joshsh / stream42 Goto Github PK

Integrate LinkedDataSail as described in the PhD thesis

There's an issue with joining partial solutions from patterns that do not share any variables.
While partial solutions for the individual patterns are found, SolutionIndex.joinSolutions() does not join them, because, looking only for other solutions that have the same variables bound, they are not found.

As an example, the query

SELECT * WHERE {
  ?item ?pred true .
  ?item2 ?pred2 false .
}

with two statements

queryEngine.addStatements(DEFAULT_EVENT_TTL, new StatementImpl(
  new URIImpl("urn:test:created:uri:1"), new URIImpl("urn:test:created:pred:1"),
  new LiteralImpl("true", XMLSchema.BOOLEAN)));
queryEngine.addStatements(DEFAULT_EVENT_TTL, new StatementImpl(
  new URIImpl("urn:test:created:uri:2"), new URIImpl("urn:test:created:pred:2"),
  new LiteralImpl("false", XMLSchema.BOOLEAN)));

should create one result
[item=urn:test:created:uri:1;pred=urn:test:created:pred:1;item2=urn:test:created:uri:2;pred2=urn:test:created:pred:2]
but doesn't.

As a fix, in Query.bind() I add the unbound query variables with a value of null (along with some minor changes to cope with them), which works and still runs your tests (albeit it creates more partial solutions, obviously).

Proposed fix: 3aabf1b

Use the formal definitions of a continuous query database from the thesis. Try to find any ways in which SesameStream deviates from those definitions.

Depends on #2

Evaluate SS+LDS analogously to [Har11a] for the continuous query scenario from the motivating example in the thesis. Precisely state and back up the claim that in a facilitated environment where a sufficient number of individuals are interacting with the same objects of interest, the latency for retrieving metadata about objects of interest drops below the necessary thresholds with a certain probability over time.
Criteria for completion:
• formal description my solution, analogous to link traversal based query execution, as documented in [Har11b]
• demonstration of the successful execution of continuous SPARQL queries whose answers require both streaming data and dynamically-fetched Linked Data

Like the title. Failing query scenarios have been discovered.

As described in the thesis