Как я написал приложение, которое за 15 минут делало то же самое, что и регулярное выражение за 5 дней

$ python ./flashtext_regex_timing_keyword_replace.py 
Count  | FlashText | Regex    
-------------------------------
1      | 0.00968   | 0.00412   |
1001   | 0.01066   | 0.00425   |
2001   | 0.01156   | 0.00418   |
3001   | 0.01201   | 0.00436   |
4001   | 0.01136   | 0.00436   |
5001   | 0.01133   | 0.00439   |
6001   | 0.01215   | 0.00444   |
7001   | 0.01196   | 0.00442   |
8001   | 0.01209   | 0.00445   |
9001   | 0.01218   | 0.00446   |
10001  | 0.01218   | 0.00451   |
11001  | 0.01230   | 0.00446   |
12001  | 0.01239   | 0.00452   |
13001  | 0.01263   | 0.00457   |
14001  | 0.01265   | 0.00458   |
15001  | 0.01308   | 0.00457   |
16001  | 0.01283   | 0.00451   |
17001  | 0.01301   | 0.00455   |
18001  | 0.01296   | 0.00457   |
19001  | 0.01303   | 0.00458   |
20001  | 0.01318   | 0.00462   |

#!/usr/bin/env perl -w
use strict;
use warnings;
use feature 'say';

use DDP;
use String::Random;
use List::Util qw/shuffle uniq/;
use Time::HiRes qw/gettimeofday tv_interval/;

use Text::TabularDisplay;
use Storable;

my @columns = qw/Count PerlRegexp PythonFlashText PythonRegexp/;
my $table = Text::TabularDisplay->new(@columns);

# generate a random word of given length
my $string_gen = String::Random->new;
my @all_words = map { $string_gen->randpattern('c' x (3 .. 8)[int rand 6] ) } 1 .. 100_000;

my @rows;
for (my $keywords_length = 1; $keywords_length <= 20001; $keywords_length += 1000) { 
  # chose 5000 terms and create a string to search in.
  my $story = join ' ', (shuffle(@all_words))[1 .. 5000];

  # get unique keywords from the list of words generated.
  my @unique_keywords_sublist = uniq((shuffle(@all_words))[1 .. $keywords_length]);

  my ($t_start, $t_end);
  if ($keywords_length <= 17_000) {
    my %rep = map { $_ => '_keyword_' } @unique_keywords_sublist;
    my $compiled_re = qr/(@{[join '|', keys %rep]})/;

    $t_start = [gettimeofday];
    $story =~ s/$compiled_re/$rep{$1}/g;
    $t_end = [gettimeofday];

  } else {
    my @data;
    while (my @uniq = splice @unique_keywords_sublist, 0, 17000) {
      push @data, {
        rep => { map { $_ => '_keyword_' } @uniq},
        compiled_re => qr/(@{[join '|', @uniq]})/,
      }
    }
    $t_start = [gettimeofday];
    $story =~ s/$_->{compiled_re}/$_->{rep}{$1}/g for @data;
    $t_end = [gettimeofday];
    
  }
  push @rows, [$keywords_length, sprintf("%.5f",tv_interval($t_start, $t_end))];
  # say sprintf("%6d | %.5f |", $keywords_length, tv_interval($t_start, $t_end));
}

my %python_rows = (
  1     => [0.01188, 0.0005 ],
  1001  => [0.01347, 0.07646],
  2001  => [0.01381, 0.19431],
  3001  => [0.01370, 0.29055],
  4001  => [0.01381, 0.37468],
  5001  => [0.01407, 0.45641],
  6001  => [0.01418, 0.54225],
  7001  => [0.01425, 0.60971],
  8001  => [0.01449, 0.68178],
  9001  => [0.01484, 0.74831],
  10001 => [0.01478, 0.81751],
  11001 => [0.01497, 0.88625],
  12001 => [0.01496, 0.94201],
  13001 => [0.01512, 0.98292],
  14001 => [0.01520, 1.04620],
  15001 => [0.01551, 1.07280],
  16001 => [0.01553, 1.16832],
  17001 => [0.01611, 1.25071],
  18001 => [0.01581, 1.37876],
  19001 => [0.01585, 2.60501],
  20001 => [0.01600, 2.94057],
);

push @$_, @{$python_rows{$_->[0]}} for @rows;

$table->add(@$_) for @rows;
say $table->render;