Just to make things more confusing, I just learned about the
Burrows-Wheeler transform in the December 2003 Dr. Dobb’s. It
apparently can yield both compression and free-text searchability for
any substring (not just words as defined by an indexing engine). I want
to understand more about it and the peripheral subjects mentioned in the
article that are required to compute the location of the search
substring within the document, but it seems like a very interesting
algorithm even without that part!
Al
···
-----Original Message-----
From: Clifford Heath [mailto:cjh_nospam@managesoft.com]
Sent: Sunday, December 07, 2003 3:38 PM
To: ruby-talk ML
Subject: Re: Idea: Linux PIM in Ruby
I built such a program in 1983, in C on a PC - MSDOS 2.
When started from a 5 1/4" floppy disk with a search argument,
it displayed the first matching record in 1.2 seconds(!), from
a 2,500 record collection that filled the 1.2MByte disk. It was
designed for people answering phones, who wanted instant recall
of their caller’s details.
Anyhow, it used a B-tree index with the key structure being
"word/field#/record" so that all similar words were interleaved