|
The Hash Antisemijoin variant of the Hash Join algorithm performs an antisemijoin between the left-hand side and the right-hand side. The right-hand side is only used to determine which rows from the left-hand side appear in the result. With Hash Antisemijoin, the right-hand side is read to form an in-memory hash table that is subsequently probed by each row from the left-hand side. Each left-hand row is output only if it fails to match any row from the right-hand side. Hash Antisemijoin is used in cases where the join's inputs include table expressions from a quantified (NOT IN, ALL, NOT EXISTS) nested query that can be rewritten as an antijoin. Hash antisemijoin tends to outperform the evaluation of the search condition referencing the quantified query if a suitable index does not exist to make indexed retrieval of the right-hand side sufficiently inexpensive.
As with Hash Join, the Hash Antisemijoin algorithm may revert to a nested loops strategy if there is insufficient cache memory to enable the operation to complete. Should this occur, a performance counter is incremented. You can read this monitor with the QueryLowMemoryStrategy database or connection property, or in the Query: Low Memory Strategies counter in the Windows Performance Monitor.
Memory governor limits are dependent upon the server's multiprogramming level and the number of active connections. See Threading in SQL Anywhere, and Setting the database server's multiprogramming level.
NoteThe Windows Performance Monitor may not be available on Windows CE. |
For more information, see QueryLowMemoryStrategy in Connection-level properties.