Skip to content

08_reset.md

Latest commit

 

History

History
124 lines (80 loc) · 3.71 KB

08_reset.md

File metadata and controls

124 lines (80 loc) · 3.71 KB

Git Reset: Undoing Changes

Now that we know how to add commits and merge branches, we need to learn how to get our files back if we mistakenly rewrite them.

The Setup: Making a Mistake

Let's intentionally simulate a disaster where we overwrite a good file with bad data.

echo "# Mistake" > title.md
git add title.md
git commit -m "Mistake made"

At this point, our title.md file is ruined, and worse, we have saved that ruin into the history.

The git reset command is the primary tool to undo recent commits. It can modify the Commit History, the Staging Area (Index), and the Working Directory, depending on the "mode" you use.

1. The Soft Reset (--soft)

The "Undo Commit" Button.

This moves the branch pointer back, but it preserves your work. It leaves your changes in the "Staging Area" (Index), so they are ready to be committed again (perhaps with a fixed file or a different message).

Let's undo that mistake commit:

git reset --soft HEAD~1

Note: HEAD~ notation works reliably on linear history. If you have complex merges, be careful, as discussed in the previous chapter.*

Now, if we check the status:

$ git status
On branch features
Changes to be committed:
  (use "git restore --staged <file>..." to unstage)
        modified:   title.md

We can verify exactly what the changes are (the mistake is still there in the file, but it is no longer "committed"):

git diff --staged
# Or
git log -p --oneline

Output:

-# Titles
-
-- A River Runs Through It
-- Fight Club
-- Dune
+# Mistake

Summary: With git reset --soft, you go back in time in the history, but you keep the current state of your files staged.

2. The Hard Reset (--hard)

The "Delete Forever" Button.

This moves the branch pointer back and makes your Working Directory match that older commit perfectly. Any work done in the commits you jump over is destroyed.

Let's pretend we made the "Mistake" commit again. This time, we want to completely destroy it and revert the file to how it looked before.

git reset --hard HEAD~1

What happens:

  1. The commit "Mistake made" is removed from the history.
  2. The file title.md is physically rewritten on your disk to match the previous commit.
  3. The text "# Mistake" is gone. The original list of titles is back.

FAQ: Reset Nuances

Does git reset --hard delete untracked files?

Generally, No.

If you create a new file (e.g., secret_notes.txt) but never add it to git (it is "Untracked"), git reset --hard will leave it alone. Git only modifies files it is tracking.

However, if the commit you are resetting to contains a file named secret_notes.txt, Git might overwrite your untracked file to match the commit. But typically, untracked files are safe from a reset.

Is it possible to recover files after git reset --hard?

Yes, but with a catch.

If you committed the files (even if you just reset them away), Git still has them in its database for a short time. You can find them using the Reflog.

  1. Check the Reflog:

    git reflog

    This shows a log of everywhere HEAD has been, even if you reset.

    a1b2c3d HEAD@{0}: reset: moving to HEAD~1
99z88y7 HEAD@{1}: commit: Mistake made

  2. Recover the lost commit: Identify the hash of the commit you just deleted (99z88y7 in the example above) and reset back to it:

    git reset --hard 99z88y7

Warning: If you had changes in your working directory that you never added or committed, and you ran git reset --hard, those specific changes are lost forever. Git cannot recover what it never saw.