Re: jdbc and postgresql function session question

On Tue, May 31, 2016 at 3:36 PM, David G. Johnston <
david(dot)g(dot)johnston(at)gmail(dot)com> wrote:

> On Tue, May 31, 2016 at 6:01 PM, Adrian Klaver <adrian(dot)klaver(at)aklaver(dot)com>
> wrote:
>
>> On 05/31/2016 02:20 PM, Michael Moore wrote:
>>
>>> When the postgres jdbc driver calls a postgres function, does it get a
>>> new session every time? If I create a PREPARED statement on the first
>>> call, will it still be available on the 2nd call? If not, is there any
>>> way I could get the prepared statement to be used on successive calls?
>>>
>>
>> I think this is going to need more explanation, so:
>>
>> 1) When you say session are you talking about connection or a
>> transaction? A function will run in its own transaction each time, but can
>> be run multiple times in a connection.
>>
>> 2) Where is the PREPARED statement being built, in the function or
>> outside it?
>>
>> 3) Can show an outline of what you are doing?
>>
>>
> ​Making some assumptions...
>
> con = DriverManager.getConnection(); -- opens a database session
>
> Any JDBC Statement objects you create using this con are able to access
> named PostgreSQL PREPAREd statement previously created.
> However, beware of
> DEALLOCATE ALL
> [1]
> ​
> While you are unlikely to issue such a command explicitly you need to be
> aware that connection poolers may do so in some configurations.
>
> ​CallableStatement stmt = con.prepareCall(sql); --sees any session-scoped
> data since con was created.​
>
> con.close();​
>
> ​ -- now the database session is done​
>
> https://www.postgresql.org/docs/current/static/sql-deallocate.html
> ​
> ​David J.​
>
> A little background. This is a PL/SQL to pgplsql conversion. A Java
process calls a pgsql function. The function does a few table lookups and
builds a moderately complex (3 way join) SQL SELECT statement and executes
it dynamically returning a set of about 5 rows. There are about 20 possible
structural variations of this same select statement. By that I mean
different columns being referenced in the WHERE clause and possibly a 4th
table being joined. There are no skewed data statistics, so it is unlikely
that the same select statement with different predicate values will choose
a different plan. This happens about about 10 times per second so
performance is critical. I am trying to determine if there could be
performance gains by having my pgsql function execute a PREPARE statement.
I am going to study the information at David's link and get information
about how java is handling the connection then I should be able to come
back here and ask more specific questions.
Thanks,Mike