[
https://issues.apache.org/jira/browse/JENA-1128?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Andy Seaborne resolved JENA-1128.
---------------------------------
Resolution: Fixed
Fix Version/s: Jena 3.1.0
> sdbquery doesn't work with MINUS
> --------------------------------
>
> Key: JENA-1128
> URL: https://issues.apache.org/jira/browse/JENA-1128
> Project: Apache Jena
> Issue Type: Bug
> Components: SDB
> Affects Versions: Jena 3.0.1
> Environment: jena-sdb-3.1.0-SNAPSHOT dated 2 Feb 2016
> apache-jena-3.1.0-SNAPSHOT dated 2 Feb 2016
> Reporter: Osma Suominen
> Assignee: Andy Seaborne
> Fix For: Jena 3.1.0
>
>
> I'm running a SPARQL query against a SDB loaded with SKOS data. The intent of
> the query is to check for broken links, i.e. skos:closeMatch relationships
> that point to nonexistent concepts in another SKOS dataset. I have simplified
> my query to a rather minimal test case below. In this case, also the remote
> data is included in the same graph for simplicity.
> Here is my test data:
> {noformat}
> @prefix skos: <http://www.w3.org/2004/02/skos/core#>.
> @prefix local: <http://example.com/local/>.
> @prefix remote: <http://example.com/remote/>.
> local:conceptA a skos:Concept ;
> skos:prefLabel "Local concept A"@en ;
> skos:note "has a valid link to an existing remote concept" ;
> skos:closeMatch remote:conceptC .
> local:conceptB a skos:Concept ;
> skos:prefLabel "Local concept B"@en ;
> skos:note "has a broken link to a nonexistent remote concept" ;
> skos:closeMatch remote:conceptD .
> remote:conceptC a skos:Concept ;
> skos:prefLabel "Remote concept C"@en .
> {noformat}
> This is my SPARQL query:
> {noformat}
> PREFIX skos: <http://www.w3.org/2004/02/skos/core#>
> SELECT * WHERE {
> ?local skos:closeMatch ?remote .
> MINUS { ?remote a skos:Concept }
> }
> {noformat}
> If I run the query using the command line tool "sparql" from the apache-jena
> distribution, it returns the correct result, i.e. the one concept with the
> broken link:
> {noformat}
> ------------------------------------------------------------------------------
> | local | remote |
> ==============================================================================
> | <http://example.com/local/conceptB> | <http://example.com/remote/conceptD> |
> ------------------------------------------------------------------------------
> {noformat}
> But when I load the above data into a SDB database (MySQL) and use sdbquery
> with the same SPARQL query, I get a different result (here run with the
> --debug option) which has an extra row:
> {noformat}
> PREFIX skos: <http://www.w3.org/2004/02/skos/core#>
> SELECT *
> WHERE
> { ?local skos:closeMatch ?remote
> MINUS
> { ?remote a skos:Concept }
> }
> - - - - - - - - - - - - - -
> SELECT -- V_3=?remote
> R_3.lex AS V_3_lex, R_3.datatype AS V_3_datatype, R_3.lang AS V_3_lang,
> R_3.type AS V_3_type
> FROM
> ( SELECT -- ?remote:(T_2.s=>T_2.X_1)
> T_2.s AS X_1
> FROM Triples AS T_2 -- ?remote
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> skos:Concept
> WHERE ( T_2.p = -6430697865200335348 -- Const:
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
> AND T_2.o = 1728757386496985884 -- Const: skos:Concept
> )
> ) AS T_2 -- ?remote:(T_2.s=>T_2.X_1)
> LEFT OUTER JOIN
> Nodes AS R_3 -- Var: ?remote
> ON ( T_2.X_1 = R_3.hash )
> (minus
> (SQL '''SqlSelectBlock/S_1 -- V_1=?local V_2=?remote
> R_1.lex/V_1_lex R_1.datatype/V_1_datatype R_1.lang/V_1_lang
> R_1.type/V_1_type
> R_2.lex/V_2_lex R_2.datatype/V_2_datatype R_2.lang/V_2_lang
> R_2.type/V_2_type
> Join/left outer
> Join/left outer
> SqlSelectBlock/T_1
> T_1.p = 2699241716664962559
> Table T_1 -- ?local skos:closeMatch ?remote
> Table R_1 -- Var: ?local
> Condition T_1.s = R_1.hash
> Table R_2 -- Var: ?remote
> Condition T_1.o = R_2.hash''')
> (SQL '''SqlSelectBlock/S_2 -- V_3=?remote
> R_3.lex/V_3_lex R_3.datatype/V_3_datatype R_3.lang/V_3_lang
> R_3.type/V_3_type
> Join/left outer
> SqlSelectBlock/T_2 -- ?remote:(T_2.s=>T_2.X_1)
> T_2.s/X_1
> T_2.p = -6430697865200335348
> T_2.o = 1728757386496985884
> Table T_2 -- ?remote
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> skos:Concept
> Table R_3 -- Var: ?remote
> Condition T_2.X_1 = R_3.hash''')
> )
> SELECT -- V_1=?local V_2=?remote
> R_1.lex AS V_1_lex, R_1.datatype AS V_1_datatype, R_1.lang AS V_1_lang,
> R_1.type AS V_1_type,
> R_2.lex AS V_2_lex, R_2.datatype AS V_2_datatype, R_2.lang AS V_2_lang,
> R_2.type AS V_2_type
> FROM
> ( SELECT *
> FROM Triples AS T_1 -- ?local skos:closeMatch ?remote
> WHERE ( T_1.p = 2699241716664962559 -- Const: skos:closeMatch
> )
> ) AS T_1
> LEFT OUTER JOIN
> Nodes AS R_1 -- Var: ?local
> ON ( T_1.s = R_1.hash )
> LEFT OUTER JOIN
> Nodes AS R_2 -- Var: ?remote
> ON ( T_1.o = R_2.hash )
> ------------------------------------------------------------------------------
> | local | remote |
> ==============================================================================
> | <http://example.com/local/conceptA> | <http://example.com/remote/conceptC> |
> | <http://example.com/local/conceptB> | <http://example.com/remote/conceptD> |
> ------------------------------------------------------------------------------
> {noformat}
> If I change the query to use FILTER NOT EXISTS instead of MINUS, then I get
> the correct result also with sdbquery:
> {noformat}
> PREFIX skos: <http://www.w3.org/2004/02/skos/core#>
> SELECT *
> WHERE
> { ?local skos:closeMatch ?remote
> FILTER NOT EXISTS { ?remote a skos:Concept }
> }
> - - - - - - - - - - - - - -
> (filter (notexists
> (SQL '''SqlSelectBlock/T_2 -- ?remote:(T_2.s=>T_2.X_1)
> T_2.s/X_1
> T_2.p = -6430697865200335348
> T_2.o = 1728757386496985884
> Table T_2 -- ?remote
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> skos:Concept''')
> )
> (SQL '''SqlSelectBlock/S_1 -- V_1=?local V_2=?remote
> R_1.lex/V_1_lex R_1.datatype/V_1_datatype R_1.lang/V_1_lang
> R_1.type/V_1_type
> R_2.lex/V_2_lex R_2.datatype/V_2_datatype R_2.lang/V_2_lang
> R_2.type/V_2_type
> Join/left outer
> Join/left outer
> SqlSelectBlock/T_1
> T_1.p = 2699241716664962559
> Table T_1 -- ?local skos:closeMatch ?remote
> Table R_1 -- Var: ?local
> Condition T_1.s = R_1.hash
> Table R_2 -- Var: ?remote
> Condition T_1.o = R_2.hash''')
> )
> SELECT -- V_1=?local V_2=?remote
> R_1.lex AS V_1_lex, R_1.datatype AS V_1_datatype, R_1.lang AS V_1_lang,
> R_1.type AS V_1_type,
> R_2.lex AS V_2_lex, R_2.datatype AS V_2_datatype, R_2.lang AS V_2_lang,
> R_2.type AS V_2_type
> FROM
> ( SELECT *
> FROM Triples AS T_1 -- ?local skos:closeMatch ?remote
> WHERE ( T_1.p = 2699241716664962559 -- Const: skos:closeMatch
> )
> ) AS T_1
> LEFT OUTER JOIN
> Nodes AS R_1 -- Var: ?local
> ON ( T_1.s = R_1.hash )
> LEFT OUTER JOIN
> Nodes AS R_2 -- Var: ?remote
> ON ( T_1.o = R_2.hash )
> SELECT *
> FROM Triples AS T_3 --
> <http://example.com/remote/conceptC>
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> skos:Concept
> WHERE ( T_3.s = 5972767169237582230 -- Const:
> <http://example.com/remote/conceptC>
> AND T_3.p = -6430697865200335348 -- Const:
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
> AND T_3.o = 1728757386496985884 -- Const: skos:Concept
> )
> SELECT *
> FROM Triples AS T_4 --
> <http://example.com/remote/conceptD>
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> skos:Concept
> WHERE ( T_4.s = 8175828786801660008 -- Const:
> <http://example.com/remote/conceptD>
> AND T_4.p = -6430697865200335348 -- Const:
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
> AND T_4.o = 1728757386496985884 -- Const: skos:Concept
> )
> ------------------------------------------------------------------------------
> | local | remote |
> ==============================================================================
> | <http://example.com/local/conceptB> | <http://example.com/remote/conceptD> |
> ------------------------------------------------------------------------------
> {noformat}
> However, in my actual query that this example is based on
> (https://github.com/NatLibFi/Finto-data/blob/master/tools/yso-updater-sparql/5-ysa-removed-concepts.rq)
> using FILTER NOT EXISTS is not an efficient solution, because the subtracted
> part uses a federated query and it will result in almost 30000 queries to be
> performed to the remote endpoint instead of just one.
> I'm using the most recent snapshots available from repository.apache.org.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
