W3C home > Mailing lists > Public > public-rdf-dawg-comments@w3.org > November 2004

Re: Comments on SPARQL (based on an SPARQL Engine implementation in Python)

From: Seaborne, Andy <andy.seaborne@hp.com>
Date: Sun, 07 Nov 2004 17:57:54 +0000
Message-ID: <418E6222.10902@hp.com>
To: Ivan Herman <ivan@w3.org>
Cc: public-rdf-dawg-comments@w3.org



Ivan Herman wrote:
> Andy,
> 
> thanks for the reply. Some (small) answer/remarks below
> 
> Seaborne, Andy wrote:
> 

<snip/>

>>
>> This is an area where the working draft says very little and it is being
>> worked on at the moment.
>>
>> Constraints, like triple patterns, are restrictions on the results that
>> match the query.  For any match, there is a set of bindings and
>> constraints evaluated on these function must be true.
>>
>> A function will receive its variables as arguments (and the variables
>> may be unbound due to optional, say - the function will have to cope).
>> The actual mechanism for doing that will be implementation dependent -
>> executing all the constraints after the triple pattern matching would be
>> correct but may be slower than executing a constraints as soon as its
>> variables can be bound, e.g., just after the point where all varibales
>> used have been mentioned in patterns as per your per-pattern
>> constraints.  As a query engine may choose to match the triple patterns
>> in any order, or in groups, there is lots of scope for implementation
>> optimization here.
>>
> 
> Understood. Actually, this is in line with what I implemented: on the 
> 'engine' level there is a possibility to call a constraint method per 
> triplets as well as for global, and an SQRL parser may optimize by 
> choosing the former when appropriate

Yes - there is a large body of work on data flow analysis that some 
engines may usefully dip into.

> 
> 
>>
>>> ---------------
>>>
>>> Nested Patterns. It is not clear in the draft how 'deep' nesting can
>>> go. I did only the simple one, ie, only a one level depth is managed:
>>>
>>> (?a,?b,c), {(?q,?w?,?r),(?s,t,?u)}
>>>
>>> It is not clear whether a nesting of the kind:
>>>
>>> (?a,?b,c), {(?q,?w?,?r),{(?s,t,?u),(?q,k,?o)}}
>>>
>>> is also allowed or not. Actually, if it is, it has to be defined what
>>> it really *means*.
>>
>>
>>
>> In that example, it is all conjunction and the same as:
>>
>> (?a,?b,c) (?q,?w?,?r) (?s,t,?u) (?q,k,?o)
>>
>>
> 
> 
> Then I am lost.:-( My understanding was that (just referring to Ti-s for 
> triples):
> 
> [T1 {T2 T3}] means [T1 T2] OR [T1 T2]
> 

Clarification first - [] is optional matching.

> so why would
> 
> [T1 {T2 {T3 T4}}] mean [T1 T2 T3 T4] as you write? What is then the 
> difference between
> 
> [T1 {T2 {T3 T4}}] and [T1 {T2 T3 T4}] or indeed [T1 T2 T3 T4]?

Nothing because the {} is just a set of triple patterns and juxaposition 
is conjunction so T1 {T2 {T3 T4}} is  "T1 AND (T2 AND (T3 AND T4))" which 
is the same as the other forms.

> 
> My interpretation of [T1 {T2 {T3 T4}}] would be
> 
> 
> [T1 T2] OR [T1 T3 T4]
> 
> what am I missing?
> 
>>
>>> (My initial thoughts are that it means an
>>> alternation of 'or'-s and 'and'-s, it means
>>> (?a,?b,c) and (?q,?w?,?r)
>>> or
>>> (?a,?b,c) and (?s,t,?u) and (?q,k,?o)
>>>
>>> etc, recursively.) If this is adopted, it has to be described.
>>
>>
>>
>> Noted.
>>
>>
>>> -----------------
>>>
>>> Optional Patterns. I was not clear to me *why* there might be more
>>
>>
>> than
>>
>>> one optional patterns, whereas there is only *one* where pattern. Why
>>> the asymmetry? 
>>
>>
>>
>> There may be more than one optional because there may be indendent
>> optionalk information:
>>
>> PREFIX foaf: <http://xmlns.com/foaf/0.1/>
>> SELECT ?name ?mbox ?hpage
>> WHERE  ( ?x foaf:name  ?name )
>>        [ ( ?x foaf:mbox ?mbox ) ]
>>        [ ( ?x foaf:homepage ?hpage ) ]
>>
>> Here, "mbox" and "hpage" get added to the results independently of each
>> other.
>>
> 
> I presume you meant
>  > PREFIX foaf: <http://xmlns.com/foaf/0.1/>
>  > SELECT ?name ?mbox ?hpage
>  > WHERE  ( ?x foaf:name  ?name )
>  > OPTIONAL  [ ( ?x foaf:mbox ?mbox ) ]
>  >           [ ( ?x foaf:homepage ?hpage ) ]

Not quite : [] is the optional block itself not merely statement grouping.

OPTIONAL { T1 T2 } is the same as [ T1 T2 ]

> 
> but is the difference between that and
> 
> PREFIX foaf: <http://xmlns.com/foaf/0.1/>
> SELECT ?name ?mbox ?hpage
> WHERE  ( ?x c  ?name )
> OPTIONAL  { ( ?x foaf:mbox ?mbox ) ( ?x foaf:homepage ?hpage ) }

Consider the dataset:

_:a   foaf:name   "Sally" .
_:a    foaf:mbox   <mailto:ceo@example.org> .

then

PREFIX foaf: <http://xmlns.com/foaf/0.1/>
SELECT ?name ?mbox ?hpage
WHERE  ( ?x foaf:name  ?name )
        [ ( ?x foaf:mbox ?mbox ) ]
        [ ( ?x foaf:homepage ?hpage ) ]

has a solution:

?name = "Sally"  ?mbox=<mailto:ceo@example.org>  ?hpage unset

whereas:

OPTIONAL  { ( ?x foaf:mbox ?mbox ) ( ?x foaf:homepage ?hpage ) }

requires both triples to match in order to match so the solution is:

?name = "Sally"  ?mbox unset   ?hpage unset

It would not have a solution with just ?mbox bound because to match
{ T1 T2 } both T1 and T2 must match.

I hope that is a bit clearer.

<snip/>

	Andy
Received on Sunday, 7 November 2004 17:58:29 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:14:47 GMT