From: "Shot (Piotr Szotkowski)" <shot@...>
Date: 2009-03-30T07:38:59+09:00
Subject: Re: 1.9.1 regex 6.5 times slower than 1.8.6 in at least one case

--IS0zKkzwUGydFO0o
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

// I=E2=80=99m slowly trawling through my ruby-talk backlog, so apologies
// if the below is irrelevant to anyone at this point in time. :)

Michael Brooks:

> However, in one case, when using a regular expressions (posted here by=20
> someone a long time ago) which I use to determine what numbers in=20
> 0..10000 are prime numbers, version 1.9.1 was at least 6.5 times=20
> slower  (1.8.6 =3D 67 secs, 1.9.1 =3D 457 secs).

> The regular express is:

>     ((("1" * self) =3D~ /^1$|^(11+?)\1+$/) =3D=3D nil)

> which I've used in different versions of my program running either
> on  its own or as part of an overloading function as follows:

>   # Add an "is_prime?" method to the built-in numeric class
>   # which returns true if the number is a prime number.
>   class Fixnum
>     def is_prime?
>       ((("1" * self) =3D~ /^1$|^(11+?)\1+$/) =3D=3D nil)
>     end
>   end

Note that this is a really ineffective way of testing primarity:

ruby 1.9.1p0 (2009-01-30 revision 21907) [i686-linux]
>> class Fixnum
>>   def is_prime?
>>     ((("1" * self) =3D~ /^1$|^(11+?)\1+$/) =3D=3D nil)
>>   end
>> end
=3D> nil
>> require 'prime' # adds Fixnum#prime? (among others)
=3D> true
>> start =3D Time.now; (0..1000).select(&:is_prime?); Time.now - start
=3D> 0.405157579
>> start =3D Time.now; (0..1000).select(&:prime?); Time.now - start
=3D> 0.016556705
>> start =3D Time.now; (0..5000).select(&:is_prime?); Time.now - start
=3D> 30.165383712
>> start =3D Time.now; (0..5000).select(&:prime?); Time.now - start
=3D> 0.105887884

Note that the Prime class has two bugs (as of 1.9.1-p0, both will be
fixed in the next 1.9.1 release), but your regex also has one of them:

>> 0.is_prime?
=3D> true # BUG
>> 1.is_prime?
=3D> false
>> 0.prime?
=3D> true # BUG
>> 1.prime?
=3D> true # BUG

(Neither 0 nor 1 are prime numbers.)

> I also have a version of the prime-number calculation program that=20
> which doesn't use the above regex (i.e. it uses a traditional brute=20
> force approach instead) and it runs 2 times faster in 1.9.1.

I highly recommend http://ruby-doc.org/core-1.9/ =E2=86=92 Prime class (et =
al.).

=E2=80=94 Shot
--=20
>> The Ruby community should proceed with all deliberate
>> speed towards ISO standardization of the language.
> Yeah, look what it did to Forth.
Don=E2=80=99t just say it, show it. http://vividpicture.com/aleks/atari/for=
th.jpg
        [M. Edward (Ed) Borasky, Matt Lawrence, Gregory Brown, ruby-talk]

--IS0zKkzwUGydFO0o
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: Digital signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEYEARECAAYFAknP9kYACgkQi/mCfdEo8UpwAACfRdRlYmKafSNntNzvguDvzMME
floAoJ+BO1HQ9wMWOCvqpCrkqbrXWz2X
=PCnV
-----END PGP SIGNATURE-----

--IS0zKkzwUGydFO0o--