Search before asking
Motivation
Currently, Fluss virtual log tables ($changelog and $binlog) only implement ScanTableSource.
Compared with normal Fluss tables, virtual log tables do not support projection pushdown or filter pushdown yet. As a result, queries on $changelog / $binlog always read and deserialize full records even when only a few columns are required, and filters cannot be used to prune partitions, log batches, or metadata ranges early.
This issue aims to align virtual log tables with normal Fluss tables by adding projection and filter pushdown support.
Solution
No response
Anything else?
No response
Willingness to contribute
Search before asking
Motivation
Currently, Fluss virtual log tables (
$changelogand$binlog) only implementScanTableSource.Compared with normal Fluss tables, virtual log tables do not support projection pushdown or filter pushdown yet. As a result, queries on
$changelog/$binlogalways read and deserialize full records even when only a few columns are required, and filters cannot be used to prune partitions, log batches, or metadata ranges early.This issue aims to align virtual log tables with normal Fluss tables by adding projection and filter pushdown support.
Solution
No response
Anything else?
No response
Willingness to contribute