How to use sed to extract the text I have a text file look like this: /home1/users/usr1/bill.txt /home1/users/usr2/bom.txt /tmp/cost.txt All filename extension ended with txt. I need to output the as follow: /home1/users/usr1 /home1/users/usr2 /tmp Any idea? Please help.

Hi: IMO, sed is not the best solution for this problem. Most unix variants supply the dirname command: #!/bin/ksh while read str do dirname $str done xfile # need a less than sign between done # and xfile. This forum wipes it out You can also use the korn shell pattern matching operator to perform the same function: while read str do echo ${str%/*} done xfile # need a less than sign between done #and xfile. This forum wipes it out Regards, Nails

sed 's#/[a-z]*.txt##' your_file -jim

Yes! I gotit! But how can it write it into a file in the while loop? Thanks Nails

Hi: Here's three ways: 1) Redirect all of standard output to a file. If the script is called a.ss: a.ss > output.file 2) Within the script, use file handles 1 and 2 to send all the output to a file: exec 1>output.file 2>&1 3) Withing the while loop, send the output to a file: while read str do echo ${str%/*} done xfile > output.file # need a less than sign between done #and xfile. This forum wipes it out Regards, Nails

Try this sed command. It should do the trick sed 's:/[a-z]*.txt:/:g' infile > outfile