I'm a Systems Reliability and DevOps engineer for Netdata Inc. When not working, I enjoy studying linguistics and history, playing video games, and cooking all kinds of international cuisine.
The simple explanation is that each commit in git has one or more 'parents', which are commits that come logically before it in the change history, and git rebase changes what the parent of the base commit (that is, the commit that you started the branch from) in a branch is. Essentially, if you think of each commit as a folder in the filesystem, it's parent is the folder that contains it, and git rebase is equivalent to moving it into a different folder (of course, with lots of extra stuff to sanely handle conflicts).
This is useful for a number of things:
Assuming you have a branch B based on another branch A, it lets you update B to include changes from A without needing a merge commit. This usually results in history that much is easier to understand than if you had merged A into B to do the same update. The general syntax for this is git rebase A B (or, if you have branch B checked out, git rebase A). In many workflows, this is the preferred way of updating branches that you have not published anywhere because it helps keep the history simple. This is about 95% of what most people will ever need git rebase for.
Assuming you have a branch C based on a branch B but need it to be based on a different branch A, it lets you transplant branch C so that it's based on branch A instead of branch B. This is useful when you accidentally start a branch from the wrong base. The general syntax for this is git rebase --onto A B C. This covers a significant majority of the remaining 5% of what most people will ever need it for.
Because git rebase can take tags or even commits instead of branch names, you can use it to move individual commits or sequences of consecutive commits from one branch to another with the git rebase --onto syntax. This is useful when you end up developing a fix for an existing bug as part of developing a new feature, and need to submit that fix separately from the new feature. Note however that git cherry-pick plus a carefully crafted git reset is often preferred here, as it's a lot easier to get right than doing the same thing with git rebase --onto (and it's usually easier to fix if you get it wrong).
Similarly because of the fact that it accepts commits or tags, you can use git rebase to remove a series of commits from further back in the history of a branch using the git rebase --onto syntax. This is not something that is frequently needed if you're doing your development work properly, but if you haven't published a branch yet, this may be preferable to reverting a prior commit in the branch as it will make the history more concise.
Aside from the above, you can perform arbitrary transformations to the history of a branch by doing an interactive rebase (git rebase -i, also works with the the --onto syntax, though that tends to be very confusing so it's generally not the best idea to do it in one step (you can always git rebase -i and then git rebase --onto as two separate steps)). This is only very rarely needed in a vast majority of development workflows (especially if you're doing your development work right), but it lets you do very complex things like splitting a commit into multiple commits, merging a bunch of commits into one, reordering the commits on the branch, or even just running a given command in the working tree for each commit on the branch (useful for testing proper bisectability of a sequence of changes, which is a major reason that git rebase -i exists in the first place).
There is, however, one very big caveat to using git rebase. Because it modifies history, you have to use a force push whenever you rebase a branch and then have to update a remote copy of that branch. This plus the rewritten history makes it very difficult for others to update copies of that branch in any clones of the repository (it requires them to delete their local copy of the branch, re-fetch the refs for it from the remote, and then check it out again), especially if they have any local work based on it. As a result, the general advice is to avoid rebasing any published branches outside of situations that you've agreed with your coworkers that it's acceptable to do so.
You may also want to look at git cherry-pick, which lets you copy commits from one place to another (unlike git rebase, which functionally moves commits).
I design compilers, write mobile and web apps, make distributed apps, develop augmented reality and design. I write, paint and read. A lot. Huge Linux enthusiast.
Blog: https://nirlanka.com
The simple explanation is that each commit in
githas one or more 'parents', which are commits that come logically before it in the change history, andgit rebasechanges what the parent of the base commit (that is, the commit that you started the branch from) in a branch is. Essentially, if you think of each commit as a folder in the filesystem, it's parent is the folder that contains it, andgit rebaseis equivalent to moving it into a different folder (of course, with lots of extra stuff to sanely handle conflicts).This is useful for a number of things:
git rebase A B(or, if you have branch B checked out,git rebase A). In many workflows, this is the preferred way of updating branches that you have not published anywhere because it helps keep the history simple. This is about 95% of what most people will ever needgit rebasefor.git rebase --onto A B C. This covers a significant majority of the remaining 5% of what most people will ever need it for.git rebasecan take tags or even commits instead of branch names, you can use it to move individual commits or sequences of consecutive commits from one branch to another with thegit rebase --ontosyntax. This is useful when you end up developing a fix for an existing bug as part of developing a new feature, and need to submit that fix separately from the new feature. Note however thatgit cherry-pickplus a carefully craftedgit resetis often preferred here, as it's a lot easier to get right than doing the same thing withgit rebase --onto(and it's usually easier to fix if you get it wrong).git rebaseto remove a series of commits from further back in the history of a branch using thegit rebase --ontosyntax. This is not something that is frequently needed if you're doing your development work properly, but if you haven't published a branch yet, this may be preferable to reverting a prior commit in the branch as it will make the history more concise.git rebase -i, also works with the the--ontosyntax, though that tends to be very confusing so it's generally not the best idea to do it in one step (you can alwaysgit rebase -iand thengit rebase --ontoas two separate steps)). This is only very rarely needed in a vast majority of development workflows (especially if you're doing your development work right), but it lets you do very complex things like splitting a commit into multiple commits, merging a bunch of commits into one, reordering the commits on the branch, or even just running a given command in the working tree for each commit on the branch (useful for testing proper bisectability of a sequence of changes, which is a major reason thatgit rebase -iexists in the first place).There is, however, one very big caveat to using
git rebase. Because it modifies history, you have to use a force push whenever you rebase a branch and then have to update a remote copy of that branch. This plus the rewritten history makes it very difficult for others to update copies of that branch in any clones of the repository (it requires them to delete their local copy of the branch, re-fetch the refs for it from the remote, and then check it out again), especially if they have any local work based on it. As a result, the general advice is to avoid rebasing any published branches outside of situations that you've agreed with your coworkers that it's acceptable to do so.You may also want to look at
git cherry-pick, which lets you copy commits from one place to another (unlikegit rebase, which functionally moves commits).Haha. Sarcasm :D ;)
Woah !