Git is not Subversion, we are free to rewrite history any way we see fit which is important when producing atomic commits and when we want our sequence of commits to tell a story.
Many people are either unaware that this is possible or are afraid to do but the basics are straightforward to understand.
In this talk I demonstrate my 5 most common use cases for git rebase. Each scenario has a corresponding bash script which sets up a git repository under repo/ ready to show the technique in action.
One of my favourites, allows you to fixup an existing commit
git commit --fixup=HEAD
git rebase -i HEAD^^ --autosquash
Similar to fixup but you can squash together any number of arbitrary commits and then edit the message of the resulting commit
git rebase -i HEAD^^
Simply allows you to reorder a set of commits
Allows you to remove a commit, good for when you've accidentally committed something you shouldn't have
Allows you to reword the message of a commit