Hi folks,

On 13.10.16 16:25, t...@apache.org wrote:
> Author: tv
> Date: Thu Oct 13 14:25:04 2016
> New Revision: 1764695
> URL: http://svn.apache.org/viewvc?rev=1764695&view=rev
> Log:
> Fix SortedPreferentialArray to actually do what was is supposed to do
> Modified:
> commons/proper/jcs/trunk/commons-jcs-core/src/main/java/org/apache/commons/jcs/auxiliary/disk/indexed/IndexedDiskElementDescriptor.java
> commons/proper/jcs/trunk/commons-jcs-core/src/main/java/org/apache/commons/jcs/utils/struct/SortedPreferentialArray.java
> commons/proper/jcs/trunk/commons-jcs-core/src/test/java/org/apache/commons/jcs/utils/struct/SortedPrefArrayUnitTest.java

When I looked into the tests of SortedPreferentialArray to see whether
this could be replaced with some JDK standard stuff, I found that it has
a number of severe problems. IMHO, it simply doesn't work. It doesn't
even sort correctly.

Now this class is used as a recycle bin in the IndexedDiskCache which,
to my knowledge, is a quite popular component within JCS.

Am I completely mistaken or is this functionality so unimportant that it
doesn't make any difference?

I replaced a lot of code with a reference to a ConcurrentSkipListSet and
managed to get all tests to pass again.

Could some kind soul please take a few cycles to review this commit and
the related code in IndexedDiskCache? I would like some help with
answering the following questions:

- The recycle bin contains elements of type
IndexedDiskElementDescriptor. Obviously, these elements were compared by
their size (len field) only, not their offset (pos field), so that
storage chunks of identical size are considered equal. Nevertheless,
this somehow seemed to work, even in the unit test where all elements
are of the same size. But why?

- The old implementation uses a fixed size array to store the elements.
If more elements were added, the smallest elements were discarded. Does
this actually make any sense?

- After my changes, the disk cache should recycle used storage chunks
more effectively. Can someone with a real-world application confirm this?

- Do I miss some other sophisticated feature in the code I removed?

Thank you in advance for any time you may be able to donate.

Bye, Thomas.

To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
For additional commands, e-mail: dev-h...@commons.apache.org

Reply via email to