base class for bulk DML upsert operations
This class assists with bulk upsert (SQL merge) operations into a target table.
- Submitting Data
- To use this class, queue data in the form of a hash (a single row or a set of rows) or a list of rows by calling the queueData() method.
The queueData() method queues data to be written to the database; the queue is flush()ed automatically when block_size
rows have been queued.
- Flushing and Discarding Data
- Each call to flush() (whether implicit or explicit) will cause a single call to be made to the dataserver; all queued rows are sent in a single bulk DML call, which allows for efficient processing of large amounts of data.
A call to flush() must be made before committing the transaction to ensure that any remaining rows in the internal queue have been written to the database. Because the destructor() will throw an exception if any data is left in the internal queue when the object is destroyed, a call to discard() must be made prior to the destruction of the object in case of errors.
on_success ds.commit();
on_error ds.rollback();
{
BulkUpsertOperation op1(table1);
BulkUpsertOperation op2(table2);
on_success {
op1.flush();
op2.flush();
}
on_error {
op1.discard();
op2.discard();
}
map op1.queueData($1), data1.iterator();
map op2.queueData($1), data2.iterator();
}
- Note
- Wach bulk DML object must be manually flush()ed before committing or manually discard()ed before rolling back to ensure that all data is managed properly in the same transaction and to ensure that no exception is thrown in the destructor(). See the example above for more information.