To install, you need to perform following steps:

  1. Clone repository to a stable place on your box

    For now, I'm going to assume it's local/mydots

    $ cd local/mydots
  2. Move (and merge if needed) your existing configs to dotfiles folder. Remove the dot from name, e.g.:

    $ mv ~/.gnucash dotfiles/gnucash

    If your configs live under ~/.config folder, put them under a similar folder under dotfiles:

    $ mv ~/.config/dunst dotfiles/config/dunst

    Since unlike the "classic" scheme where dotfiles live right under $HOME, with ".config", you don't want to share whole config, so you need to tell dottum to look for link targets on deeper level. This is done by adding entry to file "dotfiles/dotvault/subvaults":

    $ mkdir dotfiles/dotvault
    $ echo "config" >> dotfiles/dotvault/subvaults

    This will caulse dottum to look inside the config sub-folder rather than treat it as whole and create link at ~/.config. Note that you can do this at any depth, for example, in vault that keeps only host specific files, you can do

    $ echo "config/myapp/hosts" >> dotfiles/dotvault/subvaults

    and have dottum look for items under config/myapp/hosts and link them under ~/.config/myapp/hosts, eg. ~/.config/myapp/hosts/foo -> ~/local/mydots/dotfiles/config/myapp/foo

  3. Run dottum script with single argument: path to dotfiles folder (or how you named it), e.g.:

    $ dottum ~/local/mydots/dotfiles

    The script will create links back in your $HOME, uless they (or real files) did already exist, in which case it will only warn you. This makes it easy to exclude some of your applications from some of your machines: just skip step 2 for them.

    Pro tip: Notice how the dottum script discovers your .config/-bound files such as dunst.

    1. The path you pointed to is "path/to/dotfiles", so this will be regarded as "base path" of your config pool.

    2. Any members of the base directory will be prefixed with single dot and symlinked directly to your home--just as the ancient tradition dictates.

    3. If the "base path" has neighbor called the same but with suffix .config, i.e. path/to/dotfiles.config, it will symlink all present members to ~/.config path--just as XDG convention uses in its most common version.

    4. The case is similar for .local/share files; here you'll just need to go one step deeper: files that should be linked to ~/.local/share should be in dotfiles.local/share.

Instructions for bash

Although similar technique could be used for bash, I prefer to have defaults specified by distro. Therefore I use other technique:

Instead of step 2 above, move/merge your specific ~/.bashrc changes to either one of these, or preferably, split them as needed:


Then, replace your ~/.bashrc with your default distro's one, commonly found in /etc/skel/.bashrc, and use dotfiles/bash/setup to patch it:

$ cp /etc/skel/.bashrc ~/
$ local/mydots/dotfiles/bash/setup

From that point on, .bashrc will source appropriate fies for you, so do not make any changes to your local .bashrc, unless you really want them to be local, i.e. private only.

Instructions for Vim

If you want to keep default vim-ish behavior and have all local config in ~/.vimrc, you can use above generic steps. However, if you want to have rather more sophisticated model, and for example use my vim configuration as a starting point, do this instead of step 2:

  1. Merge your local ~/.vimrc into dotfiles/vim/vimrc

  2. Create a symlink to ~/.vim/vimrc:

    $ cd
    $ ln -s local/mydots/dotfiles/vim/vimrc .vimrc

You could actually have your .vimrc symlink point right to local/mydots/... I find this more flexible, though.


Additionally, if you want to use perlcritic.vim, you need to:

  • install Perl::Critic (from CPAN)

  • put bin/perlcritic_vim to your PATH

Instructions for git

Note: For most of config I have shared, you will need at least git 1.8. Without it, things will not work, and you might even end up with broken git!

However, if you only want to use the sharing/splitting mechanism, 1.7 should be OK. No warranties, though.

The pricnicple is the same as in bash section above: a new folder is created, local config is split into parts per user/host or global one, and original file becomes just a crossroads into these.

Except that since git does not support expansion in config paths, the file must be created by a script. So what you want to do is the same as for bash above, except that instead of applying patch to your .gitconfig, you simply delete it and run dotfiles/git/setup.

If you have a local changes that you don't want to share, you can either put them into the re-created ~.gitconfig, or (as I prefer) in dotfiles/git/private/gitconfig