I wrote a script to replace "old_var" by "new_var" in all files in the current directory.
Now, what i wanted was the script reads a lsit of "old_var" and "new_var" from an input file
for eg: the input file is old_new.txt contains
a1 a2
b1 b2
c1 c2......

Here is my script
-------------------------------------------------------
#!/bin/bash

while read old_var new_var line
do

for file in $(grep -l $old_var *)
do
sed -e "s|$old_var|$new_var|g" $file > temp.c
mv -f temp.c $file
echo "Replaced $old_var in file $file"
done

done < $1
---------------------------------------------------------

the problem is say i want to replace PROG with J_E_PROG. normally, it will replace properly in all files.

but when it finds something like C_PROG_DEF, it will replace PROG in this also making it as C_J_E_PROG_DEF .
i don't want this C_PROG_DEF => C_J_E_PROG_DEF conversion.

Can i avoid it? how do i do that? I want only complete word replacement

please help.
Thanks in advance
Gauri

Hi,

You can try using word boundaries.

A small warning: Not all versions of sed allow this (for example: sed on AIX 4.3).

Ok, here's how it's done:

sed 's/\<red\>/RED/'

or

sed 's/\bred\b/RED/'

I'm kinda busy atm (work.....) so for more details feed your fav searchengine with: sed word boundary

Hope this helps.

Thanks! that's so cool....
it works...

you can use m4

(man m4)

here's what I did:

TEST DATA:
the m4 command line: (the -Uindex undefines the (maybe)default definition of index)
result is:

lovely job!