|author||Reto Gmuer <email@example.com>||Mon Aug 15 04:44:08 2016 +0000|
|committer||Reto Gmuer <firstname.lastname@example.org>||Mon Aug 15 04:44:08 2016 +0000|
Added support for queries returning graphs
Build using Apache Maven with
You can build the documentation with:
Some more advanced questions focused on showing the differences to other proposed APIs.
Yes, any compliant implementation of BlankNode, Iri or Literal as well as Triple can be added to any implemenation of Graph, as long the Graph supports adding triples. Implementation may not require the nodes to be of a particular implementation of having been created with a specific factory.
Implementation might need to map instances of BlankNode to their internal implementation. This should be done in a way that when there is no more reference to the BlankNode object (i.e. when the object can be garbage collected) the mapping to the internal implementation is removed from memory to. This can be achieved by using a java.util.WeakHashMap
For instances of Iri or Literals you get back an object that result equal to the originally added object, i.e. an object with the same HashCode and of which the equals method return true when compared with the originally added object. Ther is no guarantee that the same instance will be returned. For instances of BlankNode the above in only guaranteed as long as the original object is referenced. When the original object becomes eligible for garbage collection the implementation may start returning a different (an not equal) object. In practice this means BlankNode objects cannot safely be serialized (using Java serialization) or passed around via RMI.
Yes, as long as this doesn't affect any BlankNode instance that is currently reachable (i.e. the Java object is in memory and is not eligible for garbage collection).
For example given the non-lean graph:
ex:a ex:p _:x . _:y ex:p _:x .
As long as there is no BlankNode instance referencing _:y the implementation can reduce the graph to:
ex:a ex:p _:x .
removing the redundancy. If however there is a reachable BlankNode instance for _:y the implementation must not remove the redundancy as the code which has access to the object can go on adding a triple:
_:y ex:p2 ex:b .
Thus creating a graph that doesn't contain any internal redundancy, namely:
ex:a ex:p _:x . _:y ex:p _:x . _:y ex:p2 ex:b .