sebb wrote:
OK, I'd not noticed that the class was usable without the DataSource.
Of course the alternative is to document the class as thread-unsafe...
I would guess that the reason we've never seen a bug filed on this issue
is that nobody uses setDataSource after the class is created. For these
users, QueryRunner is thread-safe. I think just formalizing that state is
best.
I would not attempt to synchronize this class, just leave it unsafe and let
users synchronize. We should document more explicitly that (unlike some
other classes in DbUtils) it's unsafe.
I'm not sure that the class can be made thread-safe externally.
It's easy enough to override the setters with synchronized versions,
but the getters need to be synchronized as well to ensure that the
data is published correctly. However the class stores the
unsynchronized getters in the Map. So it would be necessary to
override invoke() as well. If this is done, then the whole class has
been overridden - one might as well say it has been rewritten.
I didn't mean that users would synchronize externally by
extending/overriding, but just by synchronizing access to an instance
member, or just not sharing them across threads. *shrug*
-Dan
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
For additional commands, e-mail: dev-h...@commons.apache.org
