[Mongrel] [ANN] Mongrel 0.3.15 PR -- All The Fixes Good For You

Aleksandar Lazic al-mongrelusers at none.at
Thu Nov 16 03:57:27 EST 2006


On Mit 15.11.2006 18:20, Zed A. Shaw wrote:
>When you get this new version (and read the ChangeLog) you'll see
>mention of a BMH implementation.  This is the Boyer-Moore-Horspool
>algorithm for finding one string in another:
>I took the above code (which I guess is alright) and modified the hell
>out of it so that you can pass successive chunks of the haystack and
>it'll find all the needles ultra fast even across chunk boundaries.  My
>initial performance measurements puts it at about 3.84G/second. Yes, as
>in 26 seconds to process 100G of data with 99000 mime boundaries in it.
>Commence the arguing.

Cool, we have used also the BMH for awffull[1] and was also impressed how
fast this algo work ;-)

I have compared it whith FJS (http://cgjennings.ca/fjs) and it looks for
our issue not so fast, but it could also be a implemention problem ;-)



[1] http://www.stedee.id.au/awffull

More information about the Mongrel-users mailing list