Shell Programming and Scripting

I tried aigles' solution too, that was my first intuition, but it didn't seem too elegant after all. At least on my awk, you have to add a { print } at the end for it to do anything at all. Other than that, yes.

To continue in sed, the regular expressions would tend to get pretty dense, especially if in reality you have even more fields. I'm such a hammer-nail-regex person that I'd still probably try that before awk in real life (and then switch to perl if the sed got too wicked).

Code:

sed '/^bar:A:.*/bar:A:done/'

if there can be any number of columns before and after, something like

Code:

sed 's/\(\(^\|:\)bar:A:\)[^:]*/\1done/'

If there can be any number of fields in between, and they could be in either order, but the last field is always the one to change (don't look if you have a tendency for migraine):

Code:

sed -e 's/\(\(^\|:\)bar:\([^:]*:\)*A:\([^:]*:\)*\)[^:]*$/\1done/' -e t \
    -e 's/\(\(^\|:\)A:\([^:]*:\)*bar:\([^:]*:\)*\)[^:]*$/\1done/' file

Read up on regular expressions and back references if you would like to understand how this works.

Some sed variants might dislike the backslashes in the regex, or want more, or less, or different syntax. Consult the manual page for your local variant.

Sorry, I missed the space around the ':' delimiter in the input file, and the result is not written.
Try:

Code:

awk 'BEGIN { FS=OFS=":" } $1 == "bar " && $2 == " A " { $3="done" } 1' inputfile

Another solution with gawk :

Code:

awk 'BEGIN { FS=OFS=" : " } $1=="bar" && $2=="A" { $3="done" } 1' inputfile

Jean-Pierre.

Thank you very much era and aigles, now i think i have something to play around with for the rest of the night.
knoxo

Code:

while read line
do
echo $line | grep -w bar | grep A 1> /dev/null
if [ $? -le 0 ] 
then
	echo $line | grep -w bar | grep A | sed 's/new/done/'
else 
	echo $line
fi
done < filename