PL\SQL optimization for update or insert behavior

I am trying to create a PL statement that will either update an existing record or insert a new record if no existing records are found. This will eventually be used in a JDBC PreparedStatement for batch processing. Assume that there is a pretty random distribution of inserts and updates. I've come up with a few possible ways of doing this but I'm not sure which will run faster. The smallest of gains are important since it will be run possibly thousands of times (or more). The requirements for the project forbid me to use stored procedures so that is not an option.

Which of the follow statements would be the most efficient (or is there a better way)? Any help would be greatly appreciated.

Jon

• Statement 1: --------

DECLARE

	recordFound EXCEPTION;
	spoo ROWID;
BEGIN
	SELECT ROWID into spoo from JCTEST where account_number = '103';
	raise recordFound;
EXCEPTION
	when recordFound then
		UPDATE JCTEST SET amount = 23.23 where account_number = '103';		
	when no_data_found then
		INSERT into JCTEST VALUES ('103', 101.70);

• Statement 2 --------- DECLARE recordFound EXCEPTION; spoo ROWID; acc varchar2(5):='103'; amt NUMBER :=101.70; BEGIN SELECT ROWID into spoo from JCTEST where account_number = acc; raise recordFound; EXCEPTION when recordFound then UPDATE JCTEST SET amount = amt where account_number = acc; when no_data_found then INSERT into JCTEST VALUES (acc, amt); END;
  • Statement 3 ---------- DECLARE recordFound EXCEPTION; spoo NUMBER; BEGIN SELECT COUNT(*) into spoo from JCTEST where account_number = '103'; if spoo > 0 then UPDATE JCTEST SET amount = 23.23 where account_number = '103'; else INSERT into JCTEST VALUES ('103', 101.70); END IF; END;