Branches — parallel timelines

A branch is a file in .git/refs/heads/<name> containing a 40-char SHA — the commit it points to. Creating a branch is O(1): just write a ref. Compare to centralized VCS where a branch is a server-side directory copy. This is why Git branches are 'cheap'.

HEAD: a symbolic ref at .git/HEAD storing ref: refs/heads/<current-branch>. Detached HEAD = HEAD directly contains a SHA, not a ref. Happens on git checkout <sha>. Commits made in detached HEAD are reachable only via reflog until you create a branch from them.

switch vs checkout: git switch (Git 2.23+) is the modern, focused command for branch operations. git checkout is the legacy umbrella that also handles file restoration (now git restore). Prefer switch/restore in tutorials and scripts; both still work.

Tracking branches: a local branch can track a remote branch (git branch --set-upstream-to=origin/main main or -u shorthand on push). Once tracking, git push and git pull know where to go without explicit args. git status reports 'ahead by N, behind by M'.

Merged vs unmerged delete: git branch -d refuses to delete a branch whose tip isn't reachable from another branch — saving you from losing commits. -D (capital) forces deletion. Use git reflog to recover commits from a force-deleted branch.

Default branch: main (modern), master (legacy). Change locally with git branch -m old new then push the rename and update upstream defaults on GitHub. Renaming the default branch on a public repo breaks deep links and CI configs that hardcode the name.