Database engine independent INSERT of multiple rows

Question

In our product we try to support multiple database engines while using as much standard SQL statements as possible. Our database has around 50 tables and our queries are generally pretty straightforward.
Our initial database was MySQL, but now we successfully run the product on Apache Derby and we are also trying to support Oracle 10g Release 1.
MySQL has a feature of multiple row inserts inside a single INSERT statement.
INSERT INTO mytable(col1, col2, col3) VALUES
(val11, val12, val13),
(val21, val22, val23);
On MySQL the multiple row insert is much more efficient than splitting it into separate inserts. I found something similar available in Oracle since release 9i:
INSERT ALL
INTO mytable(col1, col2, col3) VALUES
(val11, val12, val13)
INTO mytable(col1, col2, col3) VALUES
(val21, val22, val23)
SELECT 1 FROM DUAL;
Of course this is far from portable. Does anyone know a portable solution to multiple row inserts?
Fixed invalid Oracle multiple row insert.
Message was edited by:
user536129

Answer 1

> That syntax isn't supported only by MySQL, but also by Derby and
most probably DB2 (but I didn't check that one yet).
That syntax contravenes a fundamental rule of sharable SQL - it has a variable/repeating set of commands for multiple values.
It honestly does not make any sense for a RDBMS to support this syntax. Unless said RDBMS has head-up-backside when it comes having to hard parse every single SQL statement and not bothering with a shared SQL pool.
In Oracle, the feature that achieves this abortion-syntax-of-mySQL is called BULK PROCESSING.
You pass an INSERT statement to Oracle. It is parsed (soft when it already resides in the shared pool). You then execute it by providing Oracle with an array of values to insert.
In PL/SQL this is done using the FORALL statement.
Implying that Oracle should support the mySQL/Derby/DB2 syntax is just plain silly. Oracle scales up. Not down.