6 ms·
Splitting a Git Commit
- furkanturan 2mo agoI never thought about whether this was possible, but now that I know it is, I immediately see how useful it could be for my workflow.
- jaltekruse 2mo agoI had a team that insisted on a linear history and was constantly rebasing things. This workflow strongly favors "just squash it all together" before a rebase to avoid re-resolving the same conflicts. This workflow often produces big patches that reviewers end up asking to be broken back up. This experience and a recommendation from my manager at the time made a good git GUI like sourcetree a daily driver for me. It's just second nature to review my own changes every time I stage files and quickly do a mix of staging whole files or "hunk-based", even allowing clicking to select specific lines within a hunk intuitively, rather than using the terminal UI and asking git to take a guess at splitting a hunk. I find a good git GUI invaluable, I have used sourcetree for years and love it. Unfortunately sourcetree isn't available for Linux, but I recently found SourceGit which seems to be a decent substitute. Highly recommend trying a good GUI out, I think you would likely have similar ah-ha moments about workflow optimizations, and honestly I just cannot understand how people get by in git without a convenient way to visualize the commit tree (and yeah I know there is a decent command line treeview, but the context switching all the time in the terminal to go from viewing the tree or even the simple log to exit out and ask for commit contents just is so much more fussy and tedious).
- esafak 2mo agohttps://github.com/DetachHead/rebased https://github.com/DetachHead/rebased
- farlight 2mo agogit rerere will resolve repeated conflicts for you automatically, you just need to enable it. https://git-scm.com/docs/git-rerere https://git-scm.com/docs/git-rerere
- nullsanity 2mo ago[dead]
- WorldMaker 2mo agoBe careful, it will also cache any mistakes in merging including things like bug regressions. Especially if you also have a git history full of cherry-picking and reverts.
- sunshowers 2mo agoBTW not sure how you feel about LLMs but merge conflicts are not really an issue with them any more.
- ivanjermakov 2mo agoStackOverflow is really not great at handling questions for which the right answer changes over time. I guess it doesn't matter since SO is dead anyway: https://news.ycombinator.com/item?id=46482345 https://news.ycombinator.com/item?id=46482345
- locknitpicker 2mo ago> StackOverflow is really not great at handling questions for which the right answer changes over time. The right answer doesn't change over time. It can vary with the release version you are targeting.
- roelschroeven 2mo ago> P.S. The stack overflow question for splitting commits discusses the old and cumbersome approach. The 20th answer discusses the right approach but has a meagre 2 upvotes as of today, compared to 2656 for the older top answer with the cumbersome approach. It now has 5 upvotes. A long way to go to 2656 but at least it's going in the right direction.
- hotelsacher 2mo agoStill, it shows how incredibly broken the StackOverflow system is. It's completely normal for older answers to become outdated, but they still act like the old answers are laws of nature.
- BeetleB 2mo agoI think this is harsh criticism. It's because not many people use SO any more. One of SO's selling points was that outdated answers get replaced with the more modern, better ones - and it worked for many, many years.
- hungryhobbit 2mo ago... until they decided it was a good idea to train all power users to grief new users off the site.
- jatora 2mo agoThe rise of AI and the death of SO is a beautiful beautiful thing. That community is the worst, and this opinion seems to be shared universally by anyone who had extensive interaction with it.
- hotelsacher 2mo agoAI could gorge itself once on (probably stolen) training data from SO, among others, but who will answer new questions now, and where will new training data come from?
- svyatoslavpavl 2mo agoThe bit that made this click for me: it works on any commit in history, not just HEAD. Mark the commit as edit in git rebase -i, then git reset HEAD^ to uncommit it while keeping the changes in the working tree, and rebuild the pieces from there. git add -p is the real workhorse for the rebuild: s splits a hunk into smaller ones, and e lets you hand-edit the hunk when the boundary doesn't fall on clean line breaks. Stage a coherent slice, git commit, repeat until the tree is empty, then git rebase --continue. The rebase's only job is to drop you at the right spot; add -p does the actual splitting.
- bonzini 2mo agoOr the opposite direction, with repeated "git checkout -p ..." and commit steps. Slightly less comfortable because you have to figure out (such as from "git rebase --edit-todo") the later commit, but it makes testing the pieces easier.
- svyatoslavpavl 2mo agohadn't thought of building forward like that. being able to test each slice before committing beats finding out after the rebase that half of them don't compile. thanks, stealing this
- wasting_time 2mo agoPlease don't use LLM to write HN comments. To quote the guidelines[0], "HN is for conversation between humans". [0] https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html
- diath 2mo agoThat's good to know, I'd usually go about it in a roundabout way: soft reset the commit then git add --patch to stage the hunks to split it into multiple commits.
- qsera 2mo agoI do the same, but use `git gui` to select the changes.
- merb 2mo agoWhich gui? I’m on Mac and somehow I could never really be happy with gui‘s for git. Gitkraken is the last I tried and I found it too much cluttering. Paid choices weren’t that great aswell either (yet)
- qsera 2mo agohttps://git-scm.com/docs/git-gui https://git-scm.com/docs/git-gui I use this only for the aforementioned purpose.
- insane_dreamer 2mo agotry lazygit
- ericfrederich 2mo agoYeah, I've been using git for nearly 20 years. I still rely on `git gui`. It's my crutch. I know you can do all of that in a terminal, but staging different hunks (or individual lines) is super easy.
- adregan 2mo agoI think you'll find your workflow is still required. `git history split` is really only focused on turning a commit into 2 commits, so if you wanted multiple, you would need to run the command multiple times. Add to that, it's patch functionality isn't as robust as `git add --patch`. For example, you cannot edit a hunk, so if you intended to tease out atomic changes, you won't be able to.
- matheusmoreira 2mo agoToday I learned about git history split. Thanks!!
- ajasmin 2mo agoLove the minimalist blog layout
- BeetleB 2mo agoIs this yet another case of git improving its UI due to jujutsu? (This is exactly how jj does it.) If you like this kind of convenient CLI command, I would seriously recommend giving jj a try. To give you an idea, I never bothered splitting a commit in my git days. I do it almost daily with jj. Lots of other niceties that I didn't bother with in git, but routinely do in jj. jj has fewer commands than git, yet does everything git does.
- steveklabnik 2mo agoMy understanding is that `git history` was started after seeing jj, yes. I don't have a citation for you though, I could be wrong, that's just what I remember.
- prattmic 2mo ago> “That moment when you realize that a tool simply fixes all the UI issues that you had and that you have been developing for the last 20 years was not exactly great." He had two options: despair or learn from the competition. He chose to learn from it. https://lwn.net/Articles/1057561/ https://lwn.net/Articles/1057561/ which goes on to explain that this was the motivation for git history.
- BeetleB 2mo ago"He discussed the workflow for splitting a Git commit, which involves seven separate commands with Git's current UI. Most users do not know how to do this, he said. The goal is to add several ""opinionated subcommands"" that make more modern styles of working with merge requests, such as stacked branches, much easier."
- rkangel 2mo agoYes this is how jj does it, except that jj is more powerful because it does automatic rebase of all the downstream commits. If you have any descendants of the gir history split (i.e. it's not the most recent) you'll have to do some more work (a rebase) to get back to a useful history. I think you can do this in the middle of an interactive rebase which is ok, but still not as easy.
- opello 2mo agoIt seems strangely polarizing to refer to the "old and cumbersome" interactive rebase approach in contrast to "the right approach" using `git history` which is a brand new, experimental feature as of Git 2.54 from April 20, 2026, a few months ago. I have split a lot of work with interactive rebasing. I'm excited to try `git history split` and to have learned about it here. But the stylistic flourish makes it seem like it's been a feature for years that people just haven't moved to instead of being something shiny and new to be excited about trying.
- locknitpicker 2mo agoI agree, interactive rebasing existed for ages and git history is mainly a user experience improvement over it. This way of framing git history goes way past engagement tricks and erodes the blogger's credibility.
- schacon 2mo agoNot sure if this is written by an AI, but I'll engage, because it's a bit misleading. The new `git history` command can take three verbs - fixup, reword and split. While interactive rebasing can do fixup and rewording relatively easily and the `history` variant is a slight improvement, the split sub-subcommand is really _quite_ difficult to do in other Git tooling, including interactive rebase. Before reading on, I would challenge the reader to think about how they might do it. The answer is that you would rebase back to at least the commit you're splitting, choose 'edit' in the rebase script for that commit (pick the rest), then when the rebase stops there, `git add (-p)` the parts you want, commit, then commit the rest, then `git rebase --continue`. I would guess that maybe 5% of the readers of this paragraph would have guessed that correctly, if I'm being pretty positive. Patrick wrote the `git history` stuff because tools like JJ and GitButler are pushing UX that makes this type of thing very easy and Git itself is struggling to catch up. Even at a fundamental level, the first versions of history were based on the sequencer code that rebase used but it couldn't do the job properly (messing with the index/workdir), so he rewrote it based on the `git replay` machinery instead (which _itself_ is still somewhat experimental). Interestingly, it's _still_ not optimized for agents or scripting - `split` specifically needs an interactive terminal like `git add -p` does, so not many agents are good at it. I recently tried it and Claude piped `y\ny\nn` through the command, guessing at the interactive input (y, y, n) needed to stage the hunks. The point is, interactive rebasing is a horrible solution to this problem, `git history` clearly better, at least for splitting, but really we need non-interactive solutions to problems like this so agents can do them well. GitButler does this well because we focus specifically on it - our status command gives hunks/files ids and hunk/file movement can happen between commits so splitting is creating an empty commit and then moving the parts over - but it's surprising that nobody else builds tooling for this increasingly common use case. I think `git history` will slowly get better at this over time, but "interactive rebasing" is a _much_ more error prone and unintuitive way to accomplish this.
- mateusz834 2mo agoor use jj and stop worrying about git jj split --interactive {ref}
- bywater 2mo agoThe SO upvote problem is real. Once an answer hits a few hundred votes, nobody scrolls past it to check if something better came along.
- meling 2mo agoI’m doing rebasing a lot these days, especially since I started using gh stack. The main struggle I have with rebasing is for git to recognize that a branch has actually been merged when its commits have changed (e.g., if I forgot to delete the local branch with old commits and come back months later and trying to figure out if it was actually merged or not). My understanding is that this is nicer with jj when you work locally, but if you sync with GitHub, I think your still faced with the same problem, or? What are people doing to workaround this? I’ve tried to ask LLMs, but the complexity is frankly a bit off putting (I don’t have the suggestions handy)
- skydhash 2mo agoDon’t you have an id system, like an issue tracker? That’s your source of truth for change requests. The common advice is to never have long standing branches for WIP. But if you do, squash it to have a small number of commits, then replay it on top of the default branch. The try to understand why there’s any diff or conflict (refactor, update to the design, code removal, was not merged at all,…). As I squash merge, I always remove the PR branch from the main repo. And I keep it for about a week on my local repo (In case the PR is reverted and I needed my drafting version of it).
- nulltrace 2mo ago[dead]
- schacon 2mo agoIn case you're curious if _you_ can run `git history split`, the answer is almost certainly no, unless you manually installed the newest release of Git within the last 3 months. `git history` was introduced as an experimental command less than 4 months ago, which means that there are essentially zero OS releases that packages a version that contains it by default. Also, if you're on OSX and run `brew install git`, it will not overwrite the Apple version that is too old to have it. Git 2.54 was the first release with this subcommand and it is not in Xcode CLT, Debian testing, Ubuntu 26.04 - nothing. It will be "normal" in a year, but it's ridiculous to criticize SO for "old" methods of splitting long before anyone ships with it.
- gnoack 2mo agoPoint taken, that's true. I'm admittedly developing on distributions where the updates are faster, so I missed that. But then again, I also didn't expect that this short weblog article would have such a large audience when I wrote it. :)
- nemetroid 2mo agoIt's in Fedora, Arch, and many other popular operating systems.
- Zopieux 2mo agoIt's funny how people assume Dehian/Ubuntu's ice cold release schedule is the norm, when the rest of us have been enjoying rolling releases and timely, stress-free (easy rollback) updates with NixOS and Arch back in the days. It's been at 10+ years now.
- jez 2mo agoThis is a relatively recent feature. I was first introduced to it in GitHub's release notes for Git 2.54: https://github.blog/open-source/git/highlights-from-git-2-54/ https://github.blog/open-source/git/highlights-from-git-2-54... And discussion at the time: https://news.ycombinator.com/item?id=47837698 https://news.ycombinator.com/item?id=47837698
- mohamedkoubaa 2mo agoI think I'd rather keep my old school interactive rebase skills sharp for when I actually need it
- account42 2mo agoYes I prefer to keep using the much more general rebase interface instead of memorizing more verbs for very specific tasks.
- newsomix9xl 2mo agoOstensibly elite is always the way to go.
- e40 2mo agoNo mention of Magit?