Lucian Ghinda
Random Thoughts about Building Software

Random Thoughts about Building Software

A case for using git worktree

A case for using git worktree

Lucian Ghinda's photo
Lucian Ghinda
ยทAug 3, 2022ยท

4 min read

Subscribe to my newsletter and never miss my upcoming articles

Context

Let's say I am working on a project named "short-ruby" ๐Ÿ˜‰

That project is currently located somewhere like

/home/lucian/projects/short-ruby

Let's say that I am currently working on a complex feature named "automate newsletter".

This feature has multiple branches:

  • twitter-integration branched from main
  • substack-integration branched from main
  • collect-tweets branched from automate-n/twitter-integration

To test all this stuff, I am also creating a temp merge branch all where I want to test if all work well together before deciding to integrate them to main.

For simplicity

Current task

The first 2 branches are implemented, and I am currently working on collect-tweets.

Now imagine that while working on this branch, I discover that there is something that needs to be changed, and it is related to integration with Twitter, like I need access to also moments, not only lists.

This change should be part of twitter-integration branch.

Git solution without git worktree

What are my options now?

Simple solution

The simplest one is to git stash my current changes from collect-tweets and then git checkout the twitter-integration branch do the changes there, commit them and then come back to collect-tweets and git rebase.

It might look something like this:

# work on collect-tweets
(short-ruby)$ git stash
(short-ruby)$ git checkout twitter-integration

#... do the work

(short-ruby)$ git commit -m "Added support for querying moments"
(short-ruby)$git checkout collect-tweets
(short-ruby)$ git rebase twitter-integration
(short-ruby)$ git stash pop

# resume previous work on collect-tweets

Elegant but more complex solution

Of course, another solution would be to commit the changes in the branch collect-tweets and then git cherry-pick them to the correct branch.

This is an elegant solution, but you have to be careful and remember to make sure your commit was atomic and did not contain anything from the current branch. Of course, you can also fix that with git rebase -i, but still, that is also a solution that works when you always do logical atomic commits.

More changes

Now imagine that you resumed work on "collect-tweets" and then the next day, a bug needs to be fixed on main and the bug means changing a method in a class that was refactored in twitter-integration

You will need again to do a series of stash/checkout/rebase that now will also mean multiple git checkout and rebases as you will want to propagate the changes from main to twitter-integration to collect-tweets.

Probably cannot do cherry picking while you are in twitter-integration or collect-tweets because you refactored that code already so cherry-picking will fail.

Here I think comes in handy git worktree

Git worktree

Git worktree is a simple command that will map a local folder to a branch.

Imagine I said I have the project in /home/lucian/projects/short-ruby

Here is what I can do before going into all branches:

# Rename folder short-ruby to main and create a new folder short-ruby and put main inside
(short-ruby)$ cd ... ; mv short-ruby main ;  mkdir short-ruby ; cd short-ruby
(short-ruby)$ mv ../main ./ ; cd main
(short-ruby/main)$ git checkout main
(short-ruby/main)$ git worktree add ../twitter-integration witter-integration
(short-ruby/main)$ git worktree add ../substack-integration substack-integration
(short-ruby/main)$ git worktree add ../collect-tweets collect-tweets
(short-ruby/main)$ git worktree add ../all all

Now if I execute git worktree list it will show an output like this:

/home/lucian/projects/short-ruby/main                   bff36930 [main]
/home/lucian/projects/short-ruby/twitter-integration    f840805c [twitter-integration]
/home/lucian/projects/short-ruby/substack-integration   8286fd79 [substack-integration]
/home/lucian/projects/short-ruby/collect-tweets         52d9bf8c [collect-tweets]
/home/lucian/projects/short-ruby/all                    7cc380a4 [all]

What happen is that it created those folders and inside each one it did a checkout with that specific branch.

Now my flow will look like this: I will cd short-ruby/collect-tweets and work in that folder everything related to the branch collect-tweets.

If I need to do a change on twitter-integration branch:

cd twitter-integration
# do the changes
git commit -m "Added support for querying moments"

cd collect-tweets
git rebase twitter-integration
# continue work

Do I need to do some changes on main?

Simple:

cd main
# do the changes
git commit -m "Bug is fixed"

cd twitter-integration
git rebase main

cd collect-tweets
git rebase twitter-integration

Of course, I can automate rebase on all branches (probably) with something like:

# rebase-automate-newsletter.sh
cd ~/projects/short-ruby/main ; git pull

cd ~/projects/short-ruby/twitter-integration; git pull ; git rebase main
cd ~/projects/short-ruby/substack-integration; git pull ; git rebase main

cd ~/projects/short-ruby/collect-tweets; git pull ; rebase twitter-integration
ย 
Share this