Tom Lane wrote:
> Peter Eisentraut <peter_e@gmx.net> writes:
>> Am Dienstag, 5. September 2006 05:58 schrieb Tom Lane:
>>> A couple of recently discussed FE/BE protocol issues are: not storing a
>>> plan at all for unnamed-statement cases, and thus allowing bind
>>> parameters to be treated as constants; allowing parameter types to go
>>> unresolved rather than throwing an error. Perhaps it's too late to
>>> consider these for 8.2, but they seem no more invasive than some other
>>> items on the open-issues list.
>
>> Do we have a patch for that today?
>
> We could have a patch for the first one today --- I was thinking about
> it last night and intending to code it today. The second one is merely
> a matter of removing an error check that exists now; the question really
> is do people want that behavior. (I asked that on the jdbc list and got
> zero response, so actually I was thinking that it was a dead issue; but
> as long as it's on the open-items list we ought to discuss it.)
I personally think it's a good idea to do it, as it should improve the
plans for one-shot queries. Unfortunately I don't certainly know how the
JDBC driver issues queries when called through a PreparedStatement but
without a "prepare-threshold"[*] set. If it uses the unnamed-statement,
then I guess the proposed change would be a win.
Best Regards
Michael Paesold
[*] This option determines, after how many executes of a prepared
statement, the driver will switch to server-side prepares.