Can we please not invent yet another pooling mechanism at Apache? ;-) OTOH
bringing in Apache Commons Pool might be too heavy handed.
For the two main server stacks I work on at my day job, it's not an issue
since I already use Commons Pool. Actually, now that I think of it: Commons
DBCP (which depend on Commons Pool) is already an optional dependency for
the JDBC Appender...
Discuss! :-)
Gary
On Wed, Jan 8, 2020 at 5:17 PM Volkan Yazıcı <volkan.yaz...@gmail.com>
wrote:
> On Mon, Jan 6, 2020 at 7:09 PM Matt Sicker <boa...@gmail.com> wrote:
> > Would it be useful to implement some sort of buffer pool for
> > StringBuilders and ByteBuffers? Could likely copy code from netty's
> > util library (ByteBuf et al.) or reuse stuff from commons-pool if
> > needed. This would work properly in applications, servlets, and even
> > reactive streams and lightweight threads later on.
>
> Would you mind elaborating a little bit more on the idea you have in
> mind, please? To get the discussion going, I have the following draft:
>
> public interface ObjectPool<V> {
> V acquire();
> void release(V instance);
> }
>
> public interface ObjectPoolFactory {
> // For char[], byte[], ByteBuffer, etc. (Types that cannot get
> extended implicitly.)
> ObjectPool<V> create(Class<V> clazz, Supplier<V> objectFactory);
> // For StringBuilder, etc. (Types that can get extended
> implicitly.)
> ObjectPool<V> share(Class<V> clazz);
> }
>
> One can pretty efficiently implement such an ObjectPool<V>, e.g., via
> a JCTools[1] MPMC queue. Though note that object pools necessitate
> manual object release, which is not the case for TLA-provided objects.
>
> Providing a shared object pool factory to the rest of the code base
> might need a little bit more thinking.
>
> Exposing configuration knobs to enforce certain limits on the pool and
> its factory is also a question that needs to be addressed.
>
> [1] https://github.com/JCTools/JCTools/
>
