The content below is an LLM summary of this month’s activity on the mailing list.

• Java Extension Type Refactoring: Discussions are underway to simplify Java ExtensionTypeWriter creation by transitioning from a factory-based pattern to a more type-based approach. This proposed change aims to reduce boilerplate and simplify the API but introduces breaking changes planned for the v18.4.0 release. Thread
• High-Precision Temporal Types: Initial discussions for introducing picosecond-precision timestamp and duration types have evolved. The current proposal favors a (seconds: i64, picoseconds: i64) pair over a single 128-bit integer, aligning with existing database implementations like BigQuery and Trino, and is now being considered as a canonical extension type. Thread
• Late Materialization for Parquet: The community is discussing supporting “late materialization” for new Parquet encodings (like ALP and FSST) within Arrow. This aims to allow operations on encoded data longer for performance benefits, potentially requiring new Arrow vector types or opaque vectors. Thread

• Canonical Extension Types: The concept of canonical extension types is maturing, with a TimestampWithOffset type (UTC timestamp + offset in minutes) recently voted in and merged to enhance compatibility with ANSI SQL’s TIMESTAMP WITH TIME ZONE. Thread Thread
• Arrow Flight Ecosystem Expansion: A new project, DACP (Data Access and Collaboration Protocol), leveraging Arrow Flight for cross-node data access, catalog management, and collaboration, is proposing integration with the Arrow ecosystem, indicating growing interest in Flight-based solutions. Thread
• Cross-Project Collaboration & Onboarding: A key contributor from Apache Spark has begun making small contributions to Arrow, signaling a focus on deeper integration between the two projects and highlighting the ongoing effort to onboard new community members. Thread

• CI Instability: Frequent, often unrelated, GitHub Actions failures continue to be observed, attributed to third-party dependencies or CI platform issues, which can impede consistent development and validation cycles. Thread
• Go Release Memory Leak Issues: The Go 18.5.0 release candidate (RC0) faced significant memory leak detection issues with Go 1.25’s new leak checker, necessitating fixes to both test structures and identified code leaks before a successful RC1 vote and release. Thread Thread

• Decentralizing ADBC Driver Development: A new adbc-drivers GitHub organization, “ADBC Driver Foundry,” has been established to accelerate ADBC driver development and address bottlenecks. Existing key drivers (BigQuery, Databricks, Snowflake) are migrating, signaling a strategic focus on expanding and streamlining the ADBC ecosystem. Thread
• Project Leadership Transition: Neal Richardson has stepped down as PMC chair, with Antoine Pitrou duly elected as the new VP and PMC chair, ensuring continued leadership rotation and governance within the project. Thread