On 03/26/2012 02:41 PM, Dmitry Olshansky wrote:
> On 27.03.2012 0:27, James Blewitt wrote:
>> Hello everybody,
>>
>> Thanks once again for the interest in my problem. I have posted the
>> details and source code that recreates (at least for me) the poor
>> performance.
>> I didn't know how to post the code to the forum, so I posted it to my
>> blog instead (see post update):
>>
>> http://jblewitt.com/blog/?p=462
>>
>> Again, if I'm doing something stupid in my code (which is possible) then
>> I apologise in advance.
>
> No need to apologize, but you are using 2.054, which is unfashionable :)
> More importantly 2.054 contains old and rusty version of std.regex, the
> new version was included in 2.057+.
> BTW The current release is 2.058.
>
>>
>> I'll take a look at the ctRegex as soon as I can.
>>
>
> Yup, just update compiler+phobos.
>
>> Regards,
>> James
>
>

My unofficial results comparing 2.056 to 2.058 on 64 bits:

shakespeare.txt, 2.056 -> 1868 msecs
shakespeare.txt, 2.058 ->  632 msecs

data.csv, 2.056 -> 51953 msecs
data.csv, 2.058 ->  1329 msecs

That last line is pretty impressive. :)

Ali

On 27 March 2012 11:05, Ali Çehreli <acehreli@yahoo.com> wrote:
> My unofficial results comparing 2.056 to 2.058 on 64 bits:
>
> shakespeare.txt, 2.056 -> 1868 msecs
> shakespeare.txt, 2.058 ->  632 msecs
>
> data.csv, 2.056 -> 51953 msecs
> data.csv, 2.058 ->  1329 msecs
>
> That last line is pretty impressive. :)

Dmitry did impressive work over those few version of Phobos/DMD. The performance is even more impressive when you consider that std.regex supports things like named matching and lookbehind that often slow down a regex (also kinda removes the "regular" from the name regular expression, technically)

--
James Miller

On Monday, 26 March 2012 at 22:05:34 UTC, Ali Çehreli wrote:
> My unofficial results comparing 2.056 to 2.058 on 64 bits:
>
> shakespeare.txt, 2.056 -> 1868 msecs
> shakespeare.txt, 2.058 ->  632 msecs
>
> data.csv, 2.056 -> 51953 msecs
> data.csv, 2.058 ->  1329 msecs
>
> That last line is pretty impressive. :)
>
> Ali

Unofficial 2.056/2.058/Ruby 1.9.3 Windows 32bit data.csv:



data.csv, 2.056 -> 76351 msecs
data.csv, 2.058 ->  2573 msecs
data.csv, 1.9.3 ->  9170 msecs

Also I had to modify line 48 of the ruby file not knowing what I'm doing:

if text.force_encoding("UTF-8") =~ /#{rule}/u

Couldn't build it with ctRegex (Some Error, then ran out of memory).

Great!

Thanks for the support everyone.  What a performance jump between v2.054 and v2.058!

James

On 3/27/12 1:57 AM, James Blewitt wrote:
> Great!
>
> Thanks for the support everyone. What a performance jump between v2.054
> and v2.058!
>
> James

Hi James -- you may want to link this discussion from your blog.

Cheers,

Andrei