13 releases

0.6.9 Nov 17, 2022
0.6.7 Jun 4, 2022
0.6.6 Nov 3, 2020
0.6.3 May 29, 2020
0.2.0 Mar 12, 2018

#188 in Development tools

Download history 42/week @ 2022-12-06 8/week @ 2022-12-13 24/week @ 2022-12-20 6/week @ 2022-12-27 13/week @ 2023-01-03 21/week @ 2023-01-10 12/week @ 2023-01-17 12/week @ 2023-01-24 13/week @ 2023-01-31 31/week @ 2023-02-07 56/week @ 2023-02-14 23/week @ 2023-02-21 40/week @ 2023-02-28 39/week @ 2023-03-07 12/week @ 2023-03-14 11/week @ 2023-03-21

103 downloads per month



git absorb

This is a port of Facebook's hg absorb, which I first read about on mozilla.dev.version-control:

  • Facebook demoed hg absorb which is probably the coolest workflow enhancement I've seen to version control in years. Essentially, when your working directory has uncommitted changes on top of draft changesets, you can run hg absorb and the uncommitted modifications are automagically folded ("absorbed") into the appropriate draft ancestor changesets. This is essentially doing hg histedit + "roll" actions without having to make a commit or manually make history modification rules. The command essentially looks at the lines that were modified, finds a changeset modifying those lines, and amends that changeset to include your uncommitted changes. If the changes can't be made without conflicts, they remain uncommitted. This workflow is insanely useful for things like applying review feedback. You just make file changes, run hg absorb and the mapping of changes to commits sorts itself out. It is magical.

Elevator Pitch

You have a feature branch with a few commits. Your teammate reviewed the branch and pointed out a few bugs. You have fixes for the bugs, but you don't want to shove them all into an opaque commit that says fixes, because you believe in atomic commits. Instead of manually finding commit SHAs for git commit --fixup, or running a manual interactive rebase, do this:

git absorb --and-rebase

git absorb will automatically identify which commits are safe to modify, and which staged changes belong to each of those commits. It will then write fixup! commits for each of those changes.

With the --and-rebase flag, these fixup commits will be automatically integrated into the corresponding ones. Alternatively, you can check its output manually if you don't trust it, and then fold the fixups into your feature branch with git's built-in autosquash functionality:

git absorb
git log # check the auto-generated fixup commits
git rebase -i --autosquash master


The easiest way to install git absorb is to download an artifact from the latest tagged release. Artifacts are available for Windows, MacOS, and Linux (built on Ubuntu with statically linked libgit2). If you need a commit that hasn't been released yet, check the latest CI artifact or file an issue.

Alternatively, git absorb is available in the following system package managers:

Repository Command
Arch Linux pacman -S git-absorb
Debian apt install git-absorb
DPorts pkg install git-absorb
FreeBSD Ports pkg install git-absorb
Homebrew and Linuxbrew brew install git-absorb
nixpkgs stable and unstable nix-env -iA nixpkgs.gitAndTools.git-absorb
Ubuntu apt install git-absorb
Void Linux xbps-install -S git-absorb
GNU Guix guix install git-absorb

Compiling from Source

crates.io badge Build

You will need the following:

Then cargo install git-absorb. Make sure that $CARGO_HOME/bin is on your $PATH so that git can find the command. ($CARGO_HOME defaults to ~/.cargo.)

Note that git absorb does not use the system libgit2. This means you do not need to have libgit2 installed to build or run it. However, this does mean you have to be able to build libgit2. (Due to recent changes in the git2 crate, CMake is no longer needed to build it.)

Note: cargo install does not currently know how to install manpages (cargo#2729), so if you use cargo for installation then git absorb --help will not work. Here is a manual workaround, assuming your system has a ~/.local/share/man/man1 directory that man --path knows about:

wget https://raw.githubusercontent.com/tummychow/git-absorb/master/Documentation/git-absorb.1
mv git-absorb.1 ~/.local/share/man/man1


  1. git add any changes that you want to absorb. By design, git absorb will only consider content in the git index (staging area).
  2. git absorb. This will create a sequence of commits on HEAD. Each commit will have a fixup! message indicating the message (if unique) or SHA of the commit it should be squashed into.
  3. If you are satisfied with the output, git rebase -i --autosquash to squash the fixup! commits into their predecessors. You can set the GIT_SEQUENCE_EDITOR environment variable if you don't need to edit the rebase TODO file.
  4. If you are not satisfied (or if something bad happened), git reset --soft to the pre-absorption commit to recover your old state. (You can find the commit in question with git reflog.) And if you think git absorb is at fault, please file an issue.

How it works (roughly)

git absorb works by checking if two patches P1 and P2 commute, that is, if applying P1 before P2 gives the same result as applying P2 before P1.

git absorb considers a range of commits ending at HEAD. The first commit can be specified explicitly with --base <ref>. By default the last 10 commits will be considered (see Configuration below for how to change this).

For each hunk in the index, git absorb will check if that hunk commutes with the last commit, then the one before that, etc. When it finds a commit that does not commute with the hunk, it infers that this is the right parent commit for this change, and the hunk is turned into a fixup commit. If the hunk commutes with all commits in the range, it means we have not found a suitable parent commit for this change; a warning is displayed, and this hunk remains uncommited in the index.


Stack size

When run without --base, git-absorb will only search for candidate commits to fixup within a certain range (by default 10). If you get an error like this:

WARN stack limit reached, limit: 10

edit your local or global .gitconfig and add the following section

    maxStack=50 # Or any other reasonable value for your project


  • implement force flag
  • implement remote default branch check
  • add smaller force flags to disable individual safety checks
  • stop using failure::err_msg and ensure all error output is actionable by the user
  • slightly more log output in the success case
  • more tests (esp main module and integration tests)
  • document stack and commute details
  • more commutation cases (esp copy/rename detection)
  • don't load all hunks in memory simultaneously because they could be huge
  • implement some kind of index locking to protect against concurrent modifications


~318K SLoC