Why I use git and puppet to manage my dotfiles | Hacker News Reader

Why I use git and puppet to manage my dotfiles

jaxelson.com

Seems that the de facto approach is to backup-rm-symlink. My approach is to source from existing configs whenever possible. Specifically, instead of

  mv ~/.bashrc ~/.bashrc.bak; ln -s $DOTFILES_PATH/.bashrc ~/.bashrc

I do a slightly more sophisticated version of:

  echo "source $DOTFILES_PATH/.bashrc" >> ~/.bashrc

(The sophistication comes in checking if that line already exists.)

Here's my install script: https://github.com/shazow/dotfiles/blob/master/install.sh

I've found many benefits in this, such as easily having machine-specific configuration side by side with my unified dotfiles and no backup clutter. This works especially well for things like vim configs, too.

  $ cat ~/.vimrc
  source $DOTFILES_PATH/.vimrc

interesting, as an alternative to this how about how janus does it with a .local file?

https://github.com/carlhuda/janus/blob/master/vimrc#L147

I think puppet in this case is overkill. I simple bash script with a small function to check if it already exists and what not would do just fine and you don't have to install another application (puppet).

I agree that its overkill for the use case mentioned in the article, a Makefile would be more than sufficient. I do the same for my vim configuration using `make`: https://github.com/aonic/vim-settings/blob/master/Makefile

Usage:

  git clone https://github.com/aonic/vim-settings.git
  cd vim-settings
  make

Remove:

  make clean

Nice method, but with a bash script you wouldn't have to manually add any new files to your Makefile. I use this one: https://gist.github.com/975295. Also handles forcibly linking a file if it exists already (with the -f flag).

Of course, you're just using it for one folder so it's not an issue for you ;)

Yeah I can see how this could be overkill. I initially started by writing my own script but I made that script more complicated than necessary (trying to handle removing files and such). So the end result was that I didn't trust the script with my files.

Perhaps there's room for a simple puppet-style configuration manager that lives in just a single script, in some ways similar to sqllite.

Agreed, I use a relatively simple shell script to link all of my files in a "home" directory in dropbox to my real home directory on multiple machines:

    #named link.sh, greps out some common things we don't want to link, also have vim stuff in a different dotfiles repo
    cd $(dirname $0)

    for F in $(ls -a1 | grep -v link.sh | egrep -v "^..?$" | grep -v .DS_Store | egrep -v ".hg(ignore)?$" | egrep -v "^vim$"); do
        if [ -a $HOME/$F ]; then
    		echo "**** Found existing $F, skipping..."
        elif [ -h $HOME/$F ]; then
	        echo "Already symlinked $F, skipping..."
    	else
    	    echo "Linking $F"
    	    ln -s $PWD/$F $HOME/$F 
    	fi
    done

This would be a little hard for me because I like to keep my system configuration files in the same repository (such as fstab and php), although I could setup some standard naming scheme like root.etc.fstab to link to /etc/fstab

If you were going to categorize like that why not just use an etc/ and home/ folder inside your repo?

I prefer to keep things simple, and version the home directory directly. I renamed the ~/.git repository to ~/.dotfiles.git in order to keep it out of the way, and set up a broad .gitignore, which can always be overridden with -f. When I need to work on the repository I use the GIT_DIR environment variable. No need for managing symlinks.

That wouldn't work so well for me because I like to keep my system configuration files like apache's in my configuration folder as well. But I can see why that approach appeals to you.

God, I like reading stuff like this, but personally I think articles about keeping your entire home directory under version control are more interesting (Joey Hess does this, I believe, and probably others I don't know about).

Stuff like this is why I enjoy Hacker News.

I ended up writing my own mini-utility for this: https://github.com/jrockway/dotfiles/tree/master/bin

But I think I am going to switch to puppet, because there are a few things I always need done on a new machine; adding myself as a sudoer, installing packages, editing the imapd config (for a local imap cache), etc.

I used to mv the old files, but now I just rm.

Host- specific files I handle by doing if [ -f ~/.host-bashrc ]; then . ~/.host-bashrc; fi

I also use proper extensions on my files like bashrc.sh and emacs.el for automatic syntax highlighting in editors.

My shell script just goes through the current dir turning eg. bashrc.sh into a link at ~/.bashrc

I do something similar with BitBucket and hg. Over time, I've found I rarely need system specific branches any more. Each computer has a .bashrc-local for local shell customizations, kept outside the repo. Otherwise, it's enough to detect the OS flavor and adjust symlinks that way.

I just use SpiderOak. It includes my dot files during its normal backup routine, it has file revisioning included, and all my files are available immediately upon installing SpiderOak to a new computer.

Personally, I just backup my dotfiles via FTP as they don't change very often, but I lot of my fellow Arch Linux users seem to like backing up their dotfiles to Github. I have never used GitHub myself but everyone seems happy with it.

I use Rake to manage the symlink part of this

https://github.com/mmb/config/blob/master/Rakefile

I use homesick https://github.com/technicalpickles/homesick

Reply on news.ycombinator.com