LucidDB 0.7.2 release is now available

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

LucidDB 0.7.2 release is now available

John Sichi
Download links are available from, or from

Note that the LucidDB installer does not yet support automatic upgrade,
so be sure to install the new release in a new location if you already
have an old version installed.  However, there is now a manual upgrade
procedure available:

It's a little bit of effort, but I've tested it out and it should work
(it's similar to a scripted version used internally at LucidEra).  Let
me know if you have trouble with it, or would like to contribute
enhancements to the installer.

The highlight of the release is sampling support for ANALYZE (via the
new ESTIMATE STATISTICS option).  What formerly took hours for a large
table should now take minutes, without significant loss of accuracy in
the resulting stats (in most cases).

A number of new documents have also been added (including some in-depth
information on how the optimizer handles joins, indexes, and subqueries):

Changes since 0.7.1 include

- sampling support for ANALYZE
- significant performance improvement for large DROP operations
- additional system views DBA_COLUMN_STATS and DBA_COLUMN_HISTOGRAMS
- table name mapping support in JDBC foreign data wrapper
- new system parameter "deviceSchedulerType" for controlling I/O impl
- new VJDBC listener system parameter "connectionTimeoutMillis"
- new session parameter "lastRowsRejected"
- new performance counters
   "DatabasePagesAllocatedAfterReclaim", "TempPagesAllocatedAfterReclaim"
- miscellaneous optimizer improvements and fixes
- optimize away distinct agg for semijoins where the keys are already unique
- improve rowcount inference for UDX's (FRG-289)
- support optimization for functional expressions in nested loop join
- optimize away nested loop join temp index when LHS input is empty
- fixes for a few problems in efficient subquery decorrelation
- increase applib function input/output lengths
- fixes for page versioning concurrency and other code review findings
- fix a bad cache page access during index search
- fixes for some rare bitmap index corruptions during incremental load
- some segment-level tracing fixes
- improved error reporting for failures setting large cache sizes
- fix a rare problem with simultaneous commits (FNL-68)
- fix decimal rounding mode for converting literals in VALUES clause
- fix exception handling issues with JNI (FNL-67)
- change LucidDB mbean to use local JDBC driver rather than RMI
- improve EXPLAIN PLAN readability for GROUP BY
- improve rejected row logfile naming