Subj : Re: Searching 40 Million short strings fast
To   : comp.programming
From : Peter Ammon
Date : Tue Sep 06 2005 09:16 pm

Willem wrote:
> Peter wrote:
> ) The most important optimization I do is to look for duplicate suffixes. 
> )   Rather than store five links for "bring" and six for "string," both 
> ) can link to a unique "ring" sequence, which reduces the total number of 
> ) nodes from 390158 to only 52904.
> 
> And what if you want to add "brings" ?  You'd have to separate the links
> again.  Don't you need extra housekeeping to tell that one node is pointed
> to from two locations ?
> 

In my case, the data is static, so I can afford expensive preflighting. 
  You're right that this optimization might be less useful if the data 
changes often.

-Peter

-- 
Pull out a splinter to reply.

.