Benchmarking timing performance Keyword Replace between regex and flashtext

It doesn't get you in to "comparable to FlashText" territory, but an easy regex optimisation to apply is to sort the keys before adding to the regex.

$ python flashtext_regex_timing_keyword_replace.py
Count  | FlashText | Regex | Sorted Regex
------------------------------------------
('1     ', '|', '0.01020  ', '|', '0.00007  ', '|', '0.00006  ', '|')
('1001  ', '|', '0.01195  ', '|', '0.15300  ', '|', '0.14172  ', '|')
('2001  ', '|', '0.01223  ', '|', '0.30786  ', '|', '0.28629  ', '|')
('3001  ', '|', '0.01370  ', '|', '0.43211  ', '|', '0.39781  ', '|')
('4001  ', '|', '0.01226  ', '|', '0.57421  ', '|', '0.53352  ', '|')
('5001  ', '|', '0.01331  ', '|', '0.66974  ', '|', '0.67607  ', '|')
('6001  ', '|', '0.01246  ', '|', '0.81990  ', '|', '0.70473  ', '|')
('7001  ', '|', '0.01144  ', '|', '0.92889  ', '|', '0.89696  ', '|')
('8001  ', '|', '0.01229  ', '|', '1.02987  ', '|', '0.92075  ', '|')
('9001  ', '|', '0.01200  ', '|', '1.14402  ', '|', '1.08537  ', '|')
('10001 ', '|', '0.01359  ', '|', '1.23232  ', '|', '1.07877  ', '|')
('11001 ', '|', '0.01344  ', '|', '1.38192  ', '|', '1.27324  ', '|')
('12001 ', '|', '0.01369  ', '|', '1.47766  ', '|', '1.31015  ', '|')
('13001 ', '|', '0.01565  ', '|', '1.46536  ', '|', '1.39552  ', '|')
('14001 ', '|', '0.01474  ', '|', '1.63738  ', '|', '1.54096  ', '|')
('15001 ', '|', '0.01419  ', '|', '1.75537  ', '|', '1.68527  ', '|')
('16001 ', '|', '0.01576  ', '|', '1.85689  ', '|', '1.68421  ', '|')
('17001 ', '|', '0.01501  ', '|', '1.90390  ', '|', '1.77516  ', '|')
('18001 ', '|', '0.01641  ', '|', '1.96707  ', '|', '1.77640  ', '|')
('19001 ', '|', '0.01499  ', '|', '2.06039  ', '|', '1.73332  ', '|')
('20001 ', '|', '0.01688  ', '|', '2.04545  ', '|', '1.92251  ', '|')

The speed improvement is quite varied depending on the keys chosen, but it's almost always a safe thing to do.

 sorted_compiled_re = re.compile("|".join(sorted(rep.keys())))