Question 5 - DEA-C01 discussion

When created, a stream logically takes an initial snapshot of every row in the source object (e.g. table, external table, or the underlying tables for a view) by initializing a point in time (called an offset) as the current transactional version of the object. The change tracking system utilized by the stream then records information about the DML changes after this snapshot was taken. Change records provide the state of a row before and after the change. Change information mirrors the column structure of the tracked source object and includes additional metadata columns that describe each change event.

Note that a stream itself does not contain any table data. A stream only stores an offset for the source object and returns CDC records by leveraging the versioning history for the source object.

A new table version is created whenever a transaction that includes one or more DML statements is committed to the table.

In the transaction history for a table, a stream offset is located between two table versions. Querying a stream returns the changes caused by transactions committed after the offset and at or before the current time.

Multiple queries can independently consume the same change data from a stream without changing the offset. A stream advances the offset only when it is used in a DML transaction. This behavior applies to both explicit and autocommit transactions. (By default, when a DML statement is execut-ed, an autocommit transaction is implicitly started and the transaction is committed at the completion of the statement. This behavior is controlled with the AUTOCOMMIT parameter.) Querying a stream alone does not advance its offset, even within an explicit transaction; the stream contents must be consumed in a DML statement.

To advance the offset of a stream to the current table version without consuming the change data in a DML operation, complete either of the following actions:

· Recreate the stream (using the CREATE OR REPLACE STREAM syntax).

Insert the current change data into a temporary table. In the INSERT statement, query the stream but include a WHERE clause that filters out all of the change data (e.g. WHERE 0 = 1).