Yep false is a winner here in this case.
David W. Van Couvering wrote:
Thanks, Lance. I understand the motivation for the method. My
question, is how do we respond to it. It appears like result sets are
closed if the severity of the exception is high enough, but that's not
what this API seems to be asking about it. It seems to ask if *any*
SQLException is a guarantee that result sets are closed.
Since we don't *always* close result sets, even though we do it in
*some* cases, baesd on my interpretation of the semantics of this
method, I will need to answer false. I just want to make sure there
are no objections.
David
Lance J. Andersen wrote:
This method was added as various drivers and database vendors all do
this differently. In some cases autocommit is handled explictly by
the drivers themselves, not the backend.
autocommit has been very problematic over the years and somedays it
would be great it it never existed....
David W. Van Couvering wrote:
Hi, all (and especially Lance). I just ran this test:
- Create a statement that is scrollabe and read-only
- Execute a select on the statement
- Attempt to update the row, catch the exception
- Call rs.beforeFirst(). Looking at the embedded code, if the
result set is closed, this should cause an exception saying the
result set is closed.
My test indicates that the result set is not closed, independent of
whether it is holdable or not. But this makes sense. What database
would want to close a result set just because someone tries to
update it?
That said, I am trying to answer the question to the new metadata call
autoCommitFailureClosesAllResultSets()
where the javadoc states
"Retrieves whether a SQLException while autoCommit is true inidcates
that all open ResultSets are closed, even ones that are holdable.
When a SQLException occurs while autocommit is true, it is vendor
specific whether the JDBC driver responds with a commit operation, a
rollback operation, or by doing neither a commit nor a rollback. A
potential result of this difference is in whether or not holdable
ResultSets are closed."
Based on this javadoc, it looks like our answer should be "false."
But is this the correct interpretation? Is there a better way to
test this? I feel like I'm not quite understanding the intention of
this metadata call.
Thanks,
David
