[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

Re: [ba-ohs-talk] new version of graph model paper posted


On Wed, 2002-04-17 at 10:17, Peter Jones wrote:
> Lee Iverson wrote:
> > I'm not sure I follow.  What kinds of queries would lead to these
> > "bifurcations"?  In a general sense, since equality of two nodes is
> > easy to determine, it is fairly simple to ensure that in any query
> > result multiple references to the same node get properly resolved.
> 
> I was thinking of graph structures like the following:
> 
> A--------------B1
> |              |
> |              |
> |              |
> B2-------------C
> 
> Where B1 and B2 are not identical nodes, but nonetheless contain the
> same value in respect of the particular
> property being queried for.
> So if I express my query as A-[B*]-C then some sort of decision has
> to be made about what the result should be: Two B's or two C's or what?.
> (One way would be to bin the idea of ordinary documents altogether and
> just have hyperdocuments.)    (01)

So, your query is asking for configurations of the form:
	A -> B* -> C    (02)

As far as I see it, the query engine should return two results:
	A -> B1 -> C
	A -> B2 -> C
and care must be taken that both references to C are returned in such
a way that the client can easily determine that they resolve to
the same value.  In XML this might be accomplished with something
like:    (03)

<resultset>
  <A> 
   <B1> <C name="c0"/> </B1>
   <B2> <C href="#c0"/> </B2>
  </A>
</resultset>    (04)

Am I missing something?    (05)

-- 
-------------------------------------------------------------------------------
Lee Iverson                    
leei@ece.ubc.ca                Dept of ECE, 2356 Main Mall
http://www.ai.sri.com/~leei/   Vancouver BC Canada V6T 1Z4
Office: (604) 822-3381    (06)