Row-based triggering

The row-based triggering design enables a fault-tolerant, multi-server-aware triggering mechanism that can enable maps to selectively process those rows associated with a triggered watch event. In the preceding sentence, fault-tolerant means that data is not lost if a process fails, a connection is dropped, an Launcher stops, or some other unforeseen interruption occurs. Multi-server-aware means that Launchers on different machines can connect to and "watch" the same table.

The main difference between row-based and table-based triggering is that with row-based triggering, the query executes on only those rows that have been updated or inserted, not on the entire table. Additionally, batch-like processing is possible if the When clause is used to control the time and/or date at/upon which processing should occur.

When using column-based triggering, batch processing does not occur.

To see how to construct the WHEN expression for row-based triggering, refer to the example in SELECT 1 FROM Format.