Is there is a way to replicate this git worktrees structure?

[ahmedelgabri]

I have been using a very specific structure to manage my git worktrees based on --bare clones, which looks like this

my-project/
├── .bare/            # all git data lives here
├── .git              # just a pointer file
├── main/             # worktree: main branch
└── feature/          # worktree: feature branch

I wonder if I can replicate this with jj, I tried jj workspace add but it has the same problems I had with git worktrees.

I also tried to see if jj will work with that structure and it didn't and filed with

Internal error: The repository appears broken or inaccessible

The only way I managed to get jj is working is jj git init --git-repo=../.bare inside each worktree. Which is not what I want.

I'm not married to this workflow, maybe I'm not thinking "the jj way", maybe there is a better way to do this with jj?

[joyously]

It sort of looks like you want what Breezy offers by default: each branch is a folder and the repo data is shared at the top level.
However, if you init with Breezy, Git doesn't know about it. So you have to init with Git and then Breezy can manipulate the repo (even without init), but it's in Git's format...but that's what jj needs also, with another init.

I'm not sure what you expect Git or jj to see with this layout, or what Breezy does to interface with Git in this situation (I don't use Git much).
I just now tried going to a repo that I started with Git, and have been using jj and Breezy to commit. I told Breezy to branch into a sibling folder, and it says "branched 77 revisions". Looking at the new folder, it has a .git folder in it. Isn't that what you would get with a clone (or even a push)?
I think there are still open issues for jj git push to a local folder.

[ahmedelgabri]

I'm not sure what you expect Git or jj to see with this layout, or what Breezy does to interface with Git in this situation (I don't use Git much).

I'm looking for two things:

• The tool manages these clones/workspaces, that's what currently git does, git worktree list|add|remove all work fine with this setup.
• I want to have an encapsulated project folder, not folders scattered across multiple places.

Looking at the new folder, it has a .git folder in it. Isn't that what you would get with a clone (or even a push)?

In regular clones, yes, you will have .git folder per clone. With this setup, you git a file .git pointing to the .bare folder, which contains all the .git data, for example in a main worktree, the content will look like this.

gitdir: ../.bare/worktrees/main

But in general, I'm open to changing my workflow completely. At the end, what I'm really looking for is an easy way to manage and work on multiple isolated copies of a codebase.

[jennings]

There is no concept of a "bare" repository, but you can jj new root() to get a new empty working copy. That's essentially a bare repository.

$ jj git clone git@github.com:user/repo
$ cd repo
$ jj new 'root()'
$ ls -a
.jj

Then you can ignore the whole directory and add workspaces:

$ echo '*' > .gitignore
$ jj workspace add feature-a -r main
$ ls -a
.jj   feature-a

[ahmedelgabri]

That's something I can experiment with, thanks!