Module org.apache.jena.core
Package org.apache.jena.mem2.store.fast

Class FastTripleStore

java.lang.Object
org.apache.jena.mem2.store.fast.FastTripleStore
All Implemented Interfaces:
TripleStore
public class FastTripleStore extends Object implements TripleStore
A triple store that uses hash tables to map from nodes to triple bunches.

Inner structure: - three FastHashMap, one for each node (subject, predicate, object) in the triple - each map maps from a node to a FastTripleBunch - for up to 16 triples with the same subject, the bunch is an FastArrayBunch, otherwise it is a FastHashedTripleBunch - for up to 32 triples with the same predicate and object, the bunch is an FastArrayBunch, otherwise it is a FastHashedTripleBunch

Other optimizations: - each triple is added to three FastTripleBunches. To avoid the overhead of calculating the hash code three times, the hash code is calculated once and passed to the FastTripleBunches. - the different sizes for the FastArrayBunches are chosen: - to avoid the memory-overhead of FastHashedTripleBunch for a small number of triples - the subject bunch is smaller than the predicate and object bunches, because the subject is typically used for #contains operations, which are faster for FastHashedTripleBunches. - the predicate and object bunches are the same size, because they are typically used for #find operations, which typically do not answer #contains operations. Making them much larger might slow down #remove operations on the graph. - "ANY_PRE_OBJ" matches primarily use the object bunch unless the size of the bunch is larger than 400 triples. In that case, there is a secondary lookup in the predicate bunch. Then the smaller bunch is used for the lookup. Especially for RDF graphs there are some very common object nodes like "true"/"false" or "0"/"1". In that case, a secondary lookup in the predicate bunch might be faster than a primary lookup in the object bunch. - FastTripleStore#contains uses FastTripleBunch.anyMatchRandomOrder(java.util.function.Predicate<org.apache.jena.graph.Triple>) for "ANY_PRE_OBJ" lookups. This is only faster than JenaMapSetCommon.anyMatch(java.util.function.Predicate<E>) if there are many matches and the set is ordered in an unfavorable way. "ANY_PRE_OBJ" matches usually fall into this category. This optimization was only needed because the FastTripleBunches does not use the random order of the triples in #anyMatch but the ordered dense array of triples, which is faster if there are only a few matches. - for the FastArrayBunches, the equals method of the triple is not called. Instead, only the two nodes that are not part of the key of the containing map are compared.

  • Constructor Details

    • FastTripleStore

      public FastTripleStore()

  • Method Details

    • add

      public void add(Triple triple)
      Description copied from interface: TripleStore
      Add a triple to the map.
      Specified by:
      add in interface TripleStore
      Parameters:
      triple - to add

    • remove

      public void remove(Triple triple)
      Description copied from interface: TripleStore
      Remove a triple from the map.
      Specified by:
      remove in interface TripleStore
      Parameters:
      triple - to remove

    • clear

      public void clear()
      Description copied from interface: TripleStore
      Remove all triples from the map.
      Specified by:
      clear in interface TripleStore

    • countTriples

      public int countTriples()
      Description copied from interface: TripleStore
      Return the number of triples in the map.
      Specified by:
      countTriples in interface TripleStore

    • isEmpty

      public boolean isEmpty()
      Description copied from interface: TripleStore
      Return true if the map is empty.
      Specified by:
      isEmpty in interface TripleStore

    • contains

      public boolean contains(Triple tripleMatch)
      Description copied from interface: TripleStore
      Answer true if the graph contains any triple matching t.
      Specified by:
      contains in interface TripleStore
      Parameters:
      tripleMatch - triple match pattern, which may be contained

    • stream

      public Stream<Triple> stream()
      Description copied from interface: TripleStore
      Returns a Stream of all triples in the graph. Note: BaseStream.parallel() is supported.
      Specified by:
      stream in interface TripleStore
      Returns:
      a stream of triples in this graph.

    • stream

      public Stream<Triple> stream(Triple tripleMatch)
      Description copied from interface: TripleStore
      Returns a Stream of Triples matching the given pattern. Note: BaseStream.parallel() is supported.
      Specified by:
      stream in interface TripleStore
      Parameters:
      tripleMatch - triple match pattern
      Returns:
      a stream of triples in this graph matching the pattern.

    • find

      public ExtendedIterator<Triple> find(Triple tripleMatch)
      Description copied from interface: TripleStore
      Returns an ExtendedIterator of all triples in the graph matching the given triple match.
      Specified by:
      find in interface TripleStore