Once again I try to get a handle of my various dotfiles and configs. This time I take another stab at gnu stow as it is often recommended. I do not understand it.
Here’s how I understand it: I’m supposed to manually move all my files into a new directory where the original are. So for ~ I make like this:
~
- dotfiles
- bash
dot-bashrc
dot-bash_profile
- xdg
- dot-config
user-dirs.dirs
- tealdeer
- dot-config
- tealdeer
config.toml
then cd ~/dotfiles && stow --dotfiles .
Then (if I very carefully created each directory tree) it will symlink those files back to where they came from like this:
~
.bashrc
.bash_profile
- .config
user-dirs.dirs
- tealdeer
config.toml
I don’t really understand what this application is doing because setting up the dotfiles directory is a lot more work than making symlinks afterwards. Every instructions tells me to make up this directory structure by hand but that seems to tedious with so many configs; isn’t there some kind of automation to it?
Once the symlinks are created then what?
-
Tutorials don’t really mention it but the actual manual gives me the impression this is a packager manager in some way and that’s confusing. Lots of stuff about compiling
-
I see about how to combine it with
git. Triedgit-oriented dotfile systems before and they just aren’t practical for me. And again I don’t see whatstowcontributing;gitwould be doing all the work there. -
Is there anything here about sharing configs between non-identical devices? Not everything can be copy/pasted exactly. Are you supposed to be making
gitbranches or something?
The manual is not gentle enough to learn from scratch. OTOH there are very very short tutorials which offer little information.
I feel that I’m really missing the magic that’s obvious to everyone else.
There are other dotfile managers out there, personally I use yadm. It’s just a shell wrapper around a bare git repository, no symlinks or other magic needed
yadm is the one I liked the best and tried it a few times. fact is that I am unlikely to keep a repo like this even part way up to date. New files are created all the time and not added, old ones don’t get updated or removed. There’s not even a good way to notice in any file manager what is included and what’s not as far as I know. yadm doesn’t work with tools like eza which can display the git status of files in repos. (and it probably wouldn’t be feasible.)
Plus I have some specific config collections already in change tracking and it makes more sense to keep it that way. Having so many unrelated files together in one project is too chaotic and distracting.
It’s not realistic for me to manage merges, modules, cherry picking, branches all that for so many files that change constantly without direct intervention. Quickly enough git will tie itself into some knot and I won’t be able to pick it apart.
personally i always found stow to be annoying to use and switched to a bare git-repo approach. you can read up on it over here: https://www.atlassian.com/git/tutorials/dotfiles. Like this your dotfiles just rest where they should and it’s rather minimal overhead.
That’s… all stow does, there’s nothing more to it. If you need some other feature don’t waste your time trying to make it work with stow, It’s just a meme in my opinion.
About the “package manager” functionality, stow was originally supposed to be a development tool for the Perl programming language, you download a bunch of libraries into a directory, then use stow to merge those files into the root of your project (like a caveman), as it turned out some people started using it to manage dotfiles, and here we are.
When I started trying to organize my dotfiles, I started with stow, but quickly found it very limited.
After that I found dotdrop, which is considerably more involved, but gives you total control. My config with dotdrop quickly started growing insanely huge, at some point I even had system-wide systemd services declared.
Then I found out I was basically reinventing nixos and home-manager, so I switched to that.
Recreating a path to original locations is a way to configure stow.
You can utilize bash to automate it. Mkdir can create entire trees in a single command.
After links were created you start using your configured software.
It’s a link farm built using a packaging system. You put your configs into a “package” and then link said package where it belongs.
Git is useful not as a combination for stow, but a standalone way to version control your configs.
For nonidentical devices you create additional packages prefixed with specific device name. You don’t need to link all packages at once with stow, pass a name of a package to link it alone.
For nonidentical devices you create additional packages prefixed with specific device name. You don’t need to link all packages at once with stow, pass a name of a package to link it alone.uuu
Sooo… I find some way to share the dotfiles directory across devices (rsync, syncthing, git, nextcloud, DAV) then make specific subdirs like this?:
~
- dotfiles
- bash-desktop
dot-bashrc
dot-bash_profile
- bash-laptop
dot-bashrc
dot-profile
dot-bash_profile
But what is the software doing for me? I’m manually moving all these files and putting them together in the specific way requested. Setting the whole thing up is most of the work. Anyone who can write a script to create the structure can just as easily write it to make symlinks. I’m sure I’m missing something here.
Sorry if it comes out robotic, since I don’t use stow myself I just opened the manual and parsed it for you.
You manually create the repo of packages ONCE and then use stow to deploy them without having to link stuff manually every time.
As for “what’s it doing for me”, stow is just a generic deployment script that somebody else has written for you. You can just as well create your own and do it your way.
I personally used to have a bash script with one custom function “symlink” that error checked the linking process and a list of target/destination under the function.
Now I’m trying to code a python script similar to stow that works with packages of configs but instead of recreating paths inside the package you just provide a json file with target/destination for each file in the package.
Both of my ways have same shortcoming as stow: you have to do some manual work before the script can kick in.
Hopefully my messages have been helpful.
stow works great then lazy git to store the charges in a personal git server been using it like that for more than a decade
