Re^4: Fuzzy Searching: Optimizing Algorithm Selection

[demerphq]

Im not at liberty to post my code. Sorry. All I can say is what I already have said, it is an Inline::C implementation of a Patricia tree optimized for handling a 10 character alphabet.

Regarding all your bollocks in this reply, first off i was very clear in my node that i misunderstood the question and that my proposals only handled the issue of finding exact matches of the fuzzy word list. Not finding substrings. Since that is what you insulted and rubbished me about then that is what i am talking about.

I have already stated that I ran my tests against live data (35 million telephone numbers) and that it took 0.00006 seconds per telephone numbers, or 1 million lookups per minute, or the full 35 million in just over 30 minutes. (This includes the overhead of reading the file and splitting the records by tabs.)

Now comparing your algorithm against mine will mean that you only have to test to see if there is a fuzzy match from the first character. So if you want to provide a subroutine that i can run over my test data set then we are in business. But im not going to help you write it, and in fact im not going to do any more than run it to see. As it is I should be working on speeding up the site not sparring with someone who doesnt treat me with the decency I afford to him.