On 05/29/2018 08:26 AM, Brian Dunavant wrote:
On Sun, May 27, 2018 at 6:04 PM, Stuart McGraw <smcg4191@xxxxxxxxxx <mailto:smcg4191@xxxxxxxxxx>> wrote:
Is there some way to to test if a savepoint of a given name
exists? Or better yet, the number of stacked savepoints of
that name?
This is kind of backwards, but you can determine if a savepoint by a certain name exists by trying to release it. You can do so without damaging the transaction by creating a savepoint immediately beforehand, but this will cause you to lose the state of the named savepoint. Example below.
[...snipped for brevity...]
Thanks. That's interesting, protecting a questionable
release with another savepoint. I wouldn't have thought
of that. But I'm not sure it helps my ultimate goal of
simplifying my application code or that it doesn't shift
the problem of not leaking savepoints to the test
savepoints.
