Planner rule that tries to push limit into a
FlinkLogicalTableSourceScan, which table is
a
TableSourceTable. And the table source in the table is a
SupportsLimitPushDown.
The original limit will still be retained. The reasons why the limit still be retained: 1.If the
source is required to return the exact number of limit number, the implementation of the source
is highly required. The source is required to accurately control the record number of split, and
the parallelism setting also need to be adjusted accordingly. 2.When remove the limit, maybe
filter will be pushed down to the source after limit pushed down. The source need know it should
do limit first and do the filter later, it is hard to implement. 3.We can support limit with
offset, we can push down offset + fetch to table source.