I have the following file : First string second string third string 932 --end of file---- The output should look like: First string First string First string second string third string third string 932 932 --end of file---- I want keep copying a line to next line(s) till I reach another string/number using "sed".

In my opinion, sed is not the best tool for this problem. It will take a sed programmer who is smarter than I am. If this is homework and you have to use sed, take a look at this link: http://www.student.northpark.edu/pe... In the above thread, Eric Pement explains how to use if comparisons in sed. Perhaps it will help you. In my shell solution below, I read the file a line at a time and if the line is defined I print it out and save it in another variable - saveline. If the line is undefined, I print the variable saveline: #!/bin/ksh while read line do if [[ -z "$line" ]] then echo "$saveline" else saveline="$line" echo "$line" fi done < file

excellent ..it works...thanks . No I don't NEED to use 'sed' ..I thought it might be the best way...But I was wrong :) Another question . If I have a file with tab seperated columns ..how can I use your script only for one particular column?

Actually, you don't have to do anything, and the script should work as it does now. That's because the default field separater, IFS, in the shell is whitespace. Whitespace is defined as spaces, tabs, and carriage returns. BTW, you're welcome.

sed -e '/[^ ]/{h;b}' -e g junk.txt The sed for this is pretty simple actually. If the line contains any characters, it will print it and shove it into the hold buffer. If it is an empty line (or only blanks), it will print whatever is in the hold buffer.