Omnidex Release Summary & Fixes

• Support for native 64-bit on Windows, Linux and HP-UX
• Support for Oracle 11g on hpux2
• Support for MySQL 5.01 on Windows and Linux using the DataAccess ODBC Driver

• Support for detailed, real-time statistics from the UPDATE STATISTICS command
• Support in ODXSQL for api=JDBC
• Support for UPDATE INDEXES command to build Omnidex Indexes from ODXSQL or oaexecsql
• Support for LOAD OFX, UPDATE ROLLUPS and UPDATE TEXT in oaexecsql as well as ODXSQL
• Support for XML environment files in OmniAccess, with support in oadecomp and oahelper
• Support more efficient given name synonym lists with a maximum of eight synonyms
• Change oaglobal.env to Omnidex.env, allowing “connect omnidex” without a path
• Change OAGLOBAL environment variable to OMNIDEX_ENV

- Support for -odxnet option in ODXSQL to automatically launch odxnet - Support for both 32-bit and 64-bit C DATETIME datatypes - Fix problems with SET CONTINUATION OFF - Fix problem with $BITMAP_INDEX and other columns in system catalog table $INDEXES - Fix signal handler in Java scenarios. [Note JBOSS must still be run with alternate Java DLL.] - Allow $SCORE and $CONTAINS even when no proximity qualifications have occurred - Support for CLOB's in OST's. - Show file sizes for flatfile databases in ODXSQL and info calls. Info call structure and Makefile changes were made.

• Support oafetchkeys with KEYWD or KWCOUNT options on Distributed Indexes
• Support the column alias also being the name of a column in the table.
• Support NULL values for OMNIDEX, ANSI, ASCII and C DATE datatypes.
• Support for Java 1.5 on UNIX platforms
• Automatically add App_Themes directory on first use of OmniSearch

• Beta release of Omnidex JDBC Intelligent Router
• Support for JDK and JRE 1.5 with our JDBC driver
• Translate environment variables in datasources for JDBC

- Support for nested queries in the FROM clauses within the WHERE clause - Optimization of nested queries against partition qualifiers within Union Views - Optimization using rollup tables within Union Views - Improved optimization in Union Views within columns using the DISTINCT_IN_PARTITION option - Improve memory mapping and exclusive access for files - Fix to avoid processing grouped COUNT(DISTINCT …) using rollup tables unless all group by columns match the rollup table - Fix display of counts for OmniSearch ActiveCounts template Makefile changes?

- Changes to compilation of JDBC driver to use new build_jdbc.ksh script - Changes to profile to set JAVA_HOME, JAXB_HOME, DBCP_HOME, POOL_HOME and JTDS_HOME

- Optimization of nested queries against partition qualifiers in Union Views. - Optimization using rollup tables in Union Views.

- Fix problems with oastatementinfo in queries against grids (DBUNION rule)

Omnidex Build 4.4.01B (July 2007)

- Support for the DataAccess ODBC driver on Linux, with testing against FreeTDS

- Fix DSEdit and OdxQuery to allow them to work with 4.4.n releases. - Fix Omnidex Addin problem with DBUNION (Grid) statements.

Makefile changes?

Yes, compile DataAccess ODBC driver on Linux

Omnidex Build 4.4.01A (June 2007)

- Optimization of TOP, SKIP, EVERY and RANDOM clause of SELECT statements to take advantage of skips within indexing. - Improve handling of TOP and SKIP operations in the grid processing with the Addin (DBUNION rule). - Change version checking between client and server so that a) ODBC and JDBC are based on an earliest compatible version, and b) CLOA is based on the Omnidex version, taking into consider whether info calls are to be used. For CLOA, the new property NoInfoCalls=Yes is supported, and if set, allows versions to sync on just x.xx, rather than x.xx.xx. – oaversion now supports showing the earliest - Alpha internal release of OmniSearch

- Improve debug messages to show better operating system information in banners and trace files - Fix abort that occurs when column alias is a literal, but order by is a column. - Fix to oainfo to resolve problems with info calls in the ODBC driver - Improve performance of Hashed Data Cache on large result sets.

Makefile changes?

- Windows: add OmniSearch: 1. Omnidex.Web.OmniSearch.dll 2. Omnidex.OmniAccess.EA.dll 3. OmniSearchWizard.exe 4. config\english\omnisearch\*.xml - Move Xerces libraries to OMNIDEX_LIBS directory - Add version.lib reference for odxdl library

Internal: Changes for explain plans and introduction of config\english\omnisearch\*.xml

Omnidex Build 4.3.4A (April 2007)

- Port to 64-bit Linux

- Change oahelper’s generation of Omnidex installation scripts from a Unix shell script to a two input files, one for installing indexes and one for building indexes - Change oahelper’s environment source generation to declare the database type as SQL Server rather than ODBC on SQL Server databases - Support for SET CONTINUATION ON | OFF in ODXSQL to support Oracle or SQL Server style of entering multi-line statements. - Fix OMNIDEX_DEBUG=ALL to display DataAccess driver messages as well - Change licensing checks in Omnidex so that more utility operations are allowed when unlicensed, while still prohibiting database and index access. - Fix problem where the OFX index files were not being deleted along with their TDF data files - Fix problem in the Omnidex kernel where the pre-intersect counts were inaccurate on bitmap-indexed columns. - Internal only: change odxget’s count parameter to a 32-bit integer to allow larger skip forward or skip backward operations.

Makefile changes?

- Changes for port to 64-bit Linux - Change location of Xerces libraries to DBS2_HOME - Change to XSD version 2.3.1

Moderate

- Support for optimized radius searches, including o Support for $distance function o Support for USAGE LATITUDE, USAGE LONGITUDE in the environment file o Optimization of $distance functions against Omnidex composite keys containing longitude and then latitude

- Support for Rollup Tables, including o Support for declaration of Rollup Tables in the environment files o Support for automatically optimizing queries using rollup tables o Support for UPDATE ROLLUPS command in ODXSQL o Support for metadata about Rollup Tables in oatableinfo

- Support for PARTITION BY clauses containing more than one predicate, such as STATE in (‘CO’) and ZIP between (‘80000’ and ‘80999’)

- Support for Oracle 8 on Windows platforms - Support for Oracle 10g on Linux platforms - Fix error in Omnidex kernel in large wildcard qualifications, such as 1995@ (Axciom) - Fix error in Omnidex kernel dealing with intersecting large bitmap ID lists - Allow partitioning of all tables for grid environments, rather than restricting to child tables - Improve table join support for grid environments - Improve temporary file naming routines to make more unique for high-speed, multithreaded operations - Disallow the LOAD OFX FOR TDF … command since it is replaced by LOAD OFX FOR TABLE … command

Makefile changes?

- Preliminary support for XML metadata requires shipping the Xerces library with Omnidex - Changes to HPUX2 compile flags - Changes to HPUX3 compile flags - Variety of new members

Info call changes made for rollup tables.

- Support for creating/attaching a temp table with a specified access level. - Improved handling of semaphores - Fix eadecomp to preserve “partition by” clauses - Fix oahelper to load correct .dll based on version number

- Support for new PRESERVE_CRITERIA option to oaselect - Improved performance when checking for blank char fields - handling of incorrectly formulated in-clauses (e.g., “gender in (M, F, G)” instead of “gender in ('M', 'F', 'G')” ) - Fix partition table problem when “env.db.tbl” is used - Fix OPEN CURSOR and CLOSE CURSOR through cloa - Fix SNOWFLAKE_UNJOIN problem of changing “or” to “and” in the where clause - Remove reliance on OMNIDEX_LIBS

Makefile changes? - modified Makefiles: da/Makefile, oa2/Makefile, odbc/Makefile - new Makefiles: pl/Makedb2, pl/Makessv, pl/Makeora, pl/Makefile, jdbc/Makefile - Macros.master changes affecting all platforms

• Support for multiple OR’d values in partition qualifiers for aggregation indexes
• Support for partition qualifiers in aggregation indexes that have partial columns as the first component
• Support for partition qualifiers in aggregation indexes that have partial columns as the first component
• Support for the DISTINCT_IN_PARTITION option to the COLUMN statement within Union Views.

This option indicates that count distinct operations on this column are better optimized because the administrator guarantees that values are distinct within each partition.

• Support for the CACHE/NOCACHE options in the DBUNION rules in environment files.

This option indicates that the grid processing should cache connections to the node so that repeated calls to CLoaconnect, CLoaopencursor, CLoaclosecursor and CLoadisconnect are not needed for each node within each query.

• Extend limit for Alt Sorts using Aggregation Indexes from 16 components to 32 components
• Enhance Union View processing to honor DISTINCT_IN_PARTITION clauses on SELECT DISTINCT … statements.
• Support for Postgres RDBMS using our DataAccess ODBC driver
• Further support for partial-length components as partition qualifiers in Aggregation Indexes.
• Change sequence of instructions in Omnidex kernel to circumvent a Microsoft compiler optimization error
• Enhance grid (DBUNION) processing to honor UNIQUE_TO_PARTITION clauses on SELECT DISTINCT … statements.
• Support for compiling and running in 32-bit mode on 64-bit Windows architecture.
• Improve performance on Record Specific processing on Windows.
• Support multiple databases in NODE section and multiple table.column in the UNIQUE_TO_NODE clause of DBUNION rule.
• Fix error on criteria with long quoted strings
• Favor ASK over MDK optimization for distinct key processing.
• Fix problem with temporary files on certain aggregations.