Community Report — Apache Doris

00Overview

This week Apache Doris merged 113 PRs with 42 contributors active, with the community concentrating on real-time data warehousing and kernel engineering while advancing the multi-modal lakehouse and cloud-native compute-storage separation tracks. Core capabilities deepened around Catalog SPI decoupling, scan execution optimization, inverted index evolution, and peer routing with cache tiering, alongside reinforced security/governance and observability—signaling a clear direction toward a high-performance, extensible lakehouse.

01Highlights

02By Scenario

Multi-modal Lakehouse

01

Merged

The multi-modal lakehouse scenario stabilized Iceberg/Hive/external-write paths. Iceberg v3 row-lineage columns are rejected to avoid hidden conflicts; COUNT(*) pushdown tolerates missing snapshot summary counters; LZ4 compression now ships to Iceberg/Hive Parquet/ORC writers; binary columns are emitted with correct Arrow types; nested decimal precision promotion is supported; external writers clean up

In progress

Multi-modal Lakehouse work strengthens external read/write paths: Paimon write P0, Iceberg position-deletes system table, ORC format v2 reader, Parquet dict bounds checks, Arrow Flight coordinator lifecycle, OSS/HTTP FS SPI normalization, MaxCompute partition-by-name parsing, Iceberg REST OAuth re-auth on 401, S3 load complex types for Parquet/ORC, and Hive-ORC column-name default —

03Community Pulse

113Merged PRs

5New issues

42Contributors

00Overview

01Highlights

01Catalog SPI decoupling and external data source expansion

02Cloud-native compute-storage separation: peer routing and cache tiering

03Scan and query execution: pruning, lazy read, and runtime filters

04Inverted index evolution: Japanese tokenizer and SNII format

05Security and governance hardening across FE and TDE

02By Scenario

Multi-modal Lakehouse

03Community Pulse