How to split string by spaces, which contains quoted values

Question

Take the following bash example (I should mention that my real output comes from vboxmanage list vms)

TEST='"ubuntu server" foo bar'
echo $TEST
"ubuntu server" foo bar

How do I get the three values separated (in a loop or separate variable)? Every solution I've found relies on space delimiter and will break the first value.

Answer 1

You may iterate over the names of the VirtualBox machines:

vboxmanage list -l vms | sed -n 's/^Name:[[:blank:]]*//p' |
while IFS= read -r vm_name; do
   printf 'There is one VM called "%s"\n' "$vm_name"
done

This lists the machines using the long format and parses out the names. The names are then read in a shell loop and a short message is printed for each.

You may also read the names into an array:

readarray -t vm_names < <(vboxmanage list -l vms | sed -n 's/^Name:[[:blank:]]*//p')
printf 'There is one VM called "%s"\n' "${vm_names[@]}"

Here, the list of names are red into the array vm_names. You may loop over this array like so:

for vm_name in "${vm_names[@]}"; do
   # Use "$vm_name" here
done

To also use the UUID:

vboxmanage list -l vms |
sed -n \
    -e '/^Name:[[:blank:]]*/{ s///; h; }' \
    -e '/^UUID:[[:blank:]]*/{ s///; G; y/\n/@/; p; }' |
while IFS=@ read -r vm_uuid vm_name; do
    printf 'VM "%s" has UUID "%s"\n' "$vm_name" "$vm_uuid"
done

Answer 2

Assuming you have it just as a string, in the format shown in the code block in your initial question (i.e. "ubuntu server" foo bar), and the data is known to be sufficiently nice, i.e. only quoted strings (no shell special characters), you could use eval to evaluate it is as part of a shell command:

str='"ubuntu server" foo bar'
eval "arr=($str)"

With that input, this would give the array arr with the three elements ubuntu server, foo, and bar. You can use the array in the usual ways, e.g. "${arr[@]}" to expand to a list of words, one for each element.

But note that using eval involves evaluating the value for all shell syntax, not just quotes. E.g. if the string contains a command substitution $(cmd...), that command will run. If the string contains the unquoted );, it'll end the compound assignment statement, and what comes after will run as a shell command.

To be on the safe side, validate the input first, e.g. by checking it only contains quotes and safe characters. E.g. this would allow letters, numbers, blanks, the underscore and the two types of quotes, all of which should be safe here. The input could still have unmatching quotes, which would cause an error on the eval.

re='[^[:alnum:][:blank:]_'\''"]'
if [[ "$str" =~ $re ]]; then
    echo "unsafe input" >&2
else
    eval "arr=($str)";
    if [[ $? -ne 0 ]]; then
        echo "error in assignment" >&2
    fi
fi

Answer 3

You can do word splitting with the set command. Here is a clumsy attempt but you may be able to improve on it.

First set the delimiter to double quote mark:

IFS='"'

then

set -- $TEST
echo $1

echo $2
ubuntu server
echo $3
 foo bar

You have the first (2-word) variable in $2 and the rest in $3

a=$2
b=$3

Now do another operation on variable b but this time with space delimiter

IFS=' '
set -- $b
echo $1
foo
echo $2
bar