[infinispan-dev] Hash calculation in DefaultConsistentHash

[Alex Kluge]

Hi,

  I expect that using the Object hashCode, even within the same
 VM, will be problematic. Two instances of such a class will not have
 the same hashCode. Hence using one as a key to insert a value into
 the cache, then using the other to retrieve the value will not
 produce the expected results. This is true of pretty much any type
 of hash table, distributed or not. Your key can not use the Object
 hashCode method.

  This is also also relevant to your discussion of using binary data (I
 presume a byte array) in the cache. You can use an array as a hash key
 in any meaningful way. However, it is perfectly reasonable to wrap the
 array in an object that computes the hash from the array contents. I
 know this from practical experience implementing a distributed version
 of JBoss Cache.

                                             Alex

--- On Tue, 10/13/09, Jason T. Greene <jason.greene at redhat.com> wrote:

> From: Jason T. Greene <jason.greene at redhat.com>
> Subject: Re: [infinispan-dev] Hash calculation in DefaultConsistentHash
> To: "infinispan -Dev List" <infinispan-dev at lists.jboss.org>
> Date: Tuesday, October 13, 2009, 12:41 PM
> 
> 
> Manik Surtani wrote:
> > On 13 Oct 2009, at 15:30, Jason T. Greene wrote:
> > 
> >> We've always had that requirement. Hes referring
> to classes which do  
> >> not
> >> define a hashCode, and thus inherit
> Object.hashCode(), which  
> >> produces a
> >> number based on the reference pointer which is
> different per-vm.
> > 
> > Sure, but for folks that don't expect to be able to
> retrieve a key  
> > from anywhere, it should still work.  But that is
> a weird use case  
> > anyway, unless the cache is in local mode in which
> case DIST doesn't  
> > apply.
> > 
> 
> Right, you can do it in local mode. His hashcode routine
> would of course 
> still work for that, that is if you wanted to have
> datacontainer alwyas 
> use it (probably a good idea).
> 
> -- 
> Jason T. Greene
> JBoss, a division of Red Hat
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>