Dear All,
The runner for Spark 2 was deprecated quite a while back in August 2022 with
the release of Beam 2.41.0 [1]. We’re planning to move ahead with this and
finally remove support for Spark 2 (beam-runners-spark) to only maintain
support for Spark 3 (beam-runners-spark-3) going forward.
N
I am not sure I understand the question, but I do see an issue.
Context: "CalcRel" is an optimized relational operation that is somewhat
like ParDo, with a small snippet of a single-assignment DSL embedded in it.
Calcite will choose to merge all the projects and filters into the node,
and then gen
Looking at the code that turns a logical CalcRel into a BeamCalcRel I do
not see any obvious cause for this:
https://github.com/apache/beam/blob/b3aa2e89489898f8c760294ba4dba2310ac53e70/sdks/java/extensions/sql/src/main/java/org/apache/beam/sdk/extensions/sql/impl/rule/BeamCalcRule.java#L69
I don'