I'm using sed.i.e search for pattern and then do the replace. I want to replace 5657 with 10000 in lines which have the pattern /xyz/data. I will be passing env variables to sed and I tried the following but doesn't work. Can some one help me on this? $cat file123 /xyz/data01,1234 /xyz/data02,1234 /xyz/data,5657 $export FX str str2 $echo $FS $str $str2 /xyz/data 5657 100000 $sed "/$FX/s/$str/$str2/g" file123 First RE may not be null $sed "%$FX%s%$str%$str2%g" file123 Unrecognized command: %/xyz/data%s%5657%100000%g $

This is a tricky one. First, if the shell variable to be embedded in the sed script contains a front slash, it must be escaped \/ FX="\/xyz\/data" Second, the RE must be surrounded by / / Third, note that I included a caret character, ^, at the beginning of the character so only lines in the file starting with: /xyz/data are modified. #!/bin/ksh FX="\/xyz\/data" str=5657 str2=10000 sed "/^${FX}/s!${str}!${str2}!g" file123

Hi nails, Ah!. You have rescued me again. Thanks nails. The solution worked well.

use awk #!/bin/bash FX="\/xyz\/data" str=5657 str2=10000 awk -v str=$str -v str2=$str2 '/xyz/{gsub(str,str2)}1' file GNU win32 packages | Gawk