Always run a shell script from the directory it lives in

Always run a shell script in the directory in which it appears, and change back to the directory you were in when you ran it even if it fails.

trap popd EXIT
pushd $PWD
cd $(dirname "$0")

Works for me in bash. The pushd command does a cd but saves the directory where you were on a stack, and popd pops the saved directory from the stack. The trap ... EXIT is a bash way to run something when the script exits, no matter how, and dirname "$0" is the directory name of the script.

(Taken from the deploy.sh script that rebuilds and deploys this blog, so if you can read this, it works.)