Xref: utzoo comp.theory:402 comp.misc:8369 comp.lang.misc:4298
Path: utzoo!censor!geac!torsqnt!jarvis.csri.toronto.edu!cs.utexas.edu!samsung!zaphod.mps.ohio-state.edu!mips!apple!rutgers!bellcore!spectral!sjs
From: sjs@spectral.ctt.bellcore.com (Stan Switzer)
Newsgroups: comp.theory,comp.misc,comp.lang.misc
Subject: Re: Modulus (Re: hashing function for strings)
Message-ID: <20457@bellcore.bellcore.com>
Date: 1 Mar 90 15:44:01 GMT
References: <2367@castle.ed.ac.uk> <12077@goofy.megatest.UUCP> <Er$cfg@cs.psu.edu>
Sender: news@bellcore.bellcore.com
Reply-To: sjs@bellcore.com (Stan Switzer)
Organization: Bellcore
Lines: 22

In article <Er$cfg@cs.psu.edu> flee@shire.cs.psu.edu (Felix Lee) writes:
> Dave Jones <djones@megatest.UUCP> wrote:
> >I differ. The first of these, -5 div 3 == -1 is wrong.
> 
> The contrary view is that abs(a/b) != abs(-a/b) is strange.  It
> depends on what you're doing.  I have no problems with either stance.

Put up or shut up.

I have often found that the anomalous behavior of "modulus"
at zero has caused me extra work to make it do the "right" thing:
  -5 div 3 == -2
  -5 mod 3 == 1
This has occurred in graphics applications, in communications handlers
(computing sliding windows using a wrap-around counter), and others.

Please, somebody, show me a single example where the other behavior
is actually USEFUL!  We are not concerned with whether your grade-school
education [ -a/b == -(a/b) ] applies to computers, but with what
actually is useful in integer computation in actual algorithms.

Stan Switzer  sjs@bellcore.com