In past, I have seen several processes create .lock files, to usually write pid of the process.

e.g. Emacs desktop package creates .emacs.desktop.lock file to save the PID of the emacs instance that created this file. Eventually, it will use this to ensure that no other instance overwrites it, and also gives a warning as such.

There have been other instances we well.

So when I started working on elixir, I always ignored mix.lock file. When I refer to other peoples elixir repo on github, I always wondered why are they committing this file to git.

I have seen junior/inexperienced developers committing everything to git, including files which are built. So I ignored such “mistakes”

How naive I was.

Recently when I started working with phoenix, I bothered looking at the .gitignore file. I was surprised that mix.lock was not in it.

People who created phoenix framework are smart.

If they are smart enough to create a “default” .gitignore file, they know what they want in there.

So I looked inside the mix.lock file, and realized that it is definitely not PID :)

I should have thought about it earlier, since mix is not a long running/daemon process. So why would it need to store the PID ?

On reading the file, I realized that it is equivalent of requirements.txt files from the python world, but automatically created !

The documentation specifically asks to commit this file to version control !

Here is the relevant part of the documentation:

Dependency fetching is repeatable, Mix will lock the version of a dependency in the lockfile to ensure that all developers will get the same version (always commit mix.lock to version control). $ mix deps.update will update the dependency and write the updated version to the lockfile.

You can read up the original/details here


Turns out python is also moving towards similar .lock file mechanism over requirements.txt. See this.