[SOLVED] sed: -e expression #1, char 57: unterminated `s' command

TashiDuks · 12-08-2016, 03:04 AM

Hi Everyone,

I have configuration file named "mymy.conf" with following content:

Code:

"subject" : "This HTTP alert on $device.node_id"

I want to replace text This HTTP alert on to something else as per the user input.

I have come up with following script:

Code:

#!/bin/bash
read -p "Do you want to change mail subject? [y/N]" responseMailSub
 case "$responseMailSub" in
      y|Y )
            #Ask for new mail subject
             echo "Enter New Mail Subject: "
             read newSub

             #Update the file
             sudo sed -e 's/\(\"subject" : "\)[^<]*\(\$device.node_id"\)/\1'$newSub' \2/g' .mymy.conf
             echo "Your mail subject changed to $newSub"
             ;;
      n|N )
             echo "Operation canceled... ";;
        * )  echo "Wrong option used. Operation terminated!!!"
             ;;

 esac;;

After running above script i get sed: -e expression #1, char 57: unterminated `s' command message.

To check the syntax i tried running same directly on TERMINAL sed -e 's/$"subject" : "$[^<]*$\$device.node_id"$/\1This FTP alert on \2/g' .mymy.conf and it worked fine.

Can anyone tell me why there is difference? Any syntax correction?

Thanks

Turbocapitalist · 12-08-2016, 03:06 AM

It looks like you forgot to escape one of the double quotes in your script.

TashiDuks · 12-08-2016, 03:16 AM

@Turbocapitalist sorry I didn't get on which double quotes you talking about.

Thanks

Turbocapitalist · 12-08-2016, 03:23 AM

In the sed options you have a lot of quotes. You've escaped one double quote but not the others. However, that probably is not the issue. You also have single quotes there around $newsub that need to be escaped or eliminated. Walk through the whole -e option and match each pair of quotes, start and end, and you'll see.

Turbocapitalist · 12-08-2016, 04:10 AM

Also, the actual content of $newsub will affect how sed runs or fails, especially if it contains a slash or whatever delimiter you happen to use for the s command in sed. You may want to sanitize that input.

grail · 12-08-2016, 07:40 AM

If the file mymy.conf contains a single line, why does the sed have to be so complicated? If it is the case that there are more lines in the file, what is unique about the line to be changed or do
we really need to regex the entire line because there is another in the file so similar that this is required?

As an example, if it were this single line, I would simply use:

Code:

sed "s/T.*on/$newSub/" mymy.conf

So please provide more details about the input file and what is required?

TashiDuks · 12-08-2016, 11:31 PM

Quote:

Originally Posted by Turbocapitalist

In the sed options you have a lot of quotes. You've escaped one double quote but not the others. However, that probably is not the issue. You also have single quotes there around $newsub that need to be escaped or eliminated. Walk through the whole -e option and match each pair of quotes, start and end, and you'll see.

The double quotes are for my following lines in my configuration file

Code:

"subject" : "This HTTP alert on $device.node_id"

Thanks

Turbocapitalist · 12-09-2016, 02:53 AM

You might benefit from temporarily putting set -x as the second line in your script. That will show you what the system really will use in each line.

rknichols · 12-09-2016, 09:00 AM

Quote:

Originally Posted by TashiDuks

Code:

             sudo sed -e 's/\(\"subject" : "\)[^<]*\(\$device.node_id"\)/\1'$newSub' \2/g' .mymy.conf

You insert $newsub unquoted. If that string contains any spaces or other field separator characters, it's going to end the sed argument prematurely. And, note that escaping that double quote that precedes "subject" is unnecessary, and there is little point to the "g" modifier since there should never be more than one occurrence of the pattern. Try:

Code:

sudo sed -e 's/\("subject" : "\)[^<]*\(\$device.node_id"\)/\1'"$newSub"' \2/' .mymy.conf

There will still be a problem if $newsub contains any "/" characters, since those would corrupt the substitution command. You might want to pre-process $newsub to escape any "/" characters.

Code:

newSub2="$(sed 's@/@\\/@g' <<<$newSub)"
sudo sed -e 's/\("subject" : "\)[^<]*\(\$device.node_id"\)/\1'"$newSub2"' \2/' .mymy.conf

There would still be a problem with backslash characters, but those are likely to be absorbed by the read command anyway.

TashiDuks · 12-13-2016, 11:33 PM

Quote:

Originally Posted by rknichols

You insert $newsub unquoted. If that string contains any spaces or other field separator characters, it's going to end the sed argument prematurely. And, note that escaping that double quote that precedes "subject" is unnecessary, and there is little point to the "g" modifier since there should never be more than one occurrence of the pattern. Try:

Code:

sudo sed -e 's/\("subject" : "\)[^<]*\(\$device.node_id"\)/\1'"$newSub"' \2/' .mymy.conf

There will still be a problem if $newsub contains any "/" characters, since those would corrupt the substitution command. You might want to pre-process $newsub to escape any "/" characters.

Code:

newSub2="$(sed 's@/@\\/@g' <<<$newSub)"
sudo sed -e 's/\("subject" : "\)[^<]*\(\$device.node_id"\)/\1'"$newSub2"' \2/' .mymy.conf

There would still be a problem with backslash characters, but those are likely to be absorbed by the read command anyway.

Hi,

Thanks for the help. I have finally got workable script using following line:

Code:

 sudo sed -i 's/\("subject" : "\)[^<]*\(\$device.node_id"\)/\1'"$newSub"' \2/' .mymy.conf

Cheers