Sing Li wrote:
I have the same problem with a CMP example - how to
create Derby SQL tables "once only" during build
and/or deployment??
The ActiveMQ code does it programmatically, but it
seems to try it again each time during server boot.
This is kind of illustrative of the problem. ActiveMQ needs a couple of
tables as backing store but finding out whether they are there already
is harder than it appears.
For example, you can query database metadata but for that to work you
need to know how the catalog and schema are defined by your databases
(setting aside minor annoyances like which case the database uses to
store the names).
The simple but pragmatic approach ActiveMQ takes is just to create the
tables and ignore errors that result, deferring the real failure until
it needs to insert a message. This works for a simple schema like they
user (I think it's just one or two tables) but is less useful for a
complex schema like that being used by an entire application.
--
Jeremy
