Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!sol.ctr.columbia.edu!srcsip!msi-s0.msi.umn.edu!umeecs!heavy
From: heavy@zip.eecs.umich.edu (Richard Scott Hall)
Newsgroups: comp.sys.atari.st.tech
Subject: Data file indexing
Summary: Need a way to index data fro retrieval and listing
Keywords: Data Indexing
Message-ID: <1990Sep21.152926.28515@zip.eecs.umich.edu>
Date: 21 Sep 90 15:29:26 GMT
Sender: Richard Hall ( heavy@zip.eecs.umich.edu )
Distribution: all
Organization: University of Michigan EECS Dept., Ann Arbor, MI
Lines: 23

I am writing a point-of-sale system and need some advice.  I have
an inventory file with appoximately 7000 records in it and I need
a way to retrieve and list records in order quickly.  Right now I
have created a separate index file which basically is a binary
search tree of the key fields.  Each node in the index file has
a key and a record number of the inventory record.  This works
great for retrieving, I do a quick binary tree search on the index
file and get the record number, then I do a fseek in the inventory
file and read the record; it is really quick, on the average about
a second.  The problem comes in when I try to list in sequence,
I have modified the index file binary search tree so that it is
linked to previous and next nodes in order, so that way I can just
skip along the index file, but it is still slow to list because
I have to fseek in the index file and in the inventory file for
each record I list.  This is painfully slow on machines without
TOS 1.4.  Is there a better way of doing this?  I am will to scrap
my whole model if someone has a better idea that satisfies both
of my criterion...

Thanks in advance,

Richard Hall
University of Michigan