Using empty line as context "group-separator" for grep

Question

I need grep output with context, in color, and blank lines as group separator. In this question, I learned how to define custom group-separator, and I have constructed my grep command like this:

grep --group-separator="" --color=always -A5

but the group separator is actually not empty, instead it still contains the color code (i.e. [[36m[[K[[m[[K). This is because I am using --color=always. But I need color in my grep command, and I need separator to be blank line (for further processing)

How can I combine these two conditions?

Answer 1

If you use the GREP_COLORS environment variable you can control specific colors for each type of match. man grep explains the use of the variable.

The following command will print a colored match, but nothing on the line that separates the group, just a blank line. Piped through od you'll see the color escape before and after the match, but only \n\n at the group separation.

GREP_COLORS='ms=01;31:mc=01;31:sl=:cx=:fn=35:ln=32:bn=32:se=' grep --group-separator="" --color=always -A5

Unsetting the se component will suppress the printing of color in the group separator.

Since my example above used all of the default values for GREP_COLORS the following will work as well.

GREP_COLORS='se=' grep --group-separator="" --color=always -A5

If you're not using a bashlike shell, you might need to export GREP_COLORS first.

Answer 2

Personally, I do that using Perl, not grep. I have a little script that will highlight a given pattern in color:

#!/usr/bin/env perl
use Getopt::Std;
use strict;
use Term::ANSIColor; 

my %opts;
getopts('hsc:l:',\%opts);
    if ($opts{h}){
      print<<EoF; 
DESCRIPTION

$0 will highlight the given pattern in color. 

USAGE

$0 [OPTIONS] -l PATTERN FILE

If FILE is ommitted, it reads from STDIN.

-c : comma separated list of colors
-h : print this help and exit
-l : comma separated list of search patterns (can be regular expressions)
-s : makes the search case sensitive

EoF
      exit(0);
    }

my $case_sensitive=$opts{s}||undef; 
my @color=('bold red','bold blue', 'bold yellow', 'bold green', 
           'bold magenta', 'bold cyan', 'yellow on_magenta', 
           'bright_white on_red', 'bright_yellow on_red', 'white on_black');
## user provided color
if ($opts{c}) {
   @color=split(/,/,$opts{c});
}
## read patterns
my @patterns;
if($opts{l}){
     @patterns=split(/,/,$opts{l});
}
else{
    die("Need a pattern to search for (-l)\n");
}

# Setting $| to non-zero forces a flush right away and after 
# every write or print on the currently selected output channel. 
$|=1;

while (my $line=<>) 
{ 
    for (my $c=0; $c<=$#patterns; $c++){
    if($case_sensitive){
        if($line=~/$patterns[$c]/){
           $line=~s/($patterns[$c])/color("$color[$c]").$1.color("reset")/ge;
        }
    }
    else{
        if($line=~/$patterns[$c]/i){
          $line=~s/($patterns[$c])/color("$color[$c]").$1.color("reset")/ige;
        }
      }
    }
    print STDOUT $line;
}

If you save that in your path as color, you can get your desired output by running

grep --group-separator="" --color=never -A5 foo | color -l foo

That way, the script is coloring the matches for you and you can tell grep not to use colors and avoid this problem.

Answer 3

I use rcg (Regexp Colored Glasses) by Rob Flickenger. Here's a copy:

#!/usr/bin/perl -w
=head1 NAME
rcg - Look at the world through Regex Colored Glasses
=head1 SYNOPSIS
rcg sudo: RED < /var/log/messages

tail /var/log/syslog | rcg '\d\d:\d\d:\d\d' "BOLD . BLUE" 


=head1 DESCRIPTION
rcg is a simple filter that uses Term::ANSIColor to colorize arbitrary
regexs, specified on the commandline.  It is intended to help visually slog
through log files.
You must pass 'rcg' an even number of commandline parameters.  The odd terms
specify the regex, the even terms specify the color.
=head1 EXAMPLES
Suppose you wanted anything with the word 'sendmail' in your messages log to
show up magenta, instead of grey:
rcg sendmail MAGENTA < /var/log/messages | less -r


The 'less -r' is optional, but handy if you have less(1) installed.
You can use any arbitrary regex as an odd term:
rcg '\d+\.\d+\.\d+\.\d+' GREEN < /var/log/maillog


Or chain colors together:
tail -50 /var/log/messages | rcg WARNING "BOLD . YELLOW . ON_RED" 


Or specify any number of regex / color code pairs on a single commandline. 
This is where teeny shell scripts would come in handy (one for messages, one
for firewall logs, one for apache...)
See the Term::ANSIColor docs for the full list of colors and combinations.
Some other useful strings:
'\w+=\S+'               variables, like TERM=xyz
'\d+\.\d+\.\d+\.\d+'        probably an IP address
'^(J|F|M|A|S|O|N|D)\w\w (\d| )\d'   might be a date
'\d\d:\d\d:\d\d'            possibly the time
'.*last message repeated.*'     make this "BOLD . WHITE"



Use your imagination.
=head1 PREREQUISITES
Term::ANSIColor
less -R (or even -r) is nice...
=head1 BUGS
=over 4
=item * Color params are just eval()'d.  You have been warned.
=item * Colorization is applied in arbitrary order, so it's not possible to
guarantee the behavior of overlapping regex's...
=back
=head1 AUTHOR
A particularly demented Rob Flickenger <rob@oreillynet.com> on a late
evening.  Use at your own risk!  Demand a refund!  See where it gets you!
=head1 SEE ALSO
regex(7).  No, don't look at that.  Look at Camel3 or Mastering Regular
Expressions in Perl...
=cut
use strict;
use Term::ANSIColor qw(:constants);
my %target = ( );
while (my $arg = shift) {
  my $clr = shift;
if(($arg =~ /^-/) | (!$clr)) {
    print "Usage: rcg [regex] [color] [regex] [color] ...\n";
    exit;
  }

Ugly, lazy, pathetic hack here

$target{$arg} = eval($clr);
}
my $rst = RESET;
while(<>) {
  foreach my $x (keys(%target)) {
    s/($x)/$target{$x}$1$rst/g;
  }
  print;
}