Retrieving multiple arguments for a single option using getopts in Bash

I need help with getopts.

I created a Bash script which looks like this when run:

$ foo.sh -i env -d directory -s subdirectory -f file

It works correctly when handling one argument from each flag. But when I invoke several arguments from each flag I am not sure how to pull the multiple variable information out of the variables in getopts.

while getopts ":i:d:s:f:" opt
   do
     case $opt in
        i ) initial=$OPTARG;;
        d ) dir=$OPTARG;;
        s ) sub=$OPTARG;;
        f ) files=$OPTARG;;
    
     esac
done

After grabbing the options I then want to build directory structures from the variables

foo.sh -i test -d directory -s subdirectory -s subdirectory2 -f file1 file2 file3

Then the directory structure would be

/test/directory/subdirectory/file1
/test/directory/subdirectory/file2
/test/directory/subdirectory/file3
/test/directory/subdirectory2/file1
/test/directory/subdirectory2/file2
/test/directory/subdirectory2/file3

Any ideas?

You can use the same option multiple times and add all values to an array.

For the very specific original question here, Ryan's mkdir -p solution is obviously the best.

However, for the more general question of getting multiple values from the same option with getopts, here it is:

#!/bin/bash

while getopts "m:" opt; do
    case $opt in
        m) multi+=("$OPTARG");;
        #...
    esac
done
shift $((OPTIND -1))

echo "The first value of the array 'multi' is '$multi'"
echo "The whole list of values is '${multi[@]}'"

echo "Or:"

for val in "${multi[@]}"; do
    echo " - $val"
done

The output would be:

$ /tmp/t
The first value of the array 'multi' is ''
The whole list of values is ''
Or:

$ /tmp/t -m "one arg with spaces"
The first value of the array 'multi' is 'one arg with spaces'
The whole list of values is 'one arg with spaces'
Or:
 - one arg with spaces

$ /tmp/t -m one -m "second argument" -m three
The first value of the array 'multi' is 'one'
The whole list of values is 'one second argument three'
Or:
 - one
 - second argument
 - three

I know this question is old, but I wanted to throw this answer on here in case someone comes looking for an answer.

Shells like BASH support making directories recursively like this already, so a script isn't really needed. For instance, the original poster wanted something like:

$ foo.sh -i test -d directory -s subdirectory -s subdirectory2 -f file1 file2 file3
/test/directory/subdirectory/file1
/test/directory/subdirectory/file2
/test/directory/subdirectory/file3
/test/directory/subdirectory2/file1
/test/directory/subdirectory2/file2
/test/directory/subdirectory2/file3

This is easily done with this command line:

pong:~/tmp
[10] rmclean$ mkdir -pv test/directory/{subdirectory,subdirectory2}/{file1,file2,file3}
mkdir: created directory ‘test’
mkdir: created directory ‘test/directory’
mkdir: created directory ‘test/directory/subdirectory’
mkdir: created directory ‘test/directory/subdirectory/file1’
mkdir: created directory ‘test/directory/subdirectory/file2’
mkdir: created directory ‘test/directory/subdirectory/file3’
mkdir: created directory ‘test/directory/subdirectory2’
mkdir: created directory ‘test/directory/subdirectory2/file1’
mkdir: created directory ‘test/directory/subdirectory2/file2’
mkdir: created directory ‘test/directory/subdirectory2/file3’

Or even a bit shorter:

pong:~/tmp
[12] rmclean$ mkdir -pv test/directory/{subdirectory,subdirectory2}/file{1,2,3}
mkdir: created directory ‘test’
mkdir: created directory ‘test/directory’
mkdir: created directory ‘test/directory/subdirectory’
mkdir: created directory ‘test/directory/subdirectory/file1’
mkdir: created directory ‘test/directory/subdirectory/file2’
mkdir: created directory ‘test/directory/subdirectory/file3’
mkdir: created directory ‘test/directory/subdirectory2’
mkdir: created directory ‘test/directory/subdirectory2/file1’
mkdir: created directory ‘test/directory/subdirectory2/file2’
mkdir: created directory ‘test/directory/subdirectory2/file3’

Or shorter, with more conformity:

pong:~/tmp
[14] rmclean$ mkdir -pv test/directory/subdirectory{1,2}/file{1,2,3}
mkdir: created directory ‘test’
mkdir: created directory ‘test/directory’
mkdir: created directory ‘test/directory/subdirectory1’
mkdir: created directory ‘test/directory/subdirectory1/file1’
mkdir: created directory ‘test/directory/subdirectory1/file2’
mkdir: created directory ‘test/directory/subdirectory1/file3’
mkdir: created directory ‘test/directory/subdirectory2’
mkdir: created directory ‘test/directory/subdirectory2/file1’
mkdir: created directory ‘test/directory/subdirectory2/file2’
mkdir: created directory ‘test/directory/subdirectory2/file3’

Or lastly, using sequences:

pong:~/tmp
[16] rmclean$ mkdir -pv test/directory/subdirectory{1..2}/file{1..3}
mkdir: created directory ‘test’
mkdir: created directory ‘test/directory’
mkdir: created directory ‘test/directory/subdirectory1’
mkdir: created directory ‘test/directory/subdirectory1/file1’
mkdir: created directory ‘test/directory/subdirectory1/file2’
mkdir: created directory ‘test/directory/subdirectory1/file3’
mkdir: created directory ‘test/directory/subdirectory2’
mkdir: created directory ‘test/directory/subdirectory2/file1’
mkdir: created directory ‘test/directory/subdirectory2/file2’
mkdir: created directory ‘test/directory/subdirectory2/file3’

getopts options can only take zero or one argument. You might want to change your interface to remove the -f option, and just iterate over the remaining non-option arguments

usage: foo.sh -i end -d dir -s subdir file [...]

So,

while getopts ":i:d:s:" opt; do
  case "$opt" in
    i) initial=$OPTARG ;;
    d) dir=$OPTARG ;;
    s) sub=$OPTARG ;;
  esac
done
shift $(( OPTIND - 1 ))

path="/$initial/$dir/$sub"
mkdir -p "$path"

for file in "$@"; do
  touch "$path/$file"
done

I fixed the same problem you had like this:

Instead of:

foo.sh -i test -d directory -s subdirectory -s subdirectory2 -f file1 file2 file3

Do this:

foo.sh -i test -d directory -s "subdirectory subdirectory2" -f "file1 file2 file3"

With the space separator you can just run through it with a basic loop. Here's the code:

while getopts ":i:d:s:f:" opt
   do
     case $opt in
        i ) initial=$OPTARG;;
        d ) dir=$OPTARG;;
        s ) sub=$OPTARG;;
        f ) files=$OPTARG;;

     esac
done

for subdir in $sub;do
   for file in $files;do
      echo $subdir/$file
   done   
done

Here's a sample output:

$ ./getopts.sh -s "testdir1 testdir2" -f "file1 file2 file3"
testdir1/file1
testdir1/file2
testdir1/file3
testdir2/file1
testdir2/file2
testdir2/file3

If you want to specify any number of values for an option, you can use a simple loop to find them and stuff them into an array. For example, let's modify the OP's example to allow any number of -s parameters:

unset -v sub
while getopts ":i:d:s:f:" opt
   do
     case $opt in
        i ) initial=$OPTARG;;
        d ) dir=$OPTARG;;
        s ) sub=("$OPTARG")
            until [[ $(eval "echo \${$OPTIND}") =~ ^-.* ]] || [ -z $(eval "echo \${$OPTIND}") ]; do
                sub+=($(eval "echo \${$OPTIND}"))
                OPTIND=$((OPTIND + 1))
            done
            ;;
        f ) files=$OPTARG;;
     esac
done

This takes the first argument ($OPTARG) and puts it into the array $sub. Then it will continue searching through the remaining parameters until it either hits another dashed parameter OR there are no more arguments to evaluate. If it finds more parameters that aren't a dashed parameter, it adds it to the $sub array and bumps up the $OPTIND variable.

So in the OP's example, the following could be run:

foo.sh -i test -d directory -s subdirectory1 subdirectory2 -f file1

If we added these lines to the script to demonstrate:

echo ${sub[@]}
echo ${sub[1]}
echo $files

The output would be:

subdirectory1 subdirectory2
subdirectory2
file1

There actually is a way to retrieve multiple arguments using getopts, but it requires some manual hacking with getopts' OPTIND variable.

See the following script (reproduced below): https://gist.github.com/achalddave/290f7fcad89a0d7c3719. There's probably an easier way, but this was the quickest way I could find.

#!/bin/sh

usage() {
cat << EOF
$0 -a <a1> <a2> <a3> [-b] <b1> [-c]
    -a      First flag; takes in 3 arguments
    -b      Second flag; takes in 1 argument
    -c      Third flag; takes in no arguments
EOF
}

is_flag() {
    # Check if $1 is a flag; e.g. "-b"
    [[ "$1" =~ -.* ]] && return 0 || return 1
}

# Note:
# For a, we fool getopts into thinking a doesn't take in an argument
# For b, we can just use getopts normal behavior to take in an argument
while getopts "ab:c" opt ; do
    case "${opt}" in
        a)
            # This is the tricky part.

            # $OPTIND has the index of the _next_ parameter; so "\${$((OPTIND))}"
            # will give us, e.g., ${2}. Use eval to get the value in ${2}.
            # The {} are needed in general for the possible case of multiple digits.

            eval "a1=\${$((OPTIND))}"
            eval "a2=\${$((OPTIND+1))}"
            eval "a3=\${$((OPTIND+2))}"

            # Note: We need to check that we're still in bounds, and that
            # a1,a2,a3 aren't flags. e.g.
            #   ./getopts-multiple.sh -a 1 2 -b
            # should error, and not set a3 to be -b.
            if [ $((OPTIND+2)) -gt $# ] || is_flag "$a1" || is_flag "$a2" || is_flag "$a3"
            then
                usage
                echo
                echo "-a requires 3 arguments!"
                exit
            fi

            echo "-a has arguments $a1, $a2, $a3"

            # "shift" getopts' index
            OPTIND=$((OPTIND+3))
            ;;
        b)
            # Can get the argument from getopts directly
            echo "-b has argument $OPTARG"
            ;;
        c)
            # No arguments, life goes on
            echo "-c"
            ;;
    esac
done

The original question deals with getopts, but there is another solution that provides more flexible functionality without getopts (this is perhaps a bit more verbose, but provides a far more flexible command line interface). Here is an example:

while [[ $# > 0 ]]
do
    key="$1"
    case $key in
        -f|--foo)
            nextArg="$2"
            while ! [[ "$nextArg" =~ -.* ]] && [[ $# > 1 ]]; do
                case $nextArg in
                    bar)
                        echo "--foo bar found!"
                    ;;
                    baz)
                        echo "--foo baz found!"
                    ;;
                    *)
                        echo "$key $nextArg found!"
                    ;;
                esac
                if ! [[ "$2" =~ -.* ]]; then
                    shift
                    nextArg="$2"
                else
                    shift
                    break
                fi
            done
        ;;
        -b|--bar)
            nextArg="$2"
            while ! [[ "$nextArg" =~ -.* ]] && [[ $# > 1 ]]; do
                case $nextArg in
                    foo)
                        echo "--bar foo found!"
                    ;;
                    baz)
                        echo "--bar baz found!"
                    ;;
                    *)
                        echo "$key $nextArg found!"
                    ;;
                esac
                if ! [[ "$2" =~ -.* ]]; then
                    shift
                    nextArg="$2"
                else
                    shift
                    break
                fi
            done
        ;;
        -z|--baz)
            nextArg="$2"
            while ! [[ "$nextArg" =~ -.* ]] && [[ $# > 1 ]]; do

                echo "Doing some random task with $key $nextArg"

                if ! [[ "$2" =~ -.* ]]; then
                    shift
                    nextArg="$2"
                else
                    shift
                    break
                fi
            done
        ;;
        *)
            echo "Unknown flag $key"
        ;;
    esac
    shift
done

In this example we are looping through all of the command line options looking for parameters that match our accepted command line flags (such as -f or --foo). Once we find a flag, we loop through every parameter until we run out of parameters or encounter another flag. This breaks us back out into our outer loop which only processes flags.

With this setup, the following commands are equivalent:

script -f foo bar baz
script -f foo -f bar -f baz
script --foo foo -f bar baz
script --foo foo bar -f baz

You can also parse incredibly disorganized parameter sets such as:

script -f baz derp --baz herp -z derp -b foo --foo bar -q llama --bar fight

To get the output:

--foo baz found!
-f derp found!
Doing some random task with --baz herp
Doing some random task with -z derp
--bar foo found!
--foo bar found!
Unknown flag -q
Unknown flag llama
--bar fight found!

#!/bin/bash
myname=$(basename "$0")

# help function
help () { cat <<EOP
   $myname: -c cluster [...] -a action [...] -i instance [...]
EOP
}

# parse sub options
get_opts () {
  rs='' && rc=0 # return string and return code
  while [[ $# -gt 0 ]]; do
    shift
    [[ "$1" =~ -.* ]] && break ||  rs="$rs $1" && rc=$((rc + 1))
  done
  echo "$rs"
}

#parse entire command-line
while [[ $# -gt 0 ]]; do
    case $1 in
        "-a") ACTS="$(get_opts $@)"
           ;;
        "-i") INSTS=$(get_opts $@)
           ;;
        "-c") CLUSTERS=$(get_opts $@)
           ;;
        "-h") help
           ;;
        ?) echo "sorry, I dont do $1"
           exit
           ;;
    esac
    shift
done

The following link should be a general solution for this requirement.

It's easy to inject and clear enough to understand also minimizes the impact on the original code.

https://stackoverflow.com/questions/16186392/multiple-option-arguments-using-getopts-bash/66731653#66731653

function getopts-extra () {
    declare i=1
    # if the next argument is not an option, then append it to array OPTARG
    while [[ ${OPTIND} -le $# && ${!OPTIND:0:1} != '-' ]]; do
        OPTARG[i]=${!OPTIND}
        let i++ OPTIND++
    done
}

# Use it within the context of `getopts`:
while getopts s: opt; do
    case $opt in
       s) getopts-extra "$@"
          args=( "${OPTARG[@]}" )
    esac
done

As you don't show how you hope to construct your list

/test/directory/subdirectory/file1
. . .
test/directory/subdirectory2/file3

it's a little unclear how to proceed, but basically you need to keep appending any new values to the appropriate variable, i.e.

 case $opt in
    d ) dirList="${dirList} $OPTARG" ;;
 esac

Note that on the first pass dir will be empty, and you'll wind up with a space leading at the from of your final value for ${dirList}. (If you really need code that doesn't include any extra spaces, front or back, there is a command I can show you, but it will be hard to understand, and it doesn't seem that you'll need it here, but let me know)

You can then wrap your list variables in for loops to emit all the values, i.e.

for dir in ${dirList} do
   for f in ${fileList} ; do
      echo $dir/$f
   done
done

Finally, it is considered good practice to 'trap' any unknown inputs to your case statement, i.e.

 case $opt in
    i ) initial=$OPTARG;;
    d ) dir=$OPTARG;;
    s ) sub=$OPTARG;;
    f ) files=$OPTARG;;
    * ) 
       printf "unknown flag supplied "${OPTARG}\nUsageMessageGoesHere\n"
       exit 1
    ;;

 esac 

I hope this helps.

This is a simple way to pass multiple arguments for a single option.

#!/bin/bash

#test.sh -i 'input1 input2'
#OR
#test.sh -i 'input*'

while getopts "i:" opt; do
    case $opt in
        i ) input=$OPTARG;;
    esac
done
inputs=( $input )

echo "First input is "$inputs""
echo "Second input is "${inputs[1]}""
echo "All inputs: "${inputs[@]}""