I am encountering a performance problem in ListOrderedSet.retainAll().
It appears in version 3.2.1 and also in revision 1365132. I attached
a test that exposes this problem and a patch that fixes it. On my
machine, for this test, the patch provides a 263X speedup.
To run the test, just do:
$ java Test
The output for the un-patched version is:
Time is 3682
The output for the patched version is:
Time is 14
The problem is that the code for method
"ListOrderedSet.retainAll(Collection<?> coll)" is:
because "collection.retainAll(coll)" has quadratic complexity if
parameter "coll" is a List. Conceptually, the solution is to call
"coll.retainAll(collection)" which has linear complexity (not
quadratic), because "collection" is always a HashSet (we know this,
because "collection" is an inherited field of ListOrderedSet, and thus
cannot have another type). See the
"java.util.AbstractCollection.retainAll()" code inlined below for why
one call has quadratic complexity, and the other has linear
Directly calling "coll.retainAll(collection)" would change the
behavior of ListOrderedSet.retainAll(), which is why the patch seems a
bit involved. In reality, the patch simulates the behavior of
"coll.retainAll(collection)" (which is just the intersection of two
collections). You can easily see this by looking at the patch and at
the "java.util.AbstractCollection.retainAll()" code inlined below.
"collection.retainAll(coll)" is "java.util.HashSet.retainAll()", which
is "java.util.AbstractCollection.retainAll()", which has the code:
Is this a bug, or am I misunderstanding the intended behavior? If so,
can you please confirm if the patch is correct?