Use sed and read in (up to) three lines with the "N" subcommand. Use the "D" and "N" subcommands to set up a loop where the next one line is added to the pattern space and the first line of the pattern space is deleted. Once you have encountered your line the first line in the pattern space is the line you searched for.

(You could also use your pattern space solely for the searching and use the hold space as your buffer. In this case use the respective hold-space-manipulation commands instead of setting up the ring buffer with a D-N-loop.

Similar concepts using a buffer of three lines where one line is added "at the bottom" and the first line deleted surrepetitously can be implemented in awk or even shell too. The following implemetation is probably the slowest and least efficient one but the most readable, so it is showing the mechanism best. Note that the "next element" is - because of the ring structure of the buffer - also the third-to-last one:

Code:

function pNextElement
{
     (( iCnt=(iCnt+1)%3 ))
}

# main()

chLineBuffer[0]=""
chLineBuffer[1]=""
chLineBuffer[2]=""
iCnt=0

cat file | while read chLineBuffer[$iCnt] ; do
     if [ "${chLineBuffer[$iCnt]}" = "string_to_search" ] ; then
          pNextElement
          print - "The result is ${chLineBuffer[$iCnt]}"
     else
          pNextElement
     fi
done

I hope this helps.

bakunin