I have a modest proposal on repository organization. There are 2 primary archetypes, multiple repositories and the monorepo pattern. In multiple repositories, you have 1 project per repository, binaries requiring internal libraries, and use tools to ensure that you have all the necessary repositories downloaded and up to date (easy with most modern package managers). In a monorepo, all projects are stored directly in the monorepo, as part of a single codebase or project, and all internal dependencies are automatically managed.

Each archetype has its own challenges (multi-repository PRs, space consideration, compartmentalization, CI), but I’d like to propose a third solution that solves none of these issues and is really dumb

Hydra Repositories

In git, you can create a branch in a repository with no connection to the existing history: an orphan branch.

Per this SO answer,

git checkout --orphan newbranch
git rm -rf .

And boom, you have a new branch with no connection to the old history of the repository and no contents. From here you can build an entirely new project isolated from any other branches.

In the hydra Repository, you maintain separate orphan branches for each of your projects and libraries. This has all the isolation and lack of easy code sharing of multiple repos, but in the same repo. Like a monorepo, you can clone every single project you care about with a single git clone command without an easy way of excluding some parts of the project. Google’s famous monorepo includes specialized tooling for dealing with a single repository that contains all Google code, and you’ll be building all that tooling on top of git!. CI is even harder than in a monorepo, and even less tooling exists for managing this structure.

Although most dependency management tools is prepared to deal with dependencies specified by git repo URL and SHA (in order to handle multiple private repositories), are they read to deal with multiple dependencies that all come from exactly the same repository? Maybe some of them make some assumption that renders them incapable of dealing with hydra repositories, and therefor that language or environment incapable of running in a hydra repository.

By using git worktree, you can “mount” these orphan branches in the working directories of other branches like so:

git worktree add repository/ <SHA>

This directory will act like another working directory for the overall git repository, allowing you to work on multiple branches at the same time.

You can also theoretically merge these repositories easily by just merging the branches directly, so if you make the horrible and inexplicable decision, you can resolve the issue by just merging the orphan branches into the same branch.

Luckily, this article is not born of personal experience.