Andrew Haley wrote:
Marco Trudel writes:
> Andrew Haley wrote:
> > Marco Trudel writes:
> > > Andrew Haley wrote:
> > > > Marco Trudel writes:
> > > > > Andrew Haley wrote:
> > > > > > Marco Trudel writes:
> > > > > > > Hey guys
> > > > > > >
> > > > > > > java.util.Arrays.binarySearch(Object[] a, Object key, Comparator c)
> > > > > > > exchanges a[i] and key, this can lead to ClassCastExceptions as shown in
> > > > > > > ComparatorTest.java. Arrays.patch fixes it.
> > > > > > >
> > > > > > >
> > > > > > > Changelog suggestion:
> > > > > > >
> > > > > > > 2006-12-13 Marco Trudel <[EMAIL PROTECTED]>
> > > > > > >
> > > > > > > * classpath/java/util/Arrays.java (binarySearch(Object[] a, Object key,
> > > > > > > Comparator c)):
> > > > > > > Fix swapped objects inCollections.compare(...) call.
> > > > > > >
> > > > > > >
> > > > > > > Comments?
> > > > > >
> > > > > > I guess I don't get it. Arrays.binarySearch() is defined to be
> > > > > > searching for a key in an array. The comparator has to be able
to
> > > > > > compare keys and values. How is this not a bug in the test case?
> > > > >
> > > > > The key might be a different object that the values. The comparator
> > > > > will then be responsible to see if they're equal. Thus the order is
> > > > > important...
> > > >
> > > > OK, but where does it say that in the spec?
> > >
> > > Sorry, I did not read the spec.
> >
> > What!!
>
> Shame on me! I was just debugging a ClassCastException on a library and
> found that one to be the cause...
>
> > > This somehow seems logical to me.
> >
> > But why? Are you assuming that the key _must_ occur on the LHS of the
> > Comparator?
>
> No. The point is that completely different objects might be passed which
> have no relation with each other. A stupid example:
>
> Object[] o1 = { "22", "23", "24" };
> Object o2 = new java.awt.Button("2");
>
> And then the comparator does:
>
> public int compare(Object o1, Object o2)
> {
> String s1 = Integer.parseInt((String)o1);
> String s2 = ((Button)o2).getName();
> return s1.compareTo(s2);
> }
Here is a quote from the definition of Comparable.compare:
"The implementor must ensure that sgn(compare(x, y)) == -sgn(compare(y, x))
for all x and y."
Ah yeah, right... Actually I read that (just forgot it).
Seems like the implementation of compare() is wrong. Funny thing is that
this is in a Sun AWT class. So Sun doesn't even follow their own
specifications ;-)
Marco
