On Sat, 21 Jul 2007, Mikko Tiihonen wrote:
> The wasNullFlag must be evaluated for each getXXX method invocation.
> Currently it is done all over the ResultSet classes. The patch moves
> the evaluation inside checkResultSet method that is already invoked at
> the beginning of each getXXX method. Also updates the javadoc to be
> explicit about the functionality.
>
One of the reasons for doing the ad-hoc null flag setting was to avoid
multiple checkResultSet calls like you've now introduced for things like
getByte. Rethinking that decision now, the performance impact of
checkResultSet is virtually non-existent, so this cleanup makes sense.
Applied with some additional minor modifications (missed
Jdbc[24]ResultSet, mention the fact that getFastXXX can no longer handle
null values).
I don't like slipping in unrelated functionality changes like this one,
which I took out:
+ if (fields[columnIndex - 1].getOID() == Oid.BOOL) {
+ return toBoolean(getFixedString(columnIndex)) ? 1 : 0;
+ }
I'm not opposed to the idea, but please raise things like this as separate
issues and not silently lump it in with a change that supposedly didn't
affect functionality. Please resubmit.
Kris Jurka