Repository navigation
native_datafusion performance improvement #3748
Description
What is the problem the feature request solves?
Comparing to native_iceberg_compat the native_datafusion takes twice more memory on pods and after #3046 the scan grew slower, prev it was at least 20% faster. It might be a problem with aggressive SchemaAdapter transformations
Describe the potential solution
No response
Additional context
No response
Activity
comphead commentedon Mar 26, 2026
Part of it apache/datafusion#21158 which bypasses schema adapter rewrite costs if there is no need in rewrite(like no filter involved and logical/physical schemas are the same)