dannyman.toldme.com : Argument Parsing in Shell Scripts

June 22, 2005
Technical

Argument Parsing in Shell Scripts

So, say you are writing a shell script (sh or bash) that needs to take arguments like so:

./script.sh start
./script.sh -v start
./script.sh -c foo.conf start
./script.sh -vc foo.conf start

This took me a bit of doing. First, I tried getopt but apparently all that does is disambiguate -v from -c. That, and the FreeBSD man page claims to have improved the example by making it completely inscrutable.

Eventually I found a reference that explained how to use getopts, which worked a lot better, but all the examples I could find didn’t tell you how to reset $@.

Eventually, I figured out that that last bit is accomplished with shift.

Here’s example code:

#!/bin/sh
#

conf=default.conf
while getopts c:hv o
do case "$o" in
    c)  conf="$OPTARG";;
    h)  echo "Usage: $0 -hv -c <config> [start|stop|restart|status]" && exit 1;;
    v)  verbose="yes";;
    \?)  echo "Usage: $0 -hv -c <config> [start|stop|restart|status]" && exit 1;;
esac
done

# Reset $@
shift `echo $OPTIND-1 | bc`

echo "conf: $conf"
echo "command: $1"

Notes:

The while loop iterates $o through $@ where it conforms to the getopts configuration.
getopts c:hv indicates that -h and -v are boolean arguments, and -c is followed by a string argument.
$OPTIND is set at the end of getopt, and indicates how many arguments through $@ it parsed.
Oh, and the shift operation resets $@ by $OPTIND-1, so that $1 ends up pointing at the start command, or whatever.
If you’re programming in ksh, you can just set $OPTIND-1. But if you’re programming in ksh you are far beyond needing to check my blog for argument parsing.

Responses

June 22nd, 2005

Andrew Ho

My standard sh command line parsing loop looks like this:

    while [ $# != 0 ]; do
        flag="$1"
        case "$flag" in
            -a) echo "You provided the -a flag, which takes no arguments"
                ;;
            -b) if [ $# -gt 1 ]; then
                    arg="$2"
                    shift
                else
                    echo "You did not provide an argument for the -b flag"
                fi
                echo "You supplied an argument for the -b flag: $arg"
                ;;
             *) echo "Unrecognized flag or argument: $flag"
                ;;
        esac
        shift
    done

If you are using bash instead of sh, you can accept both -fvalue and -f value variants:

           -b*) arg="${arg##-b}"
                echo "You attached an argument to the -b flag: $arg"
                ;;

My code doesn’t reset $@. I have a README file in my homedir that reminds me of the other neat bash-ism that I like to use: ${1+"$@"} expands to $@, but if $@ is empty, it expands to an empty list (different from “$@” which would expand to the single empty string “”).

June 24th, 2005

an_agent

FAI…

shift $((OPTIND - 1))

works in bash (not sh) without having to pipe to bc.

January 15th, 2008

Ted Henry

Thanks for this post! This is exactly what I needed :-)

Post a Public Comment

Name:		(required)
E-Mail Address:		(required)
Website URL:

dannyman.toldme.com