Community Weekly Report
Sep 21 – Sep 27, 2026
00Overview
This week, Doris continued toward more complete lakehouse reads and writes, more flexible semi-structured analytics, and broader query capabilities for resource-constrained and highly concurrent workloads; 44 contributors helped merge 97 PRs, while the community opened 11 new issues. Kernel and engineering remained the main focus, followed by real-time warehousing, pointing to stronger execution foundations alongside broader lakehouse interoperability and online query serving.
01Highlights
02By Scenario
Multi-modal Lakehouse
01For analytics combining AI and lake tables, embedding models gain independent endpoint, credential and model settings, with fallback to general AI configuration. Iceberg nested decimals now accept precision widening at unchanged scale, improving service configuration flexibility and schema-evolution compatibility.
- [Enhance](ai_func) Support dedicate embed properties in AI RESOURCE
- [fix](iceberg) Allow nested decimal precision widening
Lakehouse work pushes selective partition filtering into Hive Metastore to avoid full listings on heavily partitioned tables. Newly opened Hive/Hudi partition-value pushdown would answer partition-column queries without opening data files, extending metadata-driven execution. Oracle floating-point mappings also broaden cross-source compatibility.
- [improvement](hive) Push partition filters to HMS
- [opt](jdbc) Support Oracle BINARY_FLOAT and BINARY_DOUBLE types
- [improvement](hive) Support partition-column-value-only pushdown for Hive/Hudi scan
