As shalin says threadsafety is not given importance because DIH is
singlethreaded. But eventually we may move there .
But we indeed tried once w/ multithreaded indexing for around 3million
records and the performance gain we observed was <10% . We realized
that it was bound by Lucene's ability to index documents rather than
DIH's ability to produce it. So we dropped it in favor of simplicity
--Noble
On Tue, Nov 18, 2008 at 10:28 PM, Shalin Shekhar Mangar
<[EMAIL PROTECTED]> wrote:
> Hmm, the import operation in DIH is single threaded. Each entity has a
> separate instance of DataSource for itself. I think that we may not need the
> synchronization at all.
>
> On Tue, Nov 18, 2008 at 10:22 PM, Yonik Seeley <[EMAIL PROTECTED]> wrote:
>
>> I took a very quick peek... looks like Mark is right, there are some
>> non-thread-safe practices going on there with both "conn" and
>> "connLastUsed". Prob also want another check inside the sync block so
>> multiple threads don't pile up at that sync block and all open a new
>> connection in turn (or just put the check inside the sync block).
>>
>> Is a finalizer really needed? If so, hopefully the connection is
>> resistant to close() being called on it more than once?
>>
>> -Yonik
>>
>>
>> On Tue, Nov 18, 2008 at 11:27 AM, Noble Paul നോബിള് नोब्ळ्
>> <[EMAIL PROTECTED]> wrote:
>> > it is not . conn is protected throughout
>> >
>> > On Tue, Nov 18, 2008 at 9:48 PM, Mark Miller <[EMAIL PROTECTED]>
>> wrote:
>> >> But if your going to sync a variable, you can't read/write it outside of
>> >> proper synchronization. Maybe I am just not following the code...
>> >>
>> >> Noble Paul ??????? ?????? wrote:
>> >>>
>> >>> one JdbcDataSource has only one connection the connection object/the
>> >>> connLastUsed etc needs to be protected.
>> >>>
>> >>>
>> >>>
>> >>> On Tue, Nov 18, 2008 at 9:24 PM, Mark Miller <[EMAIL PROTECTED]>
>> >>> wrote:
>> >>>
>> >>>>
>> >>>> JdbcDataSource looks like it has a little funkiness going on. Why is
>> >>>> there a
>> >>>> synchronize block there? Can multiple threads call getConnection
>> >>>> concurrently? If they can, this is not thread safe anyway. If they
>> can't,
>> >>>> why is factory.call (or is it the close?) being protected with a sync?
>> >>>>
>> >>>> private Connection getConnection() throws Exception {
>> >>>> long currTime = System.currentTimeMillis();
>> >>>> if (currTime - connLastUsed > CONN_TIME_OUT) {
>> >>>> synchronized (this) {
>> >>>> Connection tmpConn = factory.call();
>> >>>> close();
>> >>>> connLastUsed = System.currentTimeMillis();
>> >>>> return conn = tmpConn;
>> >>>> }
>> >>>>
>> >>>> } else {
>> >>>> connLastUsed = currTime;
>> >>>> return conn;
>> >>>> }
>> >>>> }
>> >>>>
>> >>>>
>> >>>
>> >>>
>> >>>
>> >>>
>> >>
>> >>
>> >
>> >
>> >
>> > --
>> > --Noble Paul
>> >
>>
>
>
>
> --
> Regards,
> Shalin Shekhar Mangar.
>
--
--Noble Paul
