I have some shell functions that I use for personal development. I would like to create config files on a per-project basis so that the functions read the config file and use values from the config file, if present.

This is a very basic example:

path/to/project/functions.sh (works)

#!/usr/bin/env bash

PROJECT_KEY="abc"

function project_name() {
  if [ -d ".config" ]; then
    . .config
  fi

  echo $PROJECT_KEY # => "def"
}

path/to/project/.config

PROJECT_KEY="def"

The above works, but I have dozens of these utility/helper functions, and I wanted to load the config via another function if it exists.

path/to/project/functions.sh (does not work)

#!/usr/bin/env bash
function load_config() {
  if [ -f ".config" ]; then
    . ".config"
  elif [ -f "~/.config" ]; then
    . "~/.config"
  fi
}

function project_name() {
  load_config

  echo $PROJECT_KEY # => "abc"
}

I believe that the load_config() function is sourcing the config file into the load_config() function, and then when that function returns/exits, the sourced variables are gone because they are scoped to the individual function call, ie, the load_config() function.

I understand my current approach isn’t working, but I don’t know how to avoid executing source .config all over the place. I want to put it in a single function that is called.

How do I cleanly read config files in my functions without a ton of duplicate code?