README for HP Neoview Release 2.4

Database
These are the new database features:
• “Compiler and Executor Enhancements” (page 7)
• “Automatic Clustering of ON RECOMPUTE MVs” (page 7)
• “Automatic Query Retry” (page 8)
• “CONVERTOHEX Function” (page 9)
• “Enhancements to MERGE INTO” (page 9)
• “Extensions to MAINTAIN REORG” (page 9)
• “FORMAT Option” (page 9)
• “New UPDATE STATISTICS Automation CHG_AUTO_LIST Command” (page 9)
• “Parser-Syntax Error Enhancement” (page 9)
• “Schema Privileges in Release 2.4” (page 10)
• “Schema Version 2400” (page 10)
• “Sequence Generator” (page 10)
Compiler and Executor Enhancements
Several enhancements have been made to the Neoview Compiler and Executor, among them:
• The Compiler:
— Removes redundant joins and one of the associated scans form the query plan.
— Transforms a correlated subquery to joins to reduce the number of table scans.
— Now supports asymmetric joins.
— Reduces sort materialization resulting in reduced disk I/O and increased pipeline
parallelism.
— Implements Outer Child Repartitioning(OCR) to improve scalability.
• The Executor:
— Now has faster in-memory hash-join.
— Has implemented overflow improvements for big memory operators.
— Has implemented predicate reordering optimization.
Automatic Clustering of ON RECOMPUTE MVs
Before Release 2.4, it was necessary to specify a STORE BY clause for RECOMPUTE materialized
views (MVs). In Release 2.4, this is no longer required. However, because of the nature of ON
RECOMPUTE MVs, the user still has the option of specifying clustering columns for an MV. The
STORE BY clause is optional.
An ON RECOMPUTE MV can be clustered in four ways:
In this method, the MV . . .Clustering Method
Is clustered by the user-specified clustering columns and a
SYSKEY.
The user specifies a clustering key.
Has a primary key composed of the table/MV CI columns.The MV is defined on a single table that has a primary
key, and all table (clustering index (CI)) columns are
used by the MV.
Is clustered by the table CI columns used by the MV plus a
SYSKEY.
At least one of the tables on which the MV is defined
has one or more of its CI columns used by the MV.
Is a single partition MV clustered by SYSKEY.None of the tables used by the MV has one of its CI
columns used by the MV.
New Features for Release 2.4 7