W3C home > Mailing lists > Public > public-sparql-dev@w3.org > January to March 2016

RE: BIND behaviour and GroupGraphPatterns

From: Niels Andersen <niels@thinkiq.com>
Date: Thu, 17 Mar 2016 15:57:18 +0000
To: Andy Seaborne <andy@apache.org>, "public-sparql-dev@w3.org" <public-sparql-dev@w3.org>
Message-ID: <BN3PR1101MB1185905E61F807BE8A1EF5CBA08B0@BN3PR1101MB1185.namprd11.prod.outlook.com>
I had exactly the same issue yesterday.

This article explains the issue with bottom up semantics and variable scope: https://wiki.blazegraph.com/wiki/index.php/SPARQL_Bottom_Up_Semantics


Niels

-----Original Message-----
From: Andy Seaborne [mailto:andy@apache.org] 
Sent: Thursday, March 17, 2016 04:43
To: public-sparql-dev@w3.org
Subject: Re: BIND behaviour and GroupGraphPatterns

On 17/03/16 10:16, Jörn Hees wrote:
> Hi,
>
> i have a question about the "intended scope" of BIND...
>
> Trying to fix an issue in the RDFLib SPARQL processor [1], i came up with what i thought was a minimal breaking test [2] like this:

Thank you for providing a test.

>
> This works (correctly binds ?z in RDFLib):
> select * {
>   ?s ?p ?o .
>   bind(?o as ?z)
> }

Algebra:

(extend ((?z ?o))
   (bgp (triple ?s ?p ?o)))

>
> This doesn't:
> select * {
>   ?s ?p ?o .
>   { bind(?o as ?z) }
> }

Algebra:

(join
   (bgp (triple ?s ?p ?o))
   (extend ((?z ?o))
     (table unit)))

The issue is what value ?z is being bound to.

The addition of {} means that the the BIND is run on what's before it in that {}-group, then joined with the rest of the outer {}-group. SPARQL evaluates the inner {}-groups and then combines the results.

{ bind(?o as ?z) } does not have ?o bound to anything -> evaluation error.

Try this:

select * { bind(?o as ?z) }

	Andy

>
> (At this time, my expectation was that the latter should also bind ?z. 
> Later, I found out that jena disagrees (see appendix).)
>
> So i went down in our implementation and fixed the behaviour, making my tests for [1] pass.
>
> To my surprise the fix caused another test to fail, namely bind07 [3] from the SPARQL 1.1 test cases:
>
> SELECT ?s ?p ?o ?z {
>   ?s ?p ?o .
>   { BIND(?o+1 AS ?z) } UNION { BIND(?o+2 AS ?z) } }
>
> The test asserts that ?z is unbound.
> I guess that the intention of that test was to make sure ?z isn't bound to different values.
>
> What i find weird about this is that it's inconsistent with other ways to bind ?z to different things which are allowed:
>
> SELECT ?s ?p ?o ?z {
>   ?s ?p ?o .
>   { ?s ?p ?z } UNION { ?z ?p ?o }
> }
>
>
> So my questions are:
> - Is [3] correct? What is the motivation?
> - Should ?z be bound in this: select * { ?s ?p ?o .  { bind(?o as ?z) } }  ?
>
>
> Cheers,
> Jörn
>
> [1]: https://github.com/RDFLib/rdflib/issues/580

> [2]: https://github.com/RDFLib/rdflib/pull/601

> [3]: 
> https://www.w3.org/2009/sparql/docs/tests/data-sparql11/bind/bind07.rq

>
>
>
> Appendix: some tests in jena on the commandline
>
>
> This works as expected:
> $ sparql --data=https://www.w3.org/2009/sparql/docs/tests/data-sparql11/bind/data.ttl --query=<(echo '
> select * {
>   ?s ?p ?o .
>   bind(?o as ?z)
> }
> ')
> ------------------------------------------------------------
> | s                       | p                      | o | z |
> ============================================================
> | <http://example.org/s1> | <http://example.org/p> | 1 | 1 | 
> | <http://example.org/s4> | <http://example.org/p> | 4 | 4 | 
> | <http://example.org/s3> | <http://example.org/p> | 3 | 3 | 
> | <http://example.org/s2> | <http://example.org/p> | 2 | 2 |
> ------------------------------------------------------------
>
>
> This doesn't:
> $ sparql --data=https://www.w3.org/2009/sparql/docs/tests/data-sparql11/bind/data.ttl --query=<(echo '
> select * {
>   ?s ?p ?o .
>   { bind(?o as ?z) }
> }
> ')
> ------------------------------------------------------------
> | s                       | p                      | o | z |
> ============================================================
> | <http://example.org/s1> | <http://example.org/p> | 1 |   |
> | <http://example.org/s4> | <http://example.org/p> | 4 |   |
> | <http://example.org/s3> | <http://example.org/p> | 3 |   |
> | <http://example.org/s2> | <http://example.org/p> | 2 |   |
> ------------------------------------------------------------
>
>
> Test [3]:
> $ sparql 
> --data=https://www.w3.org/2009/sparql/docs/tests/data-sparql11/bind/da

> ta.ttl 
> --query=https://www.w3.org/2009/sparql/docs/tests/data-sparql11/bind/b

> ind07.rq
> --------------------
> | s   | p  | o | z |
> ====================
> | :s1 | :p | 1 |   |
> | :s4 | :p | 4 |   |
> | :s3 | :p | 3 |   |
> | :s2 | :p | 2 |   |
> | :s1 | :p | 1 |   |
> | :s4 | :p | 4 |   |
> | :s3 | :p | 3 |   |
> | :s2 | :p | 2 |   |
> --------------------
>
>
> This binds ?z:
> $ sparql --data=https://www.w3.org/2009/sparql/docs/tests/data-sparql11/bind/data.ttl --query=<(echo '
> select * {
>   ?s ?p ?o .
>   {{ ?s ?p ?z } UNION { ?z ?p ?o }}
> }
> ')
> ----------------------------------------------------------------------
> ------------
> | s                       | p                      | o | z                       |
> ======================================================================
> ============
> | <http://example.org/s1> | <http://example.org/p> | 1 | 1                       |
> | <http://example.org/s1> | <http://example.org/p> | 1 | <http://example.org/s1> |
> | <http://example.org/s4> | <http://example.org/p> | 4 | 4                       |
> | <http://example.org/s4> | <http://example.org/p> | 4 | <http://example.org/s4> |
> | <http://example.org/s3> | <http://example.org/p> | 3 | 3                       |
> | <http://example.org/s3> | <http://example.org/p> | 3 | <http://example.org/s3> |
> | <http://example.org/s2> | <http://example.org/p> | 2 | 2                       |
> | <http://example.org/s2> | <http://example.org/p> | 2 | 
> | <http://example.org/s2> |
> ----------------------------------------------------------------------
> ------------
>
>
>


Received on Thursday, 17 March 2016 16:13:10 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 17 March 2016 16:13:10 UTC