Update: 7 months later, I've written some more detailed notes here.

Background

Chezmoi seems to be dotfiles management for power users. Until a few days ago, when I realised I'd massively broken a lot of things, I'd been putting my dotfiles in a version controlled directory and using a shell script to generate symlinks in my home directory.

This had worked really well for several years. It's a great system for maintaining dotfiles on a single machine. But it's not robust or flexible enough for managing multiple machines or multiple operating systems.

Also, it's not just files that sit in $home anymore, I need to track configuration files that live in other places too¹.

I needed something more robust and flexible than generating symlinks from a bash script.

Razor

I think I read somewhere that when purchasing a new appliance you should buy the cheapest you think you can get away with, and if that doesn't work then get the best you can afford. Moving from a custom bash script to Chezmoi is an example of this. My custom bash script was the most basic approach, and it broke badly. Chezmoi seems like the most heavy duty tool for dotfiles management I could find.

For example, I think I'm able to encrypt my files using 1Password or similar. I can use Jinja templates to create scripts for different scenarios. I don't even know what that means at the moment. I've only scratched the surface but I'm happily managing dotfiles across different machines and updating/modifying them safely. Chezmoi (aliased to cm) is free and OSS so I'm confident it'll be "my" tool for the next few decades. I'm happy to invest a couple of hours to learn a few new habits and iron out a few wrinkles².

One Question

I don't understand why chezmoi cd creates a new shell in order to jump into the Chezmoi directory. Why not simply cd into the Chezmoi directory? What's the advantage/necessity of a new shell?

You have to remember to exit after you've done whatever you went there to do, but my habit is to cd or z ... I guess it's nice to exit and then immediately go back to where ever you were before, but there are other ways of doing that - you could write the current directory to an environment variable. It seems unnecessarily complex.

Command Reference

Update - the notes in my follow-up are likely much more useful.

1. Show which files have changed → cm status

2. List of managed files → cm managed

3. List of unmanaged files → cm unmanaged

4. Start tracking a file → cm add >path/to/file<

5. Update a file, add the file again → cm add ..

6. Edit tracked version of file → cm edit <file> - don't think I'm going to use this, I'd rather edit the source file, test it, and then update using cm add ..

7. Differences between local version tracked versions → cm diff <file> - this tells me which files I need to cm add again. This feels clunky and I suspect there is a better workflow. Parsing git style diff files is horrible.

8. Clobber local version with the tracked version → cm apply

9. Dry run and see diff between local version and tracked version → cm -nv apply

10. Pull the latest changes from your repo and apply them → cm update

11. Remove a file →

    • Create .chezmoiremove in the source directory
    • chezmoi apply --remove --dry-run --verbose

12. Type cm instead of chezmoi → alias cm="chezmoi"

Links

• Github
• Project Site