I have been frustrated by the fact that git, in spite of its tremendous power and flexibility, is simply not designed to deal with "projects" that are comprised of only a single file, often in a directory along with a bunch of other individual text files that I might also like to revision independently of each other.

"Just use RCS," you might say. Yeah, well, but... I like git, want to keep using it, to learn it better, and, well, RCS is just too ancient. Let's be serious.

So I finally sat down and wrote a shell script to handle this modus operandi. Basically, it creates a separate bare repo for each file, with a unique name, and takes advantage of the fact that git has a command line option to override the default repo directory name .git.

This is written to run on UNIX and UNIX-like OSes. If you want to port it to Windows or whatever, be my guest. I've put it in the public domain, so you can do with it as you please. Pull requests accepted.

Instructions

First, copy the script into your execution path. I hate the executable name git1, but haven't yet thought of something I like better. I encourage you to change it to whatever suits you. Just naming the executable file as you like will do the trick.

Create a bare repo for the file you want to revision (e.g. foo.txt), by calling this script with the init subcommand (note that these are commands to the script, not to git directly!):

  git1 init foo.txt

You should see git output indicating successful repo creation. The file has also been checked in to the new repo. Needless to say, this step is only performed once, at the outset. The repo is named .g1_ + the filename, so in this case .g1_foo.txt.

Now, you may call the present script with the filename as first argument, and follow with any normal git commands and options:

  git1 foo.txt status
  git1 foo.txt commit -m "added new blah"
  git1 foo.txt diff
  git1 foo.txt log

etc.

Because the bare repo is unique to the file foo.txt, you can create as many of these as you like in the same directory, and commands will always have the revisioned filename as first argument, to distinguish which repo git should use. Note that this script can only be used from inside the same directory where the file and its repo exist.

To see what repos exist in the current directory:

 git1 ls

To remove a repo (though this does NOT remove the tracked file):

 git1 rm foo.txt

Protip: If you'll be doing a lot of git work on a given file at a particular time, considering creating a temporary alias for git1 foo.txt:

 alias gg='git1 foo.txt'

Then, commands can be simplified: gg diff, gg commit, etc.