You can extract the strings between brackets and then arrange after you have them. For example;

• Using awk:

  echo "$logs" | awk -F'[][]' '{print $2}'

• Using sed:

  echo "$logs" | sed -n 's/.*[([^]]*)].*/1/p'

• Using grep -oP:

  echo "$logs" | grep -oP '(?<=[)[^]]*(?=])'

They will only return “apple, pineapple, wood-apple”. If it is a necessity, you can add/concatenate/append the brackets after you retrieve them.

Edit: Based on the changes of the poster’s edit and comment, I added the new solution as follows.

Your input is $logs can be processed such as:

• Using awk:

echo $logs | awk '/[/{flag=1; next} /]/{flag=0} flag {printf "%s", $0}'

/[/{flag=1; next} this section sets a flag when [ is encountered. /]/{flag=0} this section unsets the flag when ] is encountered. Final part prints lines while the flag is set, effectively concatenating the multiline input.

• Using sed:

echo $logs | sed -n '/[/,/]/p' | sed ':a;N;$!ba;s/n//g' | sed 's/.*[(.*)].*/1/'

You have to use sed multiple times. First one finds the lines between the [ and ]. Second one joins all lines together, and the last one extracts the content between [ and ].

• Using grep and tr:

echo $logs | grep -oP '(?<=[)[sS]*(?=])' | tr -d 'n'

grep section matches all characters between [ and ] (including newlines) and tr section removes newline characters to create a single line.

burcu’s answer is perfect. As I prefer using the if statement in awk , here is my answer:

awk '{
    if ($0 == "[") {
        in_bracket = 1
        next
    }else if ($0 == "]"){
        in_bracket = 0
    }
    if (in_bracket) {
        print $0
    }
}' a.txt

At least for your value of LOGS (you didn’t specify the general format of the text), the following should work in bash:

LOGS='[apple, pineapple, wood-apple]'
echo ${LOGS//[][]/}

The // causes all occurances of the pattern to be removed. The pattern is a wildcard (i.e. glob) pattern, in this case a character class [ ... ], containing the two characters ][.

$ echo "$logs" | grep -o '[.*]'
[apple, pineapple, wood-apple]

$ echo "$logs" | sed -n 's/.*([.*]).*/1/p'
[apple, pineapple, wood-apple]

$ echo "$logs" | awk -F'[][]' '{print "[" $2 "]"}'
[apple, pineapple, wood-apple]

$ echo "$logs" | awk 'match($0,/[.*]/){print substr($0,RSTART,RLENGTH)}'
[apple, pineapple, wood-apple]

$ echo "$logs" | gawk 'match($0,/[.*]/,a){print a[0]}'
[apple, pineapple, wood-apple]

$ echo "$logs" | gawk '{print gensub(/.*([.*]).*/, "\1", 1)}'
[apple, pineapple, wood-apple]

$ [[ "$logs" =~ [.*] ]] && echo "${BASH_REMATCH[0]}"
[apple, pineapple, wood-apple]

The grep command requires the non-POSIX -o extension as provided by GNU grep, the final 2 awk commands require GNU awk for the non-POSIX extensions of gensub() and the 3rd argument to match(). Everything else will work in any version of the tools but since you’re on Linux you probably have GNU tools anyway.

The above will do what you asked for given a single pair of matched [...] on a line as shown in your question but do different things for unmatched, multiple, nested, and/or overlapping pairs of [...] on a line – if you can have any of those situations and can’t solve that problem yourself then ask a new question about that showing how you’d want those cases handled.

I’m using logs instead of LOGS for the reasons described at Correct Bash and shell script variable capitalization.