices for public collaboration on git repository services (e.g. Github, Gitlab, etc.)

  1. Protect the master branch (no push)
  2. Fork developer repos, not clone
  3. Use issue tracking
  • Not just for bug tracking
  • reference code, pull requests, developers, assignees, etc.
  • labels for filtering
  • milestones
  1. Use automated services
  • continuous integration (CI): jenkins, travis-ci, gitlab built-in
  • code coverage:
  • doc generation:,
  • code lint (formatting standards)
  1. Treat github as social media
  • curate personal information carefully
  • follow/watch projects
  • potential employers/followers can watch your activity
  • pages:,,
  • chat:,
  • gist