TL;DR

git worktree rocks.

I was so sure I had already written about this that I went looking for this post. Except that I had not already done that, so I found nothing.

So… git worktree. Much like sliced bread.

The main selling point is: keep multiple working directories out of a single repository, so that we can work on them in parallel.

Many people praise this for being able to quickly work on a critical fix in a branch without having to mess with their current working directory. The stuff that was mainly addressed with git stash up to some time ago, except that now it seems everybody secretly hated that.

For starters, I think that git stash is cool. Maybe outdated by multiple working trees, but still cool.

Second, I’m the main consumer of my stuff. OK, the only consumer of my stuff. So I don’t have this urge to fix stuff in older branches and save the day. But still git worktree is extremely useful.

Consider, for example, Codeberg Pages. Many times we keep sources for a static generator in the main branch, while the published stuff lives in the pages branch. So… why not have both checked out at the same time?

Keeping in mind that we can only checkout a branch only once (i.e. we can’t have to separate working directories for e.g. branch main), we can do as little as:

git worktree add somepath somebranch

As an example, in a Jekyll project hosted in Codeberg Pages we might do the following:

rm -rf _site
git worktree add _site pages

Now every time we regenerate the site locally, the changes get directly inside the working directory aligned with the pages branch. When we’re ready… we can just go into that working directory, commit and push:

cd _site
git add .
git commit -m 'Regenerate site'
git push

Nifty!

Stay safe folks!