W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > July to September 2004

Re: New requirement: disjunction

From: Dan Connolly <connolly@w3.org>
Date: Tue, 20 Jul 2004 10:54:58 -0500
To: Yoshio Fukushige <fukushige.yoshio@jp.panasonic.com>
Cc: Rob Shearer <Rob.Shearer@networkinference.com>, RDF Data Access Working Group <public-rdf-dawg@w3.org>, fuku@w3.org
Message-Id: <1090338897.14421.1350.camel@dirk>

On Tue, 2004-07-20 at 02:41, Yoshio Fukushige wrote:
> Hello,
> 
> As a follow-up for my comments in the f2f meeting on the current wording of
> the disjunction requirement:
> #I wish I could have said this in good English
> 
> 
> The original wording says:
> 
> [[[
> The query language must include the capability to restrict matches on a
> queried graph based on a disjunction of graph patterns, at least one of
> which must be satisfied by the query.
> ]]]
> 
> My first concern is that the notion of "a disjunction of graph patterns" is
> not clear.

It's quite conventional.

http://en.wikipedia.org/wiki/Disjunction
http://plato.stanford.edu/entries/disjunction/

We could even be more conventional by moving from "graph patterns"
to "graph expressions", to save ourselves the sort of nightmare
that "RDF model" has become.

pattern isn't all that unconventional, though.



> It cannot mean "the union of the graph patterns," for the union of the
> triples means
> the conjunction of the conditions represented by the graph patterns
> (here, the word "union" is used in the same way it is used in the set
> theory).
>
> However, the notion of disjunction goes usually with the notion of union, so
> at least
> it is quite confusing.


"the disjunction of P and Q" is related to, but not the
same as "the union of P and Q".

The relationship between disjuction and union is, pretty much:

	results(P or Q, G) = union(results(P, G), results(Q, G))

where results(E, G) is the result of evaluating the query
expression E with respect to the input graph G.

(I say "pretty much" because I'm looking at the simple case
where we're only looking at the graph structure of G, not doing
anything special in case G includes an owl:union or anything like that.)

> Is it a graph pattern among those in the query that are connected
> by one or more disjunction oprator(s)? (*)

I don't think it helps to think of a query expression (aka
graph pattern) as a graph.

> But I'm not sure if it can be  called "a disjunction" itself.
> 
> Second,  I'm puzzled by the article "a" in the phrase "a disjunction."
> The only possible reading is  the (*) above, but I'm not sure if such
> wording is possible.
> 
> Third, I can't find the anteceedent of the relative pronoun "which" in
> "at least one of which must be satisfied by the query."

it refers to one of the disjuncts in the disjunction.

> If it is "graph patterns,"

yes.

>  then I cannot understand what is meant by
> "one of the graph patterns must be satisfied by the query."

Perhaps one of the articles cited above helps?


> If the antecedent is "matches,"

no.

>  I don't think the matches are well-specified
> in a situation where
> "at least one of the matches must be satisfied by the query,"
> where the matches can contain arbitrary thing.
> 
> I enven would like to say that I don't understand  what is meant when saying
> "something is satified by the query."
>
> A query is a thing to be satified, rather than a thing to satify something
> else, isn't it?

Ah, yes... we talked about that... it should be

  "... is satisfied by the queried graph".

a (cunjunctive) query expression E is satisfied by a graph G if
there is some assignment of the variables in E to terms so that
all the conjuncts in the result of the substitution are in G.

> As such the original wording is quite confusing.
> 
> So, my alternative suggestion is
> [[[
> The query language must include the capability to restrict matches

I don't think it helps to describe disjunction as restricting matches.

>  on a
> queried graph to match at least one of the graph patterns in the query
> connected by one or more disjunction operator(s).
> ]]]
> 
> What do you think?

I think clarifying "by the query" to "by the queried graph" is
worthwhile:

[[
The query language must include the capability to restrict matches on a
queried graph based on a disjunction of graph patterns, at least one of
which must be satisfied by the queried graph.
]]

> Best,
> Yoshio
> fukushige.yoshio@jp.panasonic.com

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Tuesday, 20 July 2004 11:54:20 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:20 GMT