On Fri, Jan 4, 2013 at 10:22 AM, Giuseppe Miscione < g.misci...@innovationengineering.eu> wrote:
> Hi Reto, > why this will affect also the interface? > Because it's defined wrongly in the the javadoc comment for hashCode in the triple interface. The comments needs to be fixed. Cheers, Reto > > Il 04/01/2013 10:17, Reto Bachmann-Gmür ha scritto: > > Yes, I confirm. Thanks Giuseppe. >> The bug affects the interface definition of Triple, not just the >> implementation. >> >> Cheers, >> Reto >> >> On Fri, Jan 4, 2013 at 10:06 AM, Tommaso Teofili >> <tommaso.teof...@gmail.com>**wrote: >> >> Hi Giuseppe, >>> >>> thanks a lot! This seems to me to be a bug (if anyone else feels I'm >>> wrong >>> please chime in), feel free to open a Jira issue on >>> https://issues.apache.org/**jira/browse/CLEREZZA<https://issues.apache.org/jira/browse/CLEREZZA>and >>> eventually attach a >>> patch. >>> >>> Thanks again and have a nice day! >>> Tommaso >>> >>> >>> 2013/1/4 Giuseppe Miscione >>> <g.miscione@**innovationengineering.eu<g.misci...@innovationengineering.eu> >>> > >>> >>> Hi all, >>>> I was looking at the hashCode implementation in the >>>> org.apache.clerezza.rdf.core.****impl.TripleImpl and I found that this >>>> method has this code: >>>> >>>> public int hashCode() { >>>> return (subject.hashCode() >> 1) ^ subject.hashCode() ^ >>>> (subject.hashCode() << 1); >>>> } >>>> >>>> This function uses only the subject of the triple. Wondering why only >>>> one >>>> part of the triple is used, I looked at the implementation of the >>>> corresponding class in the Jena library, and I found that there the >>>> >>> method >>> >>>> is implemented in this way: >>>> >>>> public int hashCode() { >>>> return (subject.hashCode() >> 1) ^ predicate.hashCode() ^ >>>> (object.hashCode() << 1); >>>> } >>>> >>>> And the reason is that in this way the triple (subject, predicate, >>>> >>> object) >>> >>>> has a differente hash code than (object, predicate, subject). >>>> >>>> Now, it seems to be a small bug in the TripleImpl class. This error >>>> doesn't prevent to use TripleImpl as keys in HashMaps because the >>>> corresponding equals method is correct and coherent ( a.equals(b) => >>>> a.hashCode() == b.hashCode() ), but it can yield to performance >>>> >>> degradation >>> >>>> when storing a lot of triples with the same subject in an HashMap. >>>> >>>> Regards, >>>> Giuseppe >>>> >>>> >
