git and empty folders

Question

So I see Git doesn't recognize folders, or should I say when the only change between commits is addition of empty folders to the working tree they're not show in git status after git add ..

How would you handle the need to add empty folders to the working tree (for runtime storage) and have them be reflected/created when other repositories pull from the current repository (one in which the folders were added)?

Rebarbative

possible duplicate of How do I add an empty directory to a git repository – Celerity

I've done a little blog post about this here developertipoftheday.com/2011/07/… that might help. – Dasteel

Answer 1

People often put an empty file as a placeholder in the folder to overcome that limitation...

Hypnogenesis

personally i'd make it a readme file with some brief explanation of why the folder exists – Chiquitachirico

@Chiquitachirico that or create the directory when needed. It isn't git's fault your shell scripts or build tool die when a directory doesn't exist. – Ubana

@omouse indeed this is usually a workaround for a buggy build or deploy tool/script – Chiquitachirico

Answer 2

I usually put a .gitignore in those directories as you likely want to ignore any runtime generated data anyway.

Upbraid

Another convention is to put a file called ".gitkeep" in the directory. See: https://mcmap.net/q/12621/-what-are-the-differences-between-gitignore-and-gitkeep – Intermediacy

Answer 3

I usually put empty .gitkeep in those folders.

Rooney

This is the pattern I was looking for – Golliner

Answer 4

You are absolutely correct. Git, like some other version control systems, does not take cognizance of empty folders or of properties of folders. Folders only exist insofar as there are file that are in them. If you want to simulate this, you need to drop placeholder files into them.

Ocieock

Answer 5

You should realize that you are asking for your source control system to set up some resources that are not source, not part of your build system, but essential to the operation of your application. I would do one of the following:

1. Make the build create the necessary directory.
2. Have the application create the required directories as it runs.

Similarly, if the application wanted to append to a log file, it wouldn't make sense to have that logfile start out in the revision control system, would it?

Haney

Answer 6

Either put empty placeholder files into the folders you want git to keep track of or add instructions to create these folders to your build system if possible.

Mariammarian

Answer 7

I am also facing the same issue. I think it has got to do with the way Git tracks changes. It does not track files but rather its contents. When showing the changes, commits or logs it maps the changes to specific files and shows the changes in individual files. See this video for more details Tech Talk: Linus Torvalds on git Linus specifically gives an example about moving a function from one file to another and how git is able to track the change.

Gearalt

Answer 8

For Linux / MAC OSX, use the following one-liner, to create / put empty placeholder files into the folders, recursively, to prevent them from getting ignored, by git:

find . -type d -empty -not -path "./.git/*" -exec touch {}/.gitkeep \;

Hope this helps someone.

Subito

Answer 9

I faced this problem when trying to use git as a backup tool with support for deduplication and compression.

My solution was to create my own system. It's available on http://github.com/meingbg/store

Again, my purpose was to store files, not work with code.

Amadavat

Why don't you just use rsync? – Ubana

@omouse: rsync does not track changes. meingbg might be interested in git-annex, a git-based solution that tracks changes in files. – Sanitary