The .gitkeep technique

This technique uses an empty file called .gitkeep:

The empty file ensures that Git creates the directory with minimal cost. Any other filename may be used, as Git doesn’t treat .gitkeep files any differently.

To set this up, you might create an empty file with touch:

Then ignore all files in the directory, except .gitkeep, by adding patterns in the repository’s .gitignore file:

/build/*
!/build/.gitkeep

The first pattern ignores everything in the build directory. The second one then un-ignores the .gitkeep file, allowing it to be committed.

This technique works, but it has some downsides:

1. It requires editing two files.
2. If the directory is renamed, .gitignore needs updating, which is easy to miss.
3. .gitkeep is not a name recognized by Git, so there’s no documentation on it, potentially confusing other developers.

There’s a better way that doesn’t have these flaws.

The better .gitignore technique

This technique uses only a short .gitignore file inside the directory:

build
└── .gitignore

The .gitignore file has these contents:

The first pattern ignores all files in the directory. The second one then un-ignores the .gitignore file, so it can be committed.

You can create this file with echo and file redirection:

$ echo '*\n!.gitignore' > build/.gitignore

When you add and commit the directory, Git will pick up on the .gitignore file first, skipping other files within the directory:

$ git add build

$ git status
On branch main
Changes to be committed:
        new file:   build/.gitignore

$ git commit -m "Track build directory"
[main 1cc9120] Track build directory
 1 file changed, 2 insertions(+)
 create mode 100644 build/.gitignore

The directory is now “tracked” with a single, standard file that will work even after renames.

Fin

Don’t ignore this technique,

—Adam