OK but what is “Git Patch Stacks”? It seems like I’m missing something fundamental that was never actually explained in this post.

Yes, sorry. The post doesn’t provide the full context as it is just a post covering the goals of Git Patch Stack.

Git Patch Stack is a tool that works with Git to facilitate a “Patch Stack” based workflow in a manner that also maps seemlessly to the Feature Branch/Pull Request model of GitHub, Bitbucket, GitLab, etc. that the majority of the development communities use. https://git-ps.sh is the official website for it. https://github.com/uptech/git-ps-rs is the repository.

EDIT: Oh hey I found it after reading a different blog post on that site: https://github.com/uptech/git-ps-rs

The post is also super rambling in the middle section.

I have to assume that this is something similar to Mercurial’s MQ plugin or Stacked Git (stgit), which allow you to manage revisions as a stack of diffs.

I personally enjoy using stgit, and it fits just fine into the world of feature branches and PRs.

Thanks for the feedback abound rambling. In relation to stgit Git Patch Stack is a bit different as it tries to really facilitate Git and doesn’t have any internal state it tracks (previous versions did). Now it is really more of just a different way of viewing the state of the Git tree in a way that makes sense in terms of a “Patch Stack” based workflow. I think it also differs in that it is trying to be companion app to Git so that a user can and should use both Git and Git Patch Stack together. Git Patch Stack also tries to align the commands as much as possible around normal git concepts and commands.

This part scared me:

In a Stacked Branch workflow, you are creating branches for each of your atomic commits, and then you’re defining the relationships between those branches so it can formally store a directed acyclic graph of those relationships. Why do you want to make a separate branch for every commit? That sounds like a nightmare, and it’s definitely not the intended workflow for git.

The “Stacked Branch” workflow is a different workflow than “Patch Stack” based workflow. I agree with you that you shouldn’t want to manage branches or the relations between. That is why the “Patch Stack” based workflow that Git Patch Stack facilitates eliminates this. However, there are a lot of people out there that are using a “Stacked Branch” based workflow and tooling around it like https://graphite.dev that do manage branches for each atomic commit and building a DAG (directed acyclic graph) defining their relationships so that their tooling can do things like automate rebasing, etc.

That is actually one of the main points of Git Patch Stack. Gerrit which does support a trunk based workflow of patches similarly. It however requires a completely new set of tooling both on the client and the server side of things. So your team needs to switch to a different SCM platform which isn’t GitHub, GitLab, or Bitbucket, etc. Which means adoption requires an entire team to do a massive switch.

Git Patch Stack on the other hand provides the trunk based workflow while still allowing the user to use Git however they might want, and also interact with the feature branch/pull request model of GitHub, Bitbucket, GitLab, etc. So each dev can locally do what makes sense for them rather than having to get entire team to migrate over to something like Gerrit. Also it doesn’t have any additional state creation. It just provides a meaningful view of the Git tree in a way that makes sense for the trunk based workflow, as well as additional commands to make operating conceptually in that world make more sense. So it makes it easy for people to have hybrid workflows or adopt the workflow over time, etc.