Match a line between patterns

**dennis89** · 02-10-2009

Hi everyone,

grep matches lines, but I'm trying to match a line that itself contains no patterns, but it is within a pattern.

Thus for example,

pattern1
text that I want to match
pattern2

Code:

grep 'pattern1.*pattern2' file

The above (obviously) won't work because they're not on the same line.

How would I be able to do this with the text not being on the same line as a pattern (using grep if possible)?

Any help is appreciated. (Did a quick Google and forum search but couldn't find anything.)

edit: I was thinking of removing all new lines and thus having the text in the file show up as:

pattern1 text that I want to match pattern2

and then it would be easy to match the text.
I know how to get rid of preceding and trailing whitespaces using grep's invert-match option

Code:

grep -v '^$' file

How would I use this to get rid of lines? Or does anyone have another way how to accomplish this?

**tpl** · 02-10-2009

try awk: the "range pattern" is setup for what
you want to do: something like this:

awk '/pattern2,pattern2/ {delete}' <filename

might need some tweaking to get the delete to work

**drl** · 02-11-2009

Hi.

This quickly-written demonstration perl script, "p1":

Code:

#!/usr/bin/perl

# @(#) p1       Demonstrate bracketed search.

use warnings;
use strict;

my ($debug);
$debug = 0;
$debug = 1;

my ($op) = "alpha";
my ($cp) = "omega";
my ($p)  = "Halifax";

my ($looking_for_opening) = 1;
my ($hits)                = 0;
my (@lines);
my ($sections) = 0;

while (<>) {
  if ($looking_for_opening) {    # skip everything until opening
    if (/$op/) {
      $looking_for_opening = 0;
      @lines               = ();
      push @lines, $_;
    }
  }
  else {    # after opening, look for closing, or main match
    if (/$cp/) {
      if ( $hits > 0 ) {
        push @lines, $_;
        print @lines;
      }
      @lines               = ();
      $looking_for_opening = 1;
      $sections++;
      print " (debug): found $hits hits in section $sections\n" if $debug;
      $hits = 0;
      next;
    }
    elsif (/$p/) {
      $hits++;
    }
    push @lines, $_;
  }
}

print STDERR " ( Lines read: $. )\n";

exit(0);

Will look for string Halifax on lines, but only in sections of lines which begin with string alpha and end with string omega.

Some environment information, the data file "data1" and results of running the script as

Code:

./p1 data1

are:

Code:

Environment: LC_ALL = C, LANG = C
(Versions displayed with local utility "version")
OS, ker|rel, machine: Linux, 2.6.11-x1, i686
Distribution        : Xandros Desktop 3.0.3 Business
GNU bash 2.05b.0
perl 5.8.4

 Data file data1:
# Wed Feb 11 04:42:41 CST 2009
# Data file for bracketed match.

alpha
Text1
Text2
This is a line to be matched "Halifax"
Text3
omega

beta
alpha
Text4
but there is no match in this bracket section.
Text5
omega

# End of data file for bracketed match.

 Results:
alpha
Text1
Text2
This is a line to be matched "Halifax"
Text3
omega
 (debug): found 1 hits in section 1
 (debug): found 0 hits in section 2
 ( Lines read: 18 )

Other additions, such as easily changeable patterns, printing options of sections (e.g. only the main matching lines), etc., are left as an exercise ... cheers, drl

**secondmouse** · 02-11-2009

You can try sed:
1. create a text file a.sed:
/pattern1/{
N
N
s/pattern1\n.*\npattern2/something else/
}
2. sed -f a.sed file
now the pattern lines are replaced with "something else"

OR

change a.sed to
/pattern1/{
N
N
/pattern1\n.*\npattern2/d
}
then
sed -f a.sed file
to delete the pattern lines

**dennis89** · 02-11-2009

Originally Posted by secondmouse

You can try sed:
1. create a text file a.sed:
/pattern1/{
N
N
s/pattern1\n.*\npattern2/something else/
}
2. sed -f a.sed file
now the pattern lines are replaced with "something else"

OR

change a.sed to
/pattern1/{
N
N
/pattern1\n.*\npattern2/d
}
then
sed -f a.sed file
to delete the pattern lines

Thanks for the reply secondmouse,

I was actually looking at sed for doing this before I saw your reply. After all, sed has the ability to look at multiple lines in the pattern space, which allows you to match patterns which extend over multiple lines.

The thing is I've never used sed for such a task before, and all the examples I could find have all created a sed file.

I'd like to do this on command line if possible (not by creating a file); how would the syntax look like? I tried some stuff but kept getting errors.

edit: In my first post I asked if this was possible to do with grep. I don't believe it is, but can anyone confirm this?

**secondmouse** · 02-11-2009

edit: In my first post I asked if this was possible to do with grep. I don't believe it is, but can anyone confirm this?

It's true grep is not up for this kind of task, see [ [Chapter 27] 27.11 A Multiline Context grep Using sed ], first paragraph. Maybe agrep (which is not under GPL) can do that but I haven't tried.

For deleting with sed one-liner:
sed -e '/pattern1/,/pattern2/d' file

For replacing,
sed '/^pattern1$/{N;N;/.*pattern2$/s/.*/somethingelse/;}' file

The replacement solution I got it from:
how to replace multiple lines in a file? - comp.unix | Google Groups

**drl** · 02-12-2009

Hi.

I thought you wanted a 3-pattern windowed match. If an open-close sequence is desired, cgrep can do that (and much more):

Code:

#!/usr/bin/env bash

# @(#) s2       Demonstrate patterned window feature of cgrep.
# http://www.bell-labs.com/project/wwexptools/cgrep/

echo
set +o nounset
LC_ALL=C ; LANG=C ; export LC_ALL LANG
echo "Environment: LC_ALL = $LC_ALL, LANG = $LANG"
echo "(Versions displayed with local utility \"version\")"
version >/dev/null 2>&1 && version "=o" $(_eat $0 $1) cgrep
set -o nounset
echo

FILE=${1-data3}

echo " Data file $FILE:"
cat $FILE

echo
echo " Results:"
cgrep +w "omega" "alpha" $FILE

exit 0

Producing:

Code:

% ./s2

Environment: LC_ALL = C, LANG = C
(Versions displayed with local utility "version")
OS, ker|rel, machine: Linux, 2.6.11-x1, i686
Distribution        : Xandros Desktop 3.0.3 Business
GNU bash 2.05b.0
cgrep - (local: ~/executable/cgrep Sep 28 2007 )

 Data file data3:
Thu Feb 12 04:22:38 CST 2009
# Data file for open-close sequence extraction.

alpha
Text1
omega

# Extraneous data
# is not within an opening - closing sequence.

Dalhousie is in Halifax.

beta
alpha
Text2
Text3
omega

# Only the window boundaries, no enclosed text.

alpha
omega

# End of data file for bracketed match.

 Results:
========================================
alpha
Text1
omega
========================================
alpha
Text2
Text3
omega
========================================
alpha
omega

The cgrep code is not standard, but I had no trouble obtaining and installing it -- it is available at the site noted.

If you need something yet different from this, then I will await your post of a specific example of data and expected results ... cheers, drl

**dennis89** · 02-12-2009

Originally Posted by drl

Hi.

I thought you wanted a 3-pattern windowed match. If an open-close sequence is desired, cgrep can do that (and much more):

The cgrep code is not standard, but I had no trouble obtaining and installing it -- it is available at the site noted.

If you need something yet different from this, then I will await your post of a specific example of data and expected results ... cheers, drl

Thanks drl. I tried installing cgrep but the installation failed for me because I do not have root access on this machine.

I'll continue using sed. Thank you secondmouse.

**drl** · 02-12-2009

Hi.

I installed it in a directory in HOME that I call executable, and that directory is made to be part of my PATH environment variable.

I don't recall the details for creating cgrep, but the usual sequence is:

Code:

< download: ftp, wget, etc >
./configure
make
make install

now if one stops before the install, one can copy the binary -- the executable file -- to wherever one has permission ... cheers, drl

**dennis89** · 02-12-2009

Originally Posted by drl

Hi.

I installed it in a directory in HOME that I call executable, and that directory is made to be part of my PATH environment variable.

I don't recall the details for creating cgrep, but the usual sequence is:

Code:

< download: ftp, wget, etc >
./configure
make
make install

now if one stops before the install, one can copy the binary -- the executable file -- to wherever one has permission ... cheers, drl

I get stuck at the make install step.
This is the error that I get:

Code:

make[1]: Entering directory `/home/dennis/cgrepsrc/cgrep-8.15'
/bin/sh ./mkinstalldirs /usr/bin
  /usr/bin/install -c cgrep /usr/bin/cgrep
/usr/bin/install: cannot create regular file `/usr/bin/cgrep': Permission denied
make[1]: *** [install-binPROGRAMS] Error 1
make[1]: Leaving directory `/home/dennis/cgrepsrc/cgrep-8.15'
make: *** [install-am] Error 2

I don't have much experience with tarballs. This is my first time installing one.

I'd rather use cgrep than sed, because cgrep is more suitable for the task. Thanks for helping me out.

edit: Nevermind, I found my way around it.

Thread Tools

Display

Match a line between patterns

Posting Permissions