You seem to be dealing with record oriented data where records are separated by empty lines. Often it is convenient to read a record as a time. Consider:

use strict;
use warnings;

my $data = <<DATA;
this is test only 1
Baki: 9.70

this is test only 2
Baki : 9.30

this is test only 3
Baki : 9.00
DATA

my $lastRecord;

open my $in, '<', \$data;

local $/ = "\n\n"; # Read a record at a time

while (defined (my $record = <$in>)) {
    chomp $record;
    next if $record =~ /$(?:Baki)/;
    $lastRecord = $record;
}

print $lastRecord;
[download]

Prints:

this is test only 3
Baki : 9.00
[download]

Note the use of three parameter open and a lexical file handle. Usually there would be an or die ...; for the open as well, but in this case (using a string as a file) there is no need.

• create a variable for your result (e.g. my $last_baki = "no baki there";)
• use a loop to read the whole data-structure element-by-element (in your case: line-by-line)
  • if the element matches your requirements (in your case: contains the word baki), overwrite the result-variable with this element
• when you have finished your loop, the result-variable will contain the last matching entry

What do you think?

<hi>
This is my opinion by using shell script


system ("cat file.txt |grep "Baki" |tail -1");
[download]

As a Perl one-liner, from the Unix command line, that would be:

perl -ne '/Baki/ and $x=$_}{print$x' file.txt
[download]

or:

perl -ne '/Baki/ and $x=$_; END{print$x}' file.txt
[download]

Update: BTW, your shell solution:

cat file.txt |grep "Baki" |tail -1
[download]

should read:

grep 'Baki' file.txt |tail -1
[download]

Tom Christiansen once wrote long ago on usenet something like "If you find yourself calling cat with just one argument, you're probably doing something silly".

Update: See also Re: perl one liner for csv file one field (useless use of cat and other awards References)