Dotfiles Refinement: Cleaner, Smarter, Faster

A Year Later, A Simpler Setup

Since my last update in 2015, I’ve had time to live with the modular dotfiles system. It worked. Onboarding was faster, the install scripts were smarter, and I had shell support for both Bash and Zsh.

But even well-organized setups can start to rot. Too many aliases, too many .zshrc conditionals, and a few copy-pasted hacks started to creep in. By late 2016, I felt the need to simplify.

So this wasn’t a rebuild. It was a focused refinement. Fewer features, clearer structure, and a setup I could actually understand six months later.

This post documents how I tightened the screws.

From Organized to Loadable

My 2015 setup was topic-based, with folders like bash/, zsh/, git/, and an install script to symlink everything. But each shell file still sourced other files explicitly:

source "$DOTFILES/zsh/aliases.zsh"
source "$DOTFILES/zsh/functions.zsh"

Now, I’ve introduced a loader pattern. In zsh/load.zsh:

for config_file ($ZSH/*.zsh); do
  source $config_file
done

This small shift makes everything cleaner. Adding new functionality is just dropping a .zsh file in the folder. No need to edit multiple files to register it.

It feels like dependency injection, but for your shell.

Smaller, More Focused Functions

The biggest win from this cycle was cutting. I rewrote most of my shell functions:

• extract for decompressing archives
• take for creating and entering directories
• path-append for safely editing $PATH

Each function lives in its own file under zsh/functions/. This means I can test, replace, or remove one without affecting the rest.

# extract.zsh
extract() {
  case $1 in
    *.tar.bz2)   tar xjf $1     ;;
    *.tar.gz)    tar xzf $1     ;;
    *.bz2)       bunzip2 $1     ;;
    *.rar)       unrar x $1     ;;
    *.gz)        gunzip $1      ;;
    *.tar)       tar xf $1      ;;
    *.tbz2)      tar xjf $1     ;;
    *.tgz)       tar xzf $1     ;;
    *.zip)       unzip $1       ;;
    *.Z)         uncompress $1  ;;
    *)           echo "don't know how to extract '$1'..." ;;
  esac
}

Writing functions as standalone files was the change I didn’t know I needed.

A Prompt That Gets Out of the Way

I used to tinker with the prompt a lot. In 2015, I had a verbose, colored PS1 setup showing time, path, Git branch, and exit code. It was useful — but noisy.

Now I use a much simpler prompt. Just the essentials:

autoload -Uz vcs_info
precmd() { vcs_info }
PROMPT='%n@%m %1~ ${vcs_info_msg_0_}%# '

This prompt:

• Uses vcs_info to show Git state when needed
• Shows user, host, and current directory
• Keeps the screen focused on the command, not decorations

More signal, less noise.

Install Script Now Even Dumber (in a Good Way)

The original install.sh grew too smart. It had conditionals, dependency checks, platform switches. Now it’s dumber — but better.

find * -name "*.symlink" | while read file; do
  ln -sf "$file" "$HOME/.$(basename "$file" ".symlink")"
done

I moved platform-specific setups elsewhere. This lets install.sh do just one job: create symlinks. It’s composable now.

This script now runs the same on every machine. Predictability > cleverness.

What I’ve Learned

This update wasn’t about adding — it was about subtracting. Every improvement came from asking: what can I remove?

Instead of building a framework, I’ve built a system that:

• Has no magic
• Is easy to onboard into
• Can grow without friction
• Looks boring (and that’s a feature)

I’m not chasing perfection. Just stability and sanity.

→ Check out the diff on GitHub