Skip to main content

Git magic - Bring back from the dead

I had this interesting scenario where I did hard reset and force push to a commit. However I wanted one of the change from the lost commit. 

Once again, Git magic came in handy. Git never erases anything off its memory too soon. This magic spell as I call, is one of a kind and most git enthusiasts must know it by now. 

Introducing reflog

We all know what you get when you do git log. Let me anyway spell it out, git log gives you a list of changes or in other words, commits available as of the moment in that local repository in the chronologically descending order. 

Similarly, git reflog gives you a list if changes or in other words, change history that happened in that local repository in chronologically descending order. For instance, if you have amended the same commit 5 times, you can choose to go back to the 3rd commit amend and restart your work from there.

This situation is what I call as, 

Bring back from the dead

Let us consider the following scenario,

  • You worked and pushed a change
  • Customer comes in and says that they don’t want the change and so you git reset hard to the previous commit
  • Now after a day, customer comes back and says they want that change back again

What options do you have?

  • Try and bring back the changes from your mind?
  • Look for the closed PR changes and re-do them manually from there?
  • or worse, scold your customer inside yourself and do the first point?
I’ve been at that stages. Some of my backend buddies used to use their IDE’s local cache to get them back. All this while me, a lame Sublime text user has been under the impression it’s some magic that IntelliJ does that sublime cannot. Little I knew it was all Git underneath. 

What is the real option here?

It's Bring it back from the dead. Please read on. 😊
  1. Open your git repo in terminal
  2. Do git reflog
  3. This will show a huge list of historical changes. When you look at this, you’ll know the steps you did to get your repo to where it is now. git log is like calculating displacement and git reflog is like finding the distance covered, what all stops you made.
  4. A typical reflog will look like,

    ef9080e HEAD@{1}: commit (amend): build: add changes for success
    gh0191f HEAD@{2}: commit (amend): build: add changes for success
    hi1202g HEAD@{3}: commit (amend): build: add changes for success
    f1be775 HEAD@{4}: reset: moving to HEAD@{1}
    e3g351f HEAD@{5}: reset: moving to HEAD@{13}
    6b63233 HEAD@{6}: rebase (finish): returning to refs/heads/build-failing-issue
    6b63233 HEAD@{7}: rebase (pick): build: testing build fail
    2224786 (tag: v1.9.0, upstream/master, origin/master, origin/HEAD, master) HEAD@{8}: rebase (start): checkout upstream/master

    You could know what change happened at that Head pointer by reading the type as highlighted in the orange and the change message in black.

    /**
     *
    @disclaimer
     * The commit hash and messages used are not real and wrote for demo purposes
    */
  5. Here choose a change you want to go back to. Unlike the commit hash, you’ll have to choose a head pointer. Something like HEAD@{1} or HEAD@{10} or HEAD@{15} or it could go on
  6. Copy that and do git reset --hard {copied_HEAD_pointer} which will look something like,

       
    git reset --hard HEAD@{4}

  7. Tada. You’ve moved to the change you need. Just do a git push --force and you’re ready to ship the code. 
I hope this is helpful for folks who like terminal a lot more than IDEs and for people who were curious to know the tech underneath the IDEs

Comments

Popular posts from this blog

Git - Removing a file from a commit

Once again, another git magic that might be of help to some of you. This research came up when I accidentally added a couple of unwanted files and wanted to remove them from a commit. We all know that to update an existing commit, we shall follow this git magic to amend commits . However, how do we drop changes to a particular file in a commit? While that is easy, it is also tricky. This blog post covers scenarios and respective commands that help you understand what should be done. /**  *  @disclaimer  * Please read this post fully before executing any command. My scenario might not be the same as yours. */ There are two scenarios to be handled here, Remove a newly added file Remove changes to an existing file Let's look at them separately, Remove a newly added file This is the scenario where an unwanted file added to the commit. This file might be some config.json that got generated while doing a research on a new testing tool or a bundling tool. Such scenarios are easier. 1. Re

Git - How to drop a commit in history?

Back after a while with another git magic. We already have seen how to get rid of the top most commit in this article ->  Git magic - Make commits disappear First of all, it's one of the not so recommended way of doing it and more than that, it can only get rid of sequential commits from latest. I recently happened to get into a scenario where I had to drop a commit in the history by keeping the latest ones. As part of that exploration, presenting you this article. /**  *  @disclaimer  * Please read this post fully before executing any command. My scenario might not be same as yours. */ To my greatest surprise, I didn't know a git rebase could do this. Please continue to read to know how. The steps are simple, Count until the commit line you need to drop using git log Do an interactive rebase and you're done.  Let me explain step by step. 1. Look at the commit log & count commits This is the first step. First let's list the commits using the following command,  

Git worktree - guide to flexible folder structure

As always, let's start with a story why I have written this one down. I have been using git for almost 5 years now and in one of my work places I had to work on the same project but different features in parallel.  Switching between branches back and forth was a costly operation given that I didn't discover commit amend until recently. Even though I'd have discovered commit amend, I'd still have did this. Still did what? I can sense that question deep from your throats. So, whenever I am in a situation to work on multiple features or a feature and a bug fix, I'd have two clones of my repo each with different branches, Whenever a feature is merged or the fixes are merged, I'll delete the clone in my machine and branch in my origin (Gitlab / Github etc.,) Recently when I started writing these git articles, my manager suggested me to learn about Git Worktree and it'll be useful. And when I got free and good understanding of git commit, rebase and stuff, I decid