git stash vs. git reset --hard

Many times, I used git reset --hard and found out later that all my changes had disappeared. I think this is because I do not fully understand what git reset --hard does:

1. move the branch ref pointer to the given reference
2. set HEAD to this ref
3. Depending on the option:

• –soft: working dir and stage are not touched, useful when git reset --hard HEAD^ to change the content of the commit without having to re-add everything
• –mixed: working dir is not touched, but the stage is set to match the HEAD
• –hard: stage + working dir are set to match the HEAD

On the other side, git stash will set stage + working dir to match the HEAD.

So instead of git reset --hard REF I should rather do git stash && git reset --hard REF. This would avoid cases where I forgot that I had made changes… and end up losing them all.

IDEAS ON NEW COMMANDS

• I often have committed something like @~0: edit touist.rb @~1: edit .travis.yml

I want to add a command like git ci --amend~2 that would amend the commit in @~1, not the HEAD. Here is what I must do:

vim .travis.yml
git stash
git rb -i @^^ # edit HEAD^
git stash pop
git add .travis.yml
git ci --amend
git rb --continue

Facebook’s hg extensions have ‘absorb’ which does this but on @~1, @~2… and chooses automatically in which commit it should be amended. See:

• https://stackoverflow.com/questions/24625411/amend-the-second-to-last-commit
• https://blog.filippo.io/git-fixup-amending-an-older-commit/