git

Setup

• Login with your username found on your name badge and set the initial password for your account: https://git.aspp.school/user/login
• You'll have to type that password many many times this week: choose wisely!
• We will use the exercise in the repo for the rest of the lecture
• find a partner to do pair-programming for this lecture by finding someone with the same tarot card as you
• before starting, make yourself acquainted with the meanining and the power of the tarot cards: carefully read this booklet

A cautionary quote

My first instinct is to sell all my computers, fake my own death, move to another planet, and reinvent computing from scratch, rather than try to understand Git.

I rarely actually do that, mind you. But the urge is there.

— Lars Wirzenius (Linux kernel developer)

Warm-Up

• how to start a repo from scratch?
  • git init local method
  • on an online forge (GitHub, GitLab, …): git clone
• how to revert mistakes?
  • before commit:
    • git restore <file> [discard changes in the working directory] changes files
    • git restore --staged <file> [unstage changes ➔ opposite of git add <file>, does not modify the working directory]
  • after commit:
    • git revert <commit> [creates a new commit, modifies the working directory]
    • git reset <commit> [only reset the HEAD pointer, does not modify the working directory] rewrites history ➔ can not be used if you have already pushed to some remote
    • git reset --hard <commit> [reset HEAD and modify working directory] rewrites history and changes files ➔ can not be used if you have already pushed to some remote
• how to move the whole working directory to a specific point in history?
  • git checkout <commit> ➔ DETACHED HEAD problem, changes files
  • interaction with branches: git branch <branch_name> + git switch <branch_name>
• how to copy a file from a different branch:
  • git checkout <branch> <file> ➔ the file is staged automatically
  • git restore --source=<branch> <file ➔ the file is not staged
• git gui: building commits along the way interactively (for the mess around type of workflows)
• check out these sketches for a graphical visualization of git commands!

Workflows

Look at the sketches

1. lone scientist local ➔ working alone in the cellar without Internet (local git)
2. lone scientist remote ➔ uploading their software to the Internet in the hope it can be useful for other people (local git + one personal repo on a git-forge)
3. research group ➔ sharing one software project with some other befriended scientist working in a different place (local git + group repo on a git-forge + permissions)
4. fully distributed software development ➔ using the most typical open source software workflows as used by numpy, scipy, sklearn, etc. (like above + we don't trust our contributors, i.e. work strictly with forks)

The Open Source workflow

• remotes: git pull <from_where> <what>, git push <where> <what>, git fetch <from_where> <what>, git merge <another_branch>
• git-forge: forks, branches and PRs: notice the difference between fork and clone!!!
• strategies for keeping your fork up-to-date: your main, origin's and upstream's main, short-lived and long-lived topic branches
• a more thorough and detailed explanation can be found on the SciPy Contributor's Guide. This guide can be adapted to your own needs, see gitwash.
• GitHub and GitLab are just options (git≠GitHub): there are many git-forges out there, both installable on premises (e.g. forgejo) ) or as an online service for free and open source software (e.g. codeberg