Attached is my version of Oliver's server prepared statement threshold
patch. I've submitted a patch to have server prepared statements return
the correct affected row count which eliminates a large number of
regression test failures.
http://archives.postgresql.org/pgsql-patches/2004-04/msg00195.php
I've fixed a problem with batch statements not correctly deallocating
queries in the Statement.addBatch(String) case where there is no relation
from one statement to the next. I've also explicitly forbid the mixing of
the two batch types, PreparedStatement.addBatch() and
Statement.addBatch(String), in the same batch. This would be a rather
bizarre situation to have and makes tracking statements and variables more
difficult.
The significant regression test obstacle left is that server prepared
SELECT statements don't return table and column information used in
ResultSetMetaData. I've brought this up on hackers here:
http://archives.postgresql.org/pgsql-hackers/2004-04/msg00510.php
The one remaining issue is a boolean vs. bit conversion that doesn't look
terribly important.
Kris Jurka