Community Weekly Report
Jul 27 – Aug 2, 2026
00Overview
This week the Doris community merged 79 PRs from 47 contributors with sustained issue activity, with the capability lines clearly converging on real-time data warehouse and kernel engineering while accelerating toward lakehouse and compute-storage-separated, cloud-native architectures. Headline work on Catalog SPI, cloud-native file cache, and cloud-mode MOW throughput advances alongside Iceberg/Hive read-write parity and colocation/runtime-filter/optimizer pushdowns, signaling strong momentum toward a cloud-native, lakehouse-integrated engine built around query performance and elastic scale.
01Highlights
02By Scenario
Multi-modal Lakehouse
01The lakehouse stack matures across Paimon, Iceberg and Parquet: Paimon reads on BE no longer reconstruct the catalog metastore, paimon.table-option.* passthrough and query-level dynamic options land with regression coverage, Iceberg historical reads and nested schema evolution are fixed, V2 lazy dictionary and direct predicates cut TPC-DS sparse RF scan CPU,
- [fix](paimon) Read Paimon tables on BE without reconstructing the catalog metastore
- [test](iceberg) Add partition evolution schema coverage
- [feature](paimon) Support Paimon table option passthrough
- [feature](paimon) Support query-level dynamic options
- [fix](lakehouse) Fix Iceberg historical reads and nested schema evolution
This group advances read/write quality for the multi-modal lakehouse. Recent work targets metadata cache and refresh semantics: information_schema queries on Iceberg no longer block on a loadTable per table, SHOW TABLES avoids eagerly warming every entry, Paimon JNI gains a table cache to relieve memory pressure, and fixes close gaps
