From: remove-this-spam-blocker.bstowers@pobox.com (Bradley D. Stowers)
Subject: Re: Hash function
Date: 16 Dec 1997 00:00:00 GMT
Message-ID: <3496e29d.1985986@forums.borland.com>
Content-Transfer-Encoding: 7bit
References: <34922A30.4D59@usa.net>
Content-Type: text/plain; charset=us-ascii
Organization: None
Mime-Version: 1.0
Newsgroups: borland.public.delphi.winapi


On Fri, 12 Dec 1997 22:24:48 -0800, "Horatiu A. Tanescu" <t-horatio@usa.net>
wrote:

>I have a list of long file names and I add file names to this list only
>if the file doesn't already exists in the list. I can't sort the list to
>perform a fast binary search so I decided to use a hash technique to
>compare the file name with all the files in the list. For each file in
>the list I compute an integer value (e.g. the sum of all characters in
>name) and when I search if the file name exists in list I have to
>compare only integer values (quite faster).
>
>The problem is the hash function (that converts the file name into an
>integer value). If I compute the sum of all characters in name I could
>have a large number of strings with the same associated integer value.
>Is there a better choice for the hash function?

For simple hashing, you might want to consider the Windows API Atom functions
(InitAtomTable, AddAtom, DeleteAtom, GetAtomName, etc.).  They are really just
simple hash-type functions.  The are limited in that the strings must be 255
or less characters, and they are not case sensitive.  This may be a problem
for you since I *think* long file names can be 255 characters without their
path.


Regards,
Brad Stowers
Delphi Free Stuff
http://www.pobox.com/~bstowers/delphi/