I would like to build a script with getopts, that continues in the flag, when an $OPTARG isn't set. My script looks like this:
OPTIONS=':dBhmtb:P:'
while getopts $OPTIONS OPTION
do
case "$OPTION" in
m ) echo "m"
t ) echo "t"
d ) echo "d";;
h ) echo "h";;
B ) echo "b";;
r ) echo "r";;
b ) echo "b"
P ) echo hi;;
#continue here
\? ) echo "?";;
:) echo "test -$OPTARG requieres an argument" >&2
esac
done
My aim is to continue at my comment, when there is no $OPTARG set for -P.
All I get after running ./test -P is :
test -P requieres an argument
and then it continues after the loop but I want to continue in the -P flag.
All clear?
Any Ideas?
First, fix the missing ;;
in some of the case branches.
I don't think you can: you told getopts that -P
requires an argument: two error cases
-P
without an argument is the last option. In this case getops sees that nothing follows -P and sets the OPTION variable to :
, which you handle in the case statement.
-P
is followed by another option: getopts will simply take the next word, even if the next word is another option, as OPTARG.
Change the case branch to
P ) echo "P: '$OPTARG'";;
Then:
bash script.sh -P -m -t
, the output is
P: '-m'
t
bash script.sh -Pmt
, the output is
P: 'mt'
This is clearly difficult to work around. How do you know if the user intended the option argument to be literally "mt" and not the options -m and -t?
You might be able to work around this using getopt
(see the canonical example) using an optional argument for a long option (those require an equal sign like --long=value
) so it's maybe easier to check if the option argument is missing or not.
Translating getopts
parsing to getopt
-- it's more verbose, but you have finer-grained control
die() { echo "$*" >&2; exit 1; }
tmpArgs=$(getopt -o 'dBhmt' \
--long 'b::,P::' \
-n "$(basename "$0")" \
-- "$@"
)
(( $? == 0 )) || die 'Problem parsing options'
eval set -- "$tmpArgs"
while true; do
case "$1" in
-d) echo d; shift ;;
-B) echo B; shift ;;
-h) echo h; shift ;;
-m) echo m; shift ;;
-t) echo t; shift ;;
--P) case "$2" in
'') echo "P with no argument" ;;
*) echo "P: $2" ;;
esac
shift 2
;;
--b) case "$2" in
'') echo "b with no argument" ;;
*) echo "b: $2" ;;
esac
shift 2
;;
--) shift; break ;;
*) printf "> %q\n" "$@"
die 'getopt internal error: $*' ;;
esac
done
echo "Remaining arguments:"
for ((i=1; i<=$#; i++)); do
echo "$i: ${!i}"
done
Successfully invoking the program with --P
:
$ ./myscript.sh --P -mt foo bar
P with no argument
m
t
Remaining arguments:
1: foo
2: bar
$ ./myscript.sh --P=arg -mt foo bar
P: arg
m
t
Remaining arguments:
1: foo
2: bar
This does impose higher overhead on your users, because -P
(with one dash) is invalid, and the argument must be given with =
$ ./myscript.sh --P arg -mt foo bar
P with no argument
m
t
Remaining arguments:
1: arg
2: foo
3: bar
$ ./myscript.sh --Parg mt foo bar
myscript.sh: unrecognized option `--Parg'
Problem parsing options
$ ./myscript.sh -P -mt foo bar
myscript.sh: invalid option -- P
Problem parsing options
$ ./myscript.sh -P=arg -mt foo bar
myscript.sh: invalid option -- P
myscript.sh: invalid option -- =
myscript.sh: invalid option -- a
myscript.sh: invalid option -- r
myscript.sh: invalid option -- g
Problem parsing options