Hi Mike:
Thanks for the reply!
The reason my pattern wouldn't work is because my doClose impl does
nothing to avoid the actual closing and then later my "index updater" will
do the actual closing. The problem is although my doClose is called, the
reference is decremented to 0, and ensureOpen fails upon re-use of the
indexReader. Prior to 2.3.1, my pattern worked because of lack the reference
counting logic.
I will check on how decRef/incRef would work in my case. Seems a bit
"back-door"-ish. :)
Do you think making IndexReader.close non-final is a possibility, cuz
that would solve my problem.
Thanks
-John
On Mon, Jun 30, 2008 at 2:07 AM, Michael McCandless <
[EMAIL PROTECTED]> wrote:
>
> It's true there is now reference counting logic to determine when it's
> really time to close an IndexReader, but, at that point doClose() is still
> called so your custom logic should still be running, right? It seems like
> your pattern ought to still be working?
>
> If not, another alternative might be to override decRef() with your own
> logic, or, you could externally hold your own reference to the reader (call
> its incRef()) and then decRef() it when you're done. That would probably be
> cleanest...
>
> Mike
>
>
> On Jun 29, 2008, at 11:51 AM, John Wang wrote:
>
> Hi:
>> I had some code to do indexReader pooling to avoid open and close on a
>> large index when doing lotsa searches. So I had a FilteredIndexReader
>> proxy
>> that overrides the doClose method to do nothing, and when I really want to
>> close it, I call super.doClose(). This patter worked well for me prior to
>> 2.3. With 2.3, the close is doing the reference counting out side of
>> doClose, and since close is final, my pattern is broken.
>>
>> Am I using the api incorrectly? how would I accomplish this task?
>>
>> Thanks
>>
>> -John
>>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
>
