2023)¶

New Features and Improvements¶

Bodo: - Vectorized query engine:

Several operations (join and reads) now operate in 'vectorized' mode. Only results that are in use are kept in memory or read in. This leads to drastically improved memory performance and avoids OOM errors in many cases. There is a slight performance degradation in this mode, but this will improve with coming releases.

Some operations, such as GROUP BY and window functions are not yet vectorized.

Vectorized execution mode can be disabled by setting the environment variable BODO_STREAMING_ENABLED to "0".

Common subcolumn elimination in planner
Better selectivity in metadata pushdown query
Better type size estimates in planner
Support ignore_index=True in df.drop_duplicates

BodoSQL:

Added support for the following functions:

PERCENTILE_CONT
SHA2
MD5

Parity improvements: - Make treatment of type keywords compatible with Snowflake SQL

Planner improvements:

Improved planner based common sub-expression elimination
Improved planner estimations for column selectivity
Improved planner estimations for column sizes
Planner based Snowflake pushdown support for getdate(), current_date(), and intervals