Back in the days when I was developing in Ruby most of my waking hours rbenv was a real life saver. But looking at what it does, it initially felt awkward how it "wrapped" the cd command. You could argue that the Ruby community are no strangers to guerrilla patching. But it felt less awkward when I learned that in my zsh it uses hooks instead of guerrilla patching to achieve the same goal: React on changing the directory.

In the meantime I tried a couple of things with zsh hooks to optimize work flows and what not, but as it turns out, everything boiled down to one hook, and one hook only:

autoload -U add-zsh-hook

# source .sourceme files
load-local-conf() {
  # check file exists, is regular file and is readable:
  if [[ -f .sourceme && -r .sourceme ]]; then
  	source .sourceme
  fi
}
add-zsh-hook chpwd load-local-conf

The basic idea is: Directories naturally act like contexts, projects are structured in directories, the different concerns within a project could be structured by directories, and so on. If you store contextual functions, aliases, variables, or documentation - whatever you need on the console in the given context - in .sourceme files, the hook above will automatically source the file once you enter that directory and you will have everything in there available.

Here is a (rather contrived, but never the less pretty clear) example: Let's say you have two projects that have images in a images directory. As new images come in you find yourself repeatedly call ImageMagick's mogrify command to reduce the images to a given size and each of the projects requires a specific image size. In the project's .sourceme you could define an alias

alias resize-images='mogrify --geometry 100x images/*'

There is your first contextual helper.

Other examples are:

• Running docker ps to make sure the required containers are running as I start working on a project.
• Running cat <<EOF to display some project specific commands that I have trouble to remember.
• Setting environment variables with credentials.