Skip to content
 

Before and after match in awk

The problem is: print the line n lines before or after a line that matches a given pattern. How to do this in awk?

The sample input

Here is a sample input file:

line1
foobar
line3
line4
line5
foobar
line7
foobar
line8

We want to look for matches of "foobar", and print the lines 1, 2 or 3 lines before or after the matching lines.

n lines after the match

This can be accomplished by maintaining a "queue" (implemented with an array) of the line numbers of the future lines to be printed, updated every time a match is found (ok, it's not a proper queue, but it has the same logical function). This uses the "two files" awk idiom, despite there being a single input file:

$ awk -v n=1 '/foobar/{queue[NR+n]} NR in queue' file.txt
line3
line7
line8
$ awk -v n=2 '/foobar/{queue[NR+n]} NR in queue' file.txt
line4
foobar
$ awk -v n=3 '/foobar/{queue[NR+n]} NR in queue' file.txt
line5
line8

To avoid keeping already-used keys in the array, they can of course be deleted, although this should be hardly needed (thanks igli from #awk):

$ awk -v n=5 '/foobar/{queue[NR+n]} NR in queue {print; delete queue[NR]}' file.txt

n lines before the match

This is slightly more complicated. Basically, we need to maintain a "sliding window" of n lines, and when a match is found, print the line NR-n lines before (which will be the first line in the sliding window). Modular arithmetic comes handy because using NR%n always gives a number from 0 to n-1, which can be used to access the sliding window array.

$ awk -v n=1 '/foobar/ && NR>n {print window[(NR-n)%n]}{window[NR%n]=$0}' file.txt
line1
line5
line7
$ awk -v n=2 '/foobar/ && NR>n {print window[(NR-n)%n]}{window[NR%n]=$0}' file.txt
line4
foobar
$ awk -v n=3 '/foobar/ && NR>n {print window[(NR-n)%n]}{window[NR%n]=$0}' file.txt
line3
line5

We also need to check that NR is greater than n because otherwise we may print lines before the first, which obviously does not make sense.

Be Sociable, Share!

One Comment

  1. SML says:

    very useful for pulling dates of events out of Oracle alert log

Leave a Reply

(required)