Remove a file from a older commit without affecting any other commits in gitlab

Question

I am using the gitlab repository for uploading my code files. I have made more than 500 commits. My repo size was upto 540 mb. I want to minimize the size of the repo by deleting unwanted files in the repo.

Somehow i managed and deleted the files in my local. Now my current file size is 10 mb in the local. I pushed it and it was good. Eventhough it shows 540mb. I came to know the previous commit files holds the storage. There is a zip file with a size of 400 mb which is in one of my older commit.

I found the commit by using the shell script. I want to delete the zip file in the commit. But it shows "You can only delete files when you are on the branch" in the gitlab webui. In the terminal, I found the command git rm <file_name> to remove the file. I got the file path from the commit. But I don't know how to use it on a particular commit. It holds some important files. I don't wanna play with it.

My question is: how do I remove a certain file in a particular commit in gitlab without affecting any other commits? I tried Stack Overflow and found some answers to delete the file in previous commit but not in the older commit. Help me with some solutions.

Answer 1

The short answer is grim : you can't.

That is, you could modify a commit in the past, but it would modify every commit down the line so it doesn't check the "without affecting anyother commits" part.

Take a look at git filter-branch to have an idea of what tools are at your disposal to change history, but be aware that's quite a heavy operation.

Answer 2

It's no possible to delete the file without affect commit history.

To delete the file in the history, you can use git built-in git-filter-branch or third party bfg-repo-cleaner

Refer to i-want-to-remove-a-large-file-from-ever-existing-in-repo-history of Github: git-flight-rules

recommended-technique-use-third-party-bfg

(master)$ git rm path/to/filetoremove
(master)$ git commit -m "Commit removing filetoremove"
(master)$ java -jar ~/Downloads/bfg.jar --delete-files filetoremove

built-in-technique-use-git-filter-branch

(master)$ git filter-branch --force --index-filter 'git rm --cached --ignore-unmatch filepattern' --prune-empty --tag-name-filter cat -- --all

Answer 3

There's a command-line utility called BFG Repo-Cleaner that will help you do just that.

For example, this command will go through a repo's history and delete a specific file in all commits reachable from any branch:

bfg --delete-files big-400-MB-file.zip path/to/your/repo

You can achieve the same thing using the git filter-branch command and some shell scripting, but this is way easier.

Now, about the second part of your question:

...without affecting any other commits

That's simply not possible. Deleting a file from an older commit will cause its SHA-1 to change, which in turn will change the SHA-1 of all descendant commits. In other words, you're rewriting a large chunk of history, which, depending on your particular scenario, may or may not be a problem:

The command is filter-branch, and it can rewrite huge swaths of your history, so you probably shouldn’t use it unless your project isn’t yet public and other people haven’t based work off the commits you’re about to rewrite.