Community Weekly Report
Sep 14 – Sep 20, 2026
00Overview
This week 48 contributors merged 144 PRs, with kernel engineering remaining the dominant focus, closely followed by compute-storage separation and real-time data warehouse work—signaling Doris's steady march toward a more elastic cloud-native architecture and stronger real-time processing. Progress on lakehouse format interoperability, full-text search acceleration, and enterprise security governance advanced in parallel, while emerging momentum in security governance and agent observability points to the platform's expansion beyond core performance into broader enterprise and intelligent-application scenarios.
01Highlights
02By Scenario
Multi-modal Lakehouse
01Lakehouse federation correctness improved across Iceberg, Paimon, and Hive: schema-evolution-aware equality deletes, safer name-mapping validation, missing-file handling, and identity-partition preservation for Iceberg; schema/timestamp precision fixes for Paimon; and OpenCSVSerde-compatible parsing for Hive CSV tables, reducing silent data corruption and inconsistent query results across external catalogs.
- [fix](iceberg) Fix historical scans after schema evolution
- [fix](paimon) Preserve statement schemas and timestamp precision
- [fix](iceberg) Reject invalid name mapping instead of reading NULL
- [fix](be) Truncate native ORC timestamps to microseconds
- [fix](hive) Match OpenCSVSerde semantics in FileScannerV2
Multi-modal lakehouse work fixes native Azure credential routing for Iceberg files, plus newly opened fixes for Iceberg/Parquet predicate pushdown mis-pruning matching rows due to NaN handling and legacy timestamp statistics—preventing silent data loss from inaccurate metadata—alongside AI resource-related fixes, improving federated lakehouse query correctness.
