Hi,
PL/Java implements JDBC Savepoints using BeginInternalSubTransaction/
ReleaseCurrentSubTransaction/RollbackAndReleaseCurrentSubTransaction.
That seems to be the Accepted Way of Doing Things within backend PLs
that want control over error recovery, am I right?
PL/Java also strictly enforces that such a subxact set within a Java
function must be released or rolled back by the time that function
returns.
The reasoning there is less obvious to me; my intuition would have been
that a subtransaction could remain in play for the life of its containing
transaction, which could have been started outside of this Java function;
by holding a reference to the JDBC Savepoint object, a later Java function
called in the same transaction could release it or roll it back.
But I am beginning to suspect that the limitation may be essential, given
the comments in xact.c around StartSubTransaction and how its effects would
get clobbered on exit from a Portal, so a subxact started by an actual
SAVEPOINT is started in two steps, the later one after the Portal has
exited. By contrast, within a function (being executed inside a Portal?),
I have to use BeginInternalSubTransaction, which combines the multiple steps
into one, but whose effects wouldn't survive the exit of the Portal.
Have I reverse-engineered this reasoning correctly? If so, I'll add some
comments about it in the PL/Java code where somebody may be thankful for
them later.
Or, if it turns out the limitation isn't so inescapable, and could be
relaxed to allow a subxact lifetime longer than the single function that
starts it, I could look into doing that.
Thanks!
-Chap