Using sed to delete all lines between two matching patterns including the matched lines

Question

I need to delete lines in a file between the specific pattern matched and the matched line.

In the below code, I want to delete the lines from object Host "kali" { to the next occurrence of } (not to the last occurrence of }). and also delete the empty spaces after deletion.

object Host "linux" {
import "windows"
address = "linux"
groups = ["linux"]
}


object Host "kali" {
import "linux"
address = "linux"
groups = [linux ]
}


object Host "windows" {
import "linux"
address = "linux"
groups = ["windows" ]
}

This is my code

clear
echo -e  "Enter the host to delete in config file"
cat > deletionfile.txt
clear
while read host
do
loc=`grep -il 'object.*Host.*"$host"' /home/afrith/config-file/*.conf`
sed -i "/^object.*Host.*\"$host\".*{$/,/^}$/d" $loc
done < deletionfile.txt
rm -rf deletionfile.txt

Error show while executing script:

sed: no input files

This is the expected output: ( when i give kali as a input to script. )

object Host "linux" {
import "windows"
address = "linux"
groups = ["linux"]
}


object Host "windows" {
import "linux"
address = "linux"
groups = ["windows" ]
}

Answer 1

The following sed shoud remove all lines between object Host "kali" { and }.

sed '/^object Host "kali" {$/,/^}$/d'

Answer 2

Try this:

sed -i -Ez 's/object Host \"kali\"\s*\{[^}]*\}//g' your_file

Or, you could try the following (will also remove empty lines):

sed  '/object Host "kali"\s*{$/,/[^}]*}/d; /^$/d' your_file

The output will be:

object Host "linux" {
import "windows"
address = "linux"
groups = ["linux"]
}
object Host "windows" {
import "linux"
address = "linux"
groups = ["windows" ]
}

Answer 3

One of the way of doing it will be using AWK. You need to change the record separator in to } and run this awk '!/kali/;BEGIN{RS="}"; ORS="}"}' < your_file , the output is what you are looking for.

I have saved the text given in the question in file - test.txt

ajay@DESKTOP-PCMKVK6:~$ cat test.txt
object Host "linux" {
import "windows"
address = "linux"
groups = ["linux"]
}


object Host "kali" {
import "linux"
address = "linux"
groups = [linux ]
}


object Host "windows" {
import "linux"
address = "linux"
groups = ["windows" ]
}

now we need to run this awk command awk '!/kali/;BEGIN{RS="}"; ORS="}"}' < test.txt

ajay@DESKTOP-PCMKVK6:~$ awk '!/kali/;BEGIN{RS="}"; ORS="}"}' < test.txt
object Host "linux" {
import "windows"
address = "linux"
groups = ["linux"]
}


object Host "windows" {
import "linux"
address = "linux"
groups = ["windows" ]
}
}ajay@DESKTOP-PCMKVK6:~$ 

Also, you can use sed as mentioned by other answer and you can add -i option to make that change in the same file sed -i '/^object Host "kali" {$/,/^}$/d' file_name

Answer 4

You can make gnu awk to work in block mode by setting record selector to nothing.

awk -v RS= -v ORS="\n\n" '/object Host "kali"/ {next} 1' file
object Host "linux" {
import "windows"
address = "linux"
groups = ["linux"]
}

object Host "windows" {
import "linux"
address = "linux"
groups = ["windows" ]
}

• /object Host "kali"/ {next} if block contains object Host "kali", skip it
• 1 always true, do the default action, print

Answer 5

This might work for you (GNU sed):

sed -n '/^object Host "kali" {/{:a;n;/^}/d;ba};p' file

Or:

sed '/^object Host "kali" {/{:a;N;/^}/Md;ba}' file

The first solution acts like grep and chooses not to print the lines starting with object Host "kali" { and ending with a line starting }.

The second solution gathers up the lines between the above two patterns and then deletes the lines.

The first solution will alter the file if the first pattern matches whereas the second solution only alters the file if both pattern matches. This is because the commands n and N cause sed to end processing if they are asked to read beyond the end of the file. n will end quietly if the -n option is set whereas N will print what ever is in the pattern space before ceasing.

N.B. /^}/ matches a } at the start of line, in the second solution the M flag is added which turns on the multiline matching i.e. the ^ and $ match after/before a newline this flag is specific to GNU sed. /\n}/ will have the same effect. If the match is more relaxed i.e. to match any occurrence of },/}/ is suffice.