#!/usr/bin/perl
use strict;
use warnings;
use Benchmark 'cmpthese';
use Time::HiRes qw/tv_interval gettimeofday/;

#https://perlmonks.org/index.pl?node_id=11164097

open my $fh, '<', 'try3.txt' or die $!; # try3.txt contains the origin
+al data

my $s;

{
    local $/ = undef; #local $/ = '', is paragraph mode
    $s = <$fh>;    # slurp file
}

$s = $s x 10_000; # 90_000 lines (30_000 records)

close $fh or die $!;


sub GRT {
    my $s = shift;
    map { unpack q{x4a*}, $_ }
   sort
   map {
       m{(\d+)(?=%)}
       &&
       ( ~ pack( q{N}, $1 ) . pack( q{a*}, $_ ) )
   }
   do  {
       #local $/ = q{}; # $s is slurped before passing to split functi
+on
       split m{(?<!\A)(?=>>>)}, $s;
   }    
}

sub ST {
    my $s = shift;
    map {$_->[0]}
    sort {$b->[1] <=> $a->[1]}
    map {[$_, /(\d+)%/]} split(/^(?=>>> )/m, $s)
}

my $t0;
my $elapsed;

$t0 = [gettimeofday];
GRT($s);
$elapsed = tv_interval ( $t0, [gettimeofday]);

print "GRT time: ", $elapsed, "\n";

$t0 = [gettimeofday];
ST($s);
$elapsed = tv_interval ( $t0, [gettimeofday]);

print "ST time:  ", $elapsed;

=begin
C:\Old_Data\perlp>perl GRT_ST_transform.pl
GRT time: 0.222224
ST time:  0.05873
C:\Old_Data\perlp>
=cut

#cmpthese( -1, { 'a' => GRT($s), 'b' => ST($s) } ) ;
[download]

>>> prd1701
Filesystem                                  Size  Used Avail Use% Moun
+ted on
/workspace                                  3.9T  887G  3.0T  13% /wor
+kspace/data
>>> prd1702
Filesystem                                  Size  Used Avail Use% Moun
+ted on
/workspace                                  3.9T  746G  3.1T  23% /wor
+kspace/data
>>> prd1703
Filesystem                                  Size  Used Avail Use% Moun
+ted on
/workspace                                  3.9T  687G  3.2T  18% /wor
+kspace/data
[download]

cmpthese( -1, { 'a' => GRT($s), 'b' => ST($s) } ) ;
[download]

This calls both GRT and ST just once, and uses their output as the code to eval and benchmark. That's not what we want. Instead, we need

cmpthese(-3, {
    grt => sub { GRT($s) },
    st  => sub { ST($s) }
});
[download]

It still seems that ST is much faster:

      Rate  grt   st
grt 2.75/s   -- -76%
st  11.6/s 320%   --
[download]

But, notice the regex is different when splitting the string:

       split m{(?<!\A)(?=>>>)}, $s     # GRT
       split       (/^(?=>>> )/m, $s)  # ST
[download]

So, let's add a test to verify the results stay correct

use Test::More tests => 1;
is GRT($s), ST($s), 'same';
[download]

and we can use the simpler regex from ST in GRT. Fortunately, the test remains successful, and the results are now different:

1..1
ok 1 - same
      Rate  st grt
st  11.7/s  -- -3%
grt 12.1/s  3%  --
[download]

An insignificant difference, but at least ST doesn't seem to be much faster now.

Benchmarking and optimisation are hard.

map{substr$_->[0],$_->[1]||0,1}[\*||{},3],[[]],[ref qr-1,-,-1],[{}],[sub{}^*ARGV,3]