openRDF.org Issue Tracker

[SES-1109] SPARQL 1.1 HAVING when used with aggrateions fails with "Node is not a child node"

Thu, 4 Oct 2012 22:20:26 +0200 (CEST)

The following query results in an error

SELECT ?p (COUNT(?o) as ?objects) {
[] ?p ?o .
}

GROUP BY ?p
HAVING(?objects > 1)
ORDER BY DESC(?objects)

Query error: Unexpected failure getting metadata for query 'QueryDescription ' SELECT ?p (COUNT(?o) as ?objects) { [] ?p ?o . } GROUP BY ?p HAVING(?objects > 0) ORDER BY DESC(?objects) ' (SPARQL, QueryModifier (100, 0, {}), impacts)' caused by : Node is not a child node: Group (p) StatementPattern Var (name=-anon-1, anonymous) Var (name=p) Var (name=o) GroupElem Count Var (name=o)

[SES-1108] GROUP BY aggregates incorrectly with nothing to aggregate

Wed, 26 Sep 2012 02:32:44 +0200 (CEST)

With an empty (in-memory) repository, the following query:

SELECT ?type (COUNT(DISTINCT ?instance) AS ?count) WHERE {
?instance <http://example.com#predicate> ?type .
}
GROUP BY ?type

Returns:

Type Count
(null) 0

I believe it should not return anything at all.

[SES-1107] Paging through query results causes URL to keep growing

Thu, 4 Oct 2012 01:49:52 +0200 (CEST)

The JavaScript responding to form actions in the query results page always causes the URL to grow in length. For larger query texts, his has the potential to make the URL grow too long to handle.

The code is always appending new copies of URL query parameters. (This presently works because the same code uses only the last value in the URL.) It should test for the presence of the changing parameter in the URL and replace it.

[SES-1105] Query Page javascript broken on Internet Explorer 9

Thu, 27 Sep 2012 22:29:37 +0200 (CEST)

There is an issue where Internet Explorer 9 HTML-escapes characters in the JavaScript included in the XSL template files. Specifically a test using the '>' operator gets broken by a replacement with >. This particular instance of the issue doesn't cause a problem. However, in recent checkins for ~~SES-376~~, I have introduced more JavaScript into the tuple results XSL template with '<' and '>' comparisons. This difference in IE 9 breaks proper display of results.

I have already coded and tested a fix which simply moves the JavaScript from these files into separate *.js files in the scripts folder.

[SES-1102] Mishandling of brackets in Property Paths

Tue, 30 Oct 2012 23:48:02 +0100 (CET)

The following programm executes two sparql expressions which to me should both retrieve the same subject and object. But they do not.

public static void main(String[] args) throws Exception {
Repository repo = new SailRepository(new MemoryStore());
repo.initialize();

ValueFactory f = repo.getValueFactory();

URI a = f.createURI("urn:a");
URI b = f.createURI("urn:b");
URI c = f.createURI("urn:c");

RepositoryConnection con = repo.getConnection();
con.add(a, f.createURI("property:x"), b);
con.add(b, f.createURI("property:y"), c);

//To me, these queries should result in the same output
String[] queries = new String[2];
queries[0] = "SELECT ?m ?n WHERE { ?m <property:x>/<property:y> ?n }";
queries[1] = "SELECT ?m ?n WHERE { ?m <property:x>/(<property:y>) ?n }";

int i = 0;
for (String q : queries) {
TupleQuery query = con.prepareTupleQuery(QueryLanguage.SPARQL, q);
TupleQueryResult result = query.evaluate();

String resultStr = "No Result";
while (result.hasNext()) {
BindingSet bindingSet = result.next();
Value m = bindingSet.getValue("m");
Value n = bindingSet.getValue("n");
resultStr = "" + m + " | " + n;
}
System.out.println(i + ": " + resultStr);
i++;
}
con.close();
}

It results in the following output:
0: urn:a | urn:c
1: No Result

I would expect it to result in the following output:
0: urn:a | urn:c
1: urn:a | urn:c

[SES-1100] openrdf-sesame.war webapp resets http.proxyHost JVM-wide system property

Related — Wed, 31 Oct 2012 21:48:41 +0100 (CET)

Upon container start-up, the openrdf-sesame.war resets the http.proxyHost JVM-wide system property as well as the other proxy-related system properties.

It should not do so or, at least, leave a way for the system (administrator) to manage the proxy settings directly (e.g. Mac OS X manages proxies at system level and taken into account by the JVM as "java.net.useSystemProxies" defaults to true on this platform).
For example, proxy-related system properties should not be modified when Sesame's configuration property "proxies.enabled" is set to false.

Proposed patch: modify info.aduna.app.net.ProxySettings.setProxySystemProperty() as follows:

private void setProxySystemProperty(String key, String val) {
if (getProxiesEnabled()) {
setSystemProperty(key, val);
}
// No Sesame-specific proxy configuration => leave system configuration unchanged.
// else {
// setSystemProperty(key, null);
// }
}

Web applications should never alter the system-properties set at the command-line unless explicitely instructed to do so (which thus can not be the default behaviour).

For the long term, if Sesame needs an specific proxy configuration that differs from the system default, it should create an application-wide java.net.Proxy object with this configuration and use it when establishing network connections.

Note: this bug was created with priority Major as Sesame proxy configuration is not described anywhere (a web search on "proxies.enabled" brings back nothing), resulting in hours lost in useless debugging.

[SES-1096] Position of BIND() affects query results

Followup — Fri, 14 Sep 2012 23:58:59 +0200 (CEST)

Being a declarative language, I would not expect the ordering of SPARQL clauses to affect the query results. However, BIND is a bit strange, so perhaps I have misunderstood. This might be the correct behaviour.

Steps to reproduce:

1. Create new in memory store (no inference)
2. Insert a simple class hierarchy of 3 classes:

PREFIX rdfs:<http://www.w3.org/2000/01/rdf-schema#>
PREFIX my: <http://ex#>
INSERT DATA {
  my:Concept rdfs:subClassOf my:Super1 .
  my:Super1 rdfs:subClassOf my:Super2 .
}

3. Execute query:

PREFIX rdfs:<http://www.w3.org/2000/01/rdf-schema#>
PREFIX my: <http://ex#>
SELECT ?super {
  BIND( my:Concept as ?concept )
  ?concept rdfs:subClassOf ?super .
  OPTIONAL {
    ?concept rdfs:subClassOf ?inbetweener .
    ?inbetweener rdfs:subClassOf ?super .
    FILTER( ?inbetweener != ?concept && ?inbetweener != ?super )
  }
  FILTER( ! BOUND(?inbetweener) && ?super != ?concept)
}

Gives expected result: super=<http://ex#Super1>

4. Execute the same query, but move the BIND clause from the top to the bottom:

PREFIX rdfs:<http://www.w3.org/2000/01/rdf-schema#>
PREFIX my: <http://ex#>
SELECT ?super {
  ?concept rdfs:subClassOf ?super .
  OPTIONAL {
    ?concept rdfs:subClassOf ?inbetweener .
    ?inbetweener rdfs:subClassOf ?super .
    FILTER( ?inbetweener != ?concept && ?inbetweener != ?super )
  }
  FILTER( ! BOUND(?inbetweener) && ?super != ?concept)
  BIND( my:Concept as ?concept )
}

Unexpected result: super=<http://ex#Super1>, super=<http://ex#Super2>

[SES-1095] TimeLimitIteration can cause excessive memory use

Wed, 31 Oct 2012 22:33:46 +0100 (CET)

TimeLimitIteration create instances of an anonymous class derived from TimerTask and adds them to a singleton instance of Timer.

When an iteration is complete, it cancels each TimerTask object, however Timer does not purge cancelled TimerTasks from its queue - it waits until the appropriate time to execute them and then if they are cancelled just discards them. Flushing the Timer's queue at the moment when a TimerTask is cancelled is not very efficient.

So a problem can arise when a large number of queries are executed and the Timer's queue grows, because each of the iteration objects can not be garbage collected while they are waiting to time-out (even they they are closed/finished).

I suggest a solution where The TimerTask associated with each TimeLimitIteration sets its reference to the iteration object to null at the point where it is cancelled (closed).

I will attach a suggested implementation of TimeLimitIteration to this issue.

[SES-1091] Using setBinding on query variable in property path loses that binding from result set

Thu, 6 Sep 2012 06:01:35 +0200 (CEST)

Problem reported on mailinglist by Peter Ansell (see http://sourceforge.net/mailarchive/message.php?msg_id=29766713 )

Data set:

@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix owl: <http://www.w3.org/2002/07/owl#> .
@prefix obo: <http://purl.obolibrary.org/obo/> .

obo:PO_0025131 rdfs:subClassOf owl:Thing .
obo:PO_0025131 rdf:type owl:Class .
obo:PO_0009011 rdf:type owl:Class .
obo:PO_0009011 rdfs:subClassOf obo:PO_0025131 .
obo:PO_0025117 rdf:type owl:Class .
obo:PO_0025117 rdfs:subClassOf obo:PO_0025131 .
obo:PO_0025161 rdf:type owl:Class .
obo:PO_0025161 rdfs:subClassOf obo:PO_0025131 .

Query:

SELECT ?parent ?child
WHERE {
?child a owl:Class .
?child rdfs:subClassOf+ ?parent .
}

Executed without setting any bindings, result:

[child=http://purl.obolibrary.org/obo/PO_0025131;parent=http://www.w3.org/2002/07/owl#Thing]
[child=http://purl.obolibrary.org/obo/PO_0009011;parent=http://purl.obolibrary.org/obo/PO_0025131]
[child=http://purl.obolibrary.org/obo/PO_0009011;parent=http://www.w3.org/2002/07/owl#Thing]
[child=http://purl.obolibrary.org/obo/PO_0025117;parent=http://purl.obolibrary.org/obo/PO_0025131]
[child=http://purl.obolibrary.org/obo/PO_0025117;parent=http://www.w3.org/2002/07/owl#Thing]
[child=http://purl.obolibrary.org/obo/PO_0025161;parent=http://purl.obolibrary.org/obo/PO_0025131]
[child=http://purl.obolibrary.org/obo/PO_0025161;parent=http://www.w3.org/2002/07/owl#Thing]

Executed with using setBinding on parent, to value owl:Thing, result:

[child=http://purl.obolibrary.org/obo/PO_0025131]
[child=http://purl.obolibrary.org/obo/PO_0009011]
[child=http://purl.obolibrary.org/obo/PO_0025117]
[child=http://purl.obolibrary.org/obo/PO_0025161]

While the result seems to be correct wrt to the 'child' variable (these four results are the correct set of transitive subclasses for owl:Thing), for some reason the 'parent' variable has been lost from the result.

[SES-1090] Add vocabulary constants for owl:Thing and owl:Nothing in OWL class

Wed, 5 Sep 2012 22:48:44 +0200 (CEST)

For some reason, org.openrdf.model.vocabulary.OWL does not define constants for owl:Thing and owl:Nothing. These should be added.

[SES-1086] Default logback loglevel should be INFO, is currently set to WARN

Tue, 30 Oct 2012 23:33:09 +0100 (CET)

Default logback loglevel should be INFO, is currently set to WARN

[SES-1080] Buffered RDF handler that groups incoming triples by context/subject/predicate

Wed, 15 Aug 2012 06:57:49 +0200 (CEST)

An RDFHandler implementation that can be used as a wrapper and which groups incoming statements together on context, subject, and predicate would be a useful new feature. This could be used in combination with any of the existing RDFWriters to produce slightly more ordered RDF documents (where all triples about the same subject are grouped together).

[SES-1079] RDF/XML parser should report warning on datatypes starting with 'xsd:'

Followup — Mon, 13 Aug 2012 22:44:39 +0200 (CEST)

A common mistake in (handwritten) RDF/XML is to specify the datatype of a literal using its prefixed name:

<foo:bar rdf:datatype="xsd:string">foobar</foo:bar>

However, while this is syntactically correct, it is intrepreted by the parser as meaning that the datatype URI is 'xsd:string', rather than 'http://www.w3.org/2001/XMLSchema#string'.

The correct way to specify the datatype in RDF/XML is:

<foo:bar rdf:datatype="http://www.w3.org/2001/XMLSchema#string">foobar</foo:bar>

It would be helpful if the parser, when running in VERIFY or NORMALIZE mode, would report a warning when it encounters datatype URIs that start with 'xsd'.

[SES-1078] SPARQL query doesn't sort results when needing to cast numeric xsd:string to xsd:integer

Followup — Sun, 12 Aug 2012 17:52:59 +0200 (CEST)

I may have found a bug when sorting SPARQL queries on xsd:string data items that need to be converted to xsd:integer for the sorting.

I've attached a small RDF triples file with items deliberately out of order. When loaded into Sesame, and queried with the following query, it demonstrates the issue.

PREFIX ida: <http://ida.org/example/Section#>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

SELECT DISTINCT ?num ?section WHERE {
<http://ida.org/example/Section#Document> ida:contains ?section .
?section ida:section_num ?sectionNumber .
?sectionNumber ida:content ?num .
} ORDER BY xsd:integer(?num)

The attached SPARQL/XML result file shows that the results are not being returned as expected. The appearance of the issue began when I switched from an RDF/XML data set from an N-Triples data set, but I think the input format is possibly a red herring here.

[SES-1077] GroupIterator should not evaluate entire subquery on initialize

Fri, 10 Aug 2012 06:37:47 +0200 (CEST)

Causes evaluate() calls to block for a long time, defeating lazy iteration and query abort attempts.

init should be moved to first call of hasNext() method.

[SES-1076] Support certain standard namespaces without having explicit declarations for them

Thu, 9 Aug 2012 22:44:45 +0200 (CEST)

Sesame's SPARQL engine currently needs explicitly defined namespace declarations for every qname in a query. It would be convenient if the engine would use default values for certain 'standard' namespace prefixes, such as rdf, rdfs, owl, xsd, fn, and the sesame namespace.

This can be easily dealt with in the PrefixDeclProcessor, by simply inserting these default namespaces (but only _if_ there is no explicitly defined namespace with the same value or prefix).

[SES-1074] Context ignored in subselect query

Thu, 9 Aug 2012 00:53:15 +0200 (CEST)

It seems as if subselect query with specified context ignore the context and use default graph instead.

Following queries provides different answers :

SELECT (count(?s) as ?count) FROM <http://www.ontotext.com/implicit> WHERE
{
     ?s ?p ?o .
}

SELECT * WHERE
{
{
   SELECT (count(?s) as ?count) FROM <http://www.ontotext.com/implicit> WHERE
{
    ?s ?p ?o .
}
}
}

[SES-1071] Deprecate RDBMS Sail code and disable RDBMS-related tests

Followup — Wed, 12 Sep 2012 01:38:10 +0200 (CEST)

The Sesame RDBMS Sail will be deprecated. Code needs to be marked as such, and testsuites on the Postgres/MySQL backends can be disabled.

[SES-1070] Improve streaming behavior of HTTPRepository

Duplicate — Thu, 16 Aug 2012 02:37:36 +0200 (CEST)

Currently, the HTTPRepository does a complete parse of a server response before passing the result on to the caller. This means that executing any operation with a large number of results (e.g. an export) over HTTP runs the risk of OutOfMemoryErrors at the client side, even when the user expects an iteration.

To improve, the HTTPRepository should implement a strategy similar to the SPARQLRepository, which processes server responses in the background and can return results to the user _while processing_ the server response.

[SES-1069] CLONE -Export Repository Data To Binary Format - Unsupported yet

Duplicate — Sun, 5 Aug 2012 03:33:52 +0200 (CEST)

Tomcat 6.0 container

Export Repository --> Download format -> Binary RDF

Click on Download produces the following error:

-----------------------------------------------------------------------------

HTTP Status 500 -

type Exception report

message

description The server encountered an internal error () that prevented it from fulfilling this request.

exception

java.lang.UnsupportedOperationException
org.openrdf.rio.binary.BinaryRDFWriterFactory.getWriter(BinaryRDFWriterFactory.java:40)
org.openrdf.workbench.commands.ExportServlet.service(ExportServlet.java:50)
org.openrdf.workbench.base.TransformationServlet.service(TransformationServlet.java:73)
org.openrdf.workbench.base.BaseServlet.service(BaseServlet.java:40)
org.openrdf.workbench.proxy.ProxyRepositoryServlet.service(ProxyRepositoryServlet.java:93)
org.openrdf.workbench.proxy.WorkbenchServlet.service(WorkbenchServlet.java:131)
org.openrdf.workbench.proxy.WorkbenchServlet.service(WorkbenchServlet.java:90)
org.openrdf.workbench.proxy.WorkbenchGateway.service(WorkbenchGateway.java:109)
org.openrdf.workbench.base.BaseServlet.service(BaseServlet.java:40)
org.openrdf.workbench.proxy.CookieCacheControlFilter.doFilter(CookieCacheControlFilter.java:52)

[SES-1067] Trying to HTTP DELETE a repository gives a 500 Method Not Allowed Error

Mon, 30 Jul 2012 02:53:21 +0200 (CEST)

When trying to issue a HTTP DELETE (with Accept: */*) to a repository URL e.g. http://localhost:8080/openrdf-sesame/myrepo which according to the documentation at http://www.openrdf.org/doc/sesame2/system/ch08.html#d0e320 is the way to delete a repository I receive a HTTP 500 Method Not Allowed error

Besides the fact that this is the wrong HTTP error code for method not allowed is there any obvious reason why this wouldn't work

I am using Sesame 2.6.4 and will try upgrading to see if that resolves the issue

[SES-1063] integer division should set higher scale

Mon, 23 Jul 2012 22:35:55 +0200 (CEST)

Currently, in sparql, doing

select ?v where { bind( 1/3 as ?v) }

results in:

"0"^^xsd:decimal

This is caused by the fact that MathUtil uses the input scales as defining the scale of the output decimal. Since we are dividing integers, the scale is zero, by definition.
An improvement would be to explicitly set the scale on divisions that result in non-terminating decimal sequences. An arbitrary number that seems suitable is 24 digits.

[SES-1062] IllegalArgumentException: timeout value is negative

Wed, 18 Jul 2012 19:36:14 +0200 (CEST)

We also see some cases of uncaught IllegalArgumentExceptions in the MemoryStore sync thread:

As per http://sourceforge.net/mailarchive/message.php?msg_id=29558521

ERROR 11:13:39,490 [Thread-5] STDERR Exception in thread "MemoryStore snapshot cleanup"
ERROR 11:13:39,490 [Thread-5] STDERR java.lang.IllegalArgumentException: timeout value is negative
ERROR 11:13:39,490 [Thread-5] STDERR at java.lang.Object.wait(Native Method)
ERROR 11:13:39,491 [Thread-5] STDERR at info.aduna.concurrent.locks.LockManager.waitForActiveLocks(LockManager.java:126)
ERROR 11:13:39,491 [Thread-5] STDERR at info.aduna.concurrent.locks.AbstractReadWriteLockManager.waitForActiveReaders(AbstractReadWriteLockManager.java:95)
ERROR 11:13:39,491 [Thread-5] STDERR at info.aduna.concurrent.locks.ReadPrefReadWriteLockManager.getWriteLock(ReadPrefReadWriteLockManager.java:113)
ERROR 11:13:39,491 [Thread-5] STDERR at org.openrdf.sail.memory.MemoryStore.cleanSnapshots(MemoryStore.java:916)
ERROR 11:13:39,491 [Thread-5] STDERR at org.openrdf.sail.memory.MemoryStore$2.run(MemoryStore.java:969)
ERROR 11:13:39,491 [Thread-5] STDERR at java.lang.Thread.run(Thread.java:662)

[SES-1060] REGRESSION: SameTermFilterOptimizer throws AssertionError if Term Used in UNION

Cause — Tue, 7 Aug 2012 03:18:16 +0200 (CEST)

The following query fails to evaluate when assertions are enabled (-ea).

StringBuilder query = new StringBuilder();
query.append("PREFIX foaf:<http://xmlns.com/foaf/0.1/>\n");
query.append("SELECT * {\n");
query.append(" {\n");
query.append(" ?sameTerm foaf:mbox ?mbox\n");
query.append(" FILTER sameTerm(?sameTerm,$william)\n");
query.append(" } UNION {\n");
query.append(" ?x foaf:knows ?sameTerm\n");
query.append(" FILTER sameTerm(?sameTerm,$william)\n");
query.append(" }\n");
query.append("}");

For what it's worth: this issue _only_ affects running with assertions enabled. When assertions are not enabled, the query is correctly optimized and evaluated.

The reason is that the offending assertion simply checks that a new binding does not overwrite an existing binding in the bindingset, but in this case, the new binding has the exact same value as the existing binding, so an overwrite does not cause an actual problem. Still, the assertion is correct, and we should check that we don't overwrite before adding the binding.

[SES-1059] Shutting down and re-initializing a remote repository causes an IllegalStateException

Wed, 11 Jul 2012 02:53:25 +0200 (CEST)

The following code:

Repository rep = ...
RepositoryConnection conn = rep.getConnection();
conn.clearNamespaces();
conn.commit();
rep.shutDown();
rep.initialize();
conn = rep.getConnection();
conn.getNamespaces().hasNext();

causes this exception when used with a remote repository:

java.lang.IllegalStateException: Connection factory has been shutdown.
at org.apache.commons.httpclient.MultiThreadedHttpConnectionManager.doGetConnection(MultiThreadedHttpConnectionManager.java:463)
at org.apache.commons.httpclient.MultiThreadedHttpConnectionManager.getConnectionWithTimeout(MultiThreadedHttpConnectionManager.java:416)
at org.apache.commons.httpclient.HttpMethodDirector.executeMethod(HttpMethodDirector.java:153)
at org.apache.commons.httpclient.HttpClient.executeMethod(HttpClient.java:397)
at org.apache.commons.httpclient.HttpClient.executeMethod(HttpClient.java:323)
at org.openrdf.http.client.HTTPClient.getTupleQueryResult(HTTPClient.java:1136)
at org.openrdf.http.client.HTTPClient.getNamespaces(HTTPClient.java:918)
at org.openrdf.http.client.HTTPClient.getNamespaces(HTTPClient.java:899)
at org.openrdf.repository.http.HTTPRepositoryConnection.getNamespaces(HTTPRepositoryConnection.java:387)

Which sounds like it is connected to ~~SES-999~~