批处理插入返回标识和对象引用/序列

Question

With SQL Server 2K8 from C# I'm trying to do a batch insert/updates of records to a parent/child tables to optimize.

The inserts/updates will generate a key automatically which I'd like to extract via an OUTPUT, etc. and then reassign back in the domain model. For batch inserts I need to keep track of which newly generated ID belongs to which domain object in the batch list.

This example comes close to what I need, but was wondering if there's a way to not have an extra column added to the table (SequenceNumber) and still achieve the same results: http://illdata.com/blog/2010/01/13/sql-server-batch-inserts-of-parentchild-data-with-ibatis/

ie. could we rely on the order of the inserts generated from the OUTPUT into the temp table, or pass a ref GUID set on the data model and passed temporarily to the SQL just for reference purposes?

Answer 1

In SQL Server 2008 it is possible to use merge and output to get a mapping between the generated key and the key used in the staging table.

Have a look at this question. Using merge..output to get mapping between source.id and target.id

Answer 2

Unless I've misunderstood...

A surrogate key (IDENTITY or NEWID etc) isn't your actual object identifier. It's an implementation detail and has no intrinsic meaning.

You must have another identifier (name, ISBN, serial number, transaction code/date, etc) that is the real (natural) key.

Your OUTPUT clause can return the surrogate key and the natural key. You then use this to map back