2023)¶

New Features and Improvements¶

Support for reading tables outside the default database.
Initial support for expanding Snowflake views. When Bodo encounters a Snowflake view definition it will attempt to expand the view directly into the query, which will provide increased optimization opportunities and performance. Bodo will not be able to expand views if the user under which Bodo operates does not have permissions to access the underlying table or fetch the view definition. Bodo will not attempt to expand the view if it is a materialized view or a secure view, to comply with Snowflake users' expectations. If for any reason Bodo is unable to expand a view then the query will still execute by evaluating the view in Snowflake and reading it as a table. This is the first release offering this support, so the number of views that can be expanded will increase with future releases.
Added support for the function OBJECT_CONSTRUCT including the syntactic sugar OBJECT_CONSTRUCT(*) so long as all of the values are of the same type (or trivially castable types).
Add support for the functions ARRAY_CONSTRUCT_COMPACT, ARRAY_REMOVE, ARRAY_REMOVE_AT, ARRAY_SLICE and TO_VARIANT.
Allow FLATTEN to be called on variant columns, so long as the variants contain array/json data.
Bodo can now spill to S3.
Support for GET function on array values

Reduced peak memory usage when using the FLATTEN or SPLIT_TO_TABLE functions in streaming.

Added support for the functions IS_ARRAY, IS_OBJECT and OBJECT_PICK.
Added support for TO_ARRAY for struct and map elements.
Added support for nulls in map array values.
Support for parsing array literals. Note: just like ARRAY_CONSTRUCT, only homogeneous arrays are supported at this time.
Support for nested array scalars.
View inlining now supports all valid Snowflake CREATE VIEW syntax for the view definition.

Process join data communicated across ranks in batches to reduce peak memory consumption and improve cache locality for better performance.

Support OBJECT_INSERT
Support for writing columns of unnested JSON data with homogenous key and value types to Snowflake as Objects
Improved error messages for Unsupported Snowflake UDFs. Future releases will contain increased Snowflake UDF support
Support COALESCE with Datetime and Timestamp + timezone data
Support VALUES syntax with multiple rows
Support for struct columns with no fields

Improved ability to push filters down to Snowflake and order joins based on the distinctness of columns from inlined views.

Fixed a bug where FROM_DATE incorrectly returned TIMESTAMPs when the inputs were DATEs