Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.java.programmer > #17711

Re: hashCode

Show key headers only | View raw

  To: Roedy Green
From: Peter Duniho <NpOeStPeAdM@NnOwSlPiAnMk.com>

On Sat, 11 Aug 2012 04:54:09 -0700, Roedy Green wrote:

> On Fri, 10 Aug 2012 12:45:07 -0700 (PDT), Lew <lewbloch@gmail.com>
> wrote, quoted or indirectly quoted someone who said :
>
>>        h =3D 31 * h + attribute.hashCode();
>>   }
> In my essay I recommend XOR which is an inherentely  faster operation
> than multiply.  I wonder which actually works out better. If you had a
> large number of fields, the multiply effect could fall off the left
> hand end.  It is the algorithm used for String which could have  very
> long strings, so Sun must have thought of that.

Lew's example is better than XOR. I think you should fix your essay.

There are of course even better approaches if one knows something specific 
about the input data. But the "multiply previous by prime, add next value" is a 
reasonably standard, "pretty good" composition function.

XOR a well-known problems in the context of hash codes: in many situations, 
large numbers of combinations of values wind up mapping to the same result. 
Consider a pair of integers: any time those values are identical, the result of 
XOR is 0. On the other hand, with the multiply-and-add approach, it's much less 
common for "typical" integer values to combine to a hash value of 0.

(Integers are an important scenario, because a 32-bit integer naturally is its 
own hash value...there's no reason to try to mix the bits any further as is 
done with other data types, so they don't).

Now (unless I've done my math wrong) if your component hash codes are actually 
well-distributed across the entire range of a 32-bit integer (i.e. are 
completely uncorrelated to each other and have entirely random distribution), 
the XOR should work fine.  It's just that it's very susceptible to creating 
hash value collisions when the input values aren't themselves well-distributed 
and so as a general technique it's not nearly as good as multiply-and-add.

As far as performance goes, even if the multiply-and-add costs more (and as 
Joerg points out, this may only be a 1-cycle vs 2-cycle difference anyway on 
that specific operation), your code should not be spending a lot of time 
computing hash values in the first place.  If that's a significant bottleneck, 
there are better ways to improve performance than worrying about XOR vs 
multiply-and-add, especially since a poor hash function can hurt performance 
much worse than using the wrong math operation would.

Pete

--- BBBS/Li6 v4.10 Dada-1
 * Origin: Prism bbs (1:261/38)
--- Synchronet 3.16a-Win32 NewsLink 1.98
Time Warp of the Future BBS - telnet://time.synchro.net:24

Back to comp.lang.java.programmer | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

hashCode "bob smith" <bob.smith@1:261/38.remove-t9h-this> - 2012-08-10 18:39 +0000
  Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-t9h-this> - 2012-08-10 18:39 +0000
    Re: hashCode "markspace" <markspace@1:261/38.remove-t9h-this> - 2012-08-10 18:39 +0000
      Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-t9h-this> - 2012-08-10 18:39 +0000
        Re: hashCode "rossum" <rossum@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
  Re: hashCode "Eric Sosman" <eric.sosman@1:261/38.remove-t9h-this> - 2012-08-10 18:39 +0000
    Re: hashCode "bob smith" <bob.smith@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
      Re: hashCode "Lew" <lew@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
        Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
          Re: hashCode "Lew" <lew@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
            Re: hashCode "Arne Vajhøj" <������
høj@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
              Re: hashCode "Arne Vajhøj" <������
høj@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
                Re: hashCode "Eric Sosman" <eric.sosman@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
                Re: hashCode "Arne Vajhøj" <������
høj@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
      Re: hashCode "Eric Sosman" <eric.sosman@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
        Re: hashCode "Lew" <lew@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
          Re: hashCode "Arne Vajhøj" <������
høj@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
      Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
        Re: hashCode "Robert Klemme" <robert.klemme@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
          Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-nlb-this> - 2012-08-13 18:36 +0000
            Re: hashCode "Robert Klemme" <robert.klemme@1:261/38.remove-nlb-this> - 2012-08-13 18:36 +0000
              Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-fzq-this> - 2012-08-20 18:58 +0000
              Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-fzq-this> - 2012-08-20 18:58 +0000
      Re: hashCode "Patricia Shanahan" <patricia.shanahan@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
      Re: hashCode "Joshua Cranmer" <joshua.cranmer@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
        Re: hashCode "Patricia Shanahan" <patricia.shanahan@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
          Re: hashCode "Eric Sosman" <eric.sosman@1:261/38.remove-6fd-this> - 2012-08-12 20:00 +0000
            Re: hashCode "Patricia Shanahan" <patricia.shanahan@1:261/38.remove-6fd-this> - 2012-08-12 20:00 +0000
          Re: hashCode "Arne Vajhøj" <������
høj@1:261/38.remove-nlb-this> - 2012-08-13 18:36 +0000
        Re: hashCode "Jan Burse" <jan.burse@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
  Re: hashCode "Roedy Green" <roedy.green@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
    Re: hashCode "Lew" <lew@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
      Re: hashCode "Roedy Green" <roedy.green@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
        Re: hashCode "Joerg Meier" <joerg.meier@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
          Re: hashCode "Mike Winter" <mike.winter@1:261/38.remove-ya-this> - 2012-08-11 19:19 +0000
        Re: hashCode "Peter Duniho" <peter.duniho@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
        Re: hashCode "Jan Burse" <jan.burse@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
        Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
      Re: hashCode "rossum" <rossum@1:261/38.remove-ya-this> - 2012-08-11 19:19 +0000
    Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
      Re: hashCode "Eric Sosman" <eric.sosman@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
        Re: hashCode "Arne Vajhøj" <arne.vajhøj@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
  Re: hashCode "Jan Burse" <jan.burse@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
    Re: hashCode "Jan Burse" <jan.burse@1:261/38.remove-pjg-this> - 2012-08-11 18:17 +0000
    Re: hashCode "Lew" <lew@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
      Re: hashCode "Lew" <lew@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
      Re: hashCode "Arne Vajhøj" <������
høj@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
        Re: hashCode "Lew" <lew@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
    Re: hashCode "Jan Burse" <jan.burse@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
      Re: hashCode "Jan Burse" <jan.burse@1:261/38.remove-m2z-this> - 2012-08-12 18:58 +0000
      Re: hashCode "Lew" <lew@1:261/38.remove-6fd-this> - 2012-08-12 20:00 +0000
        Re: hashCode "Arne Vajhøj" <������
høj@1:261/38.remove-nlb-this> - 2012-08-13 18:36 +0000

csiph-web