cmdline.sh

#!/bin/bash
set -e

PROGNAME=$(basename $0)

die() {
    echo "$PROGNAME: $*" >&2
    exit 1
}

usage() {
    if [ "$*" != "" ] ; then
        echo "Error: $*"
    fi

    cat << EOF
Usage: $PROGNAME [OPTION ...] [foo] [bar]
<Program description>.

Options:
  -h, --help          display this usage message and exit
  -d, --delete        delete things
  -o, --output [FILE] write output to file
EOF

    exit 1
}

foo=""
bar=""
delete=0
output="-"
args=()
while [ $# -gt 0 ] ; do
    case "$1" in
    -h|--help)
        usage
        ;;
    -d|--delete)
        delete=1
        ;;
    -o|--output)
        output="$2"
        shift
        ;;
    -*)
        usage "Unknown option '$1'"
        ;;
    *)
        args=("${args[@]}" "$1")
        ;;
    esac
    shift
done

cat <<EOF
foo=$foo
bar=$bar
delete=$delete
output=$output
EOF

for arg in "${args[@]}" ; do
    echo "'$arg'"
done

Nice template! I added an extra case clause to support using an optional equal sign to pass arguments to a flag. For example, to treat cmdline.sh --output=somefile and cmdline.sh -o=somefile exactly the same as cmdline.sh --output somefile.

When it encounters what looks like a flag containing an equals sign, it just just splits it into the flag and the argument, shifts it off and shifts the separate parts back on again and continues looping. It's a bit hacky, but it seems to work. I tend to aim for sh compatibility, but if one was okay with depending on bash then the sed invocations could probably be replaced by fancy variable expansion magic.