SED Output Column Before and After Match

Your example desired output is incorrect.  I know you asked for SED, but attached is a PERL script that will accomplish what you are asking for.  Assumption is test5.in contains the input tab delimited data you specified in the example.

test5.in:
aaron ross who is he dating
aaron spelling dating
aaron t jeffries dating profile
aarp dating service
a basic guide to dating edwin m knowles china
abbreviation for ipa in dating
abbreviations gl dating

sample output:
is he dating
aaron spelling dating
t jeffries dating profile
aarp dating service
guide to dating edwin m
ipa in dating
abbreviations gl dating

cat test5.in | perl -ne 'print "$1\n" if /(([^\s]+\s){0,2}dating(\s[^\s]+){0,2})/'

Select allOpen in new window

Your example output is correct although your description was a bit murky and I can understand why aPresence didn't understand.  Your intention is to always output exactly three columns, either the leftmost three columns if the match is leftmost, the rightmost three columns if the match is rightmost, otherwise the left ad right adjacent columns along with the matched column.  I have attached a Python program which generates the exact output you describe given the exact sample input you describe.  

A couple of notes:  the program acepts input in a file called "columnmatch.txt" and writes the output to "columnmatch.out".   It could be easily modified to accommodate other input and output schemes.  Also the match I use is rather primitive, just a case-insensitive, whole-word match on the column value.  For more sophisticated matching, there are rich regular-expression searching features available in the Python "re" library.  Just say "import re" in the program and do a little Googling to find out its capabilities.

columnmatch.py

This should work:

perl -ane 'chomp; if( $F[0] eq "dating" ){print "dating @F[1,2]\n"} elsif( $F[-1] eq "dating" ){print "@F[-3,-2] dating\n"} else{foreach $i(0..$#F){ if( $F[$i] eq "dating" ){print "$F[$i-1] dating $F[$i+1]\n"; last}}}' input_file

Here's a less terse version

use strict;
while( <> ) {
   chomp; 
   my @F = split;
   if( $F[0] eq "dating" ) {
      print "dating @F[1,2]\n";
   } elsif( $F[-1] eq "dating" ) {
      print "@F[-3,-2] dating\n";
   } else {
      foreach my $i(0..$#F) { 
	 if( $F[$i] eq "dating" ) {
	    print "$F[$i-1] dating $F[$i+1]\n"; 
	    last;
	 }
      }
   }
}

Select allOpen in new window

I think the correct output you are looking for is this (your example is incorrect):

is he dating
aaron spelling dating
jeffries dating profile
aarp dating service
to dating edwin
ipa in dating
abbreviations gl dating

Select allOpen in new window

I diff'd the requested output and my output and they look correct... except for the iipa typo on line 6 and the leading space on line 5.

Thanks!!

Hi faithless1,
out of curiousity, what made you choose shaleesh's solution over the Perl solution I offered earlier?
I believe mine produced the expected output.
Please let me know.

Thanks.

Faithless1 clearly wanted a SED solution.  I was the first one to figure out what he was really asking for and my solution was the first to create the correct output, but it was in Python.   It would have been preferable had faithles1 not categorized this question in the perl and python zones which wasted the efforts of a numer of Perl and Python experts.

Thanks Richard.