In plain English, what does “git reset” do?

Question

I come from an SVN background and Git is a whole new paradigm. I got mercurial easily, but Git is much more technical.

I think git reset is close to hg revert, but it seems there are differences.

So what exactly does git reset do? Please include detailed explanations about:

• The options --hard, --soft and --merge;
• The strange notation you use with HEAD such as HEAD^ and HEAD~1;
• concrete use cases and workflows;
• consequences on the working copy, the HEAD, and your global stress level.

Answer 1

In git you have:

• HEAD pointer that tells you what commit you're working on
• working tree that represents the state of the files on your system
• staging area (or the index) that "stages" changes so that they can later be committed together

Let us see what --hard, --soft, and --merge do: 

• The parameter --soft will move the HEAD but it will not touch the staging area or the working tree.
• The parameter --mixed will move the HEAD and will update the staging area, but it will not touch the working tree.
• The parameter --merge moves HEAD, resets the staging area, and tries to move all the changes in your working tree into the new working tree.
• The parameter --hard moves HEAD and adjusts your staging area and working tree to the new HEAD, throwing away everything.

You should use --soft only when you want to move to another commit and patch things up without "losing your place". 

$ touch foo                            

$ git add foo                          

$ git commit -m "Bad-commit-message"   

$ git reset --soft HEAD^               

$ git commit -m "Good-commit-message"         

You should use --mixed when you want to see what things look like at another commit, but you don't want to lose any changes you already have.

You should use --merge when you want to move to a new spot but incorporate the changes you already have into that the working tree.

You should use --hard to wipe everything out and start a fresh slate at the new commit.