Apache Arrow ADBC 21 (Libraries) Release

Published 07 Nov 2025
By The Apache Arrow PMC (pmc)

The Apache Arrow team is pleased to announce the version 21 release of the Apache Arrow ADBC libraries. This release includes 30 resolved issues from 23 distinct contributors.

This is a release of the libraries, which are at version 21. The API specification is versioned separately and is at version 1.1.0.

The subcomponents are versioned independently:

  • C/C++/GLib/Go/Python/Ruby: 1.9.0
  • C#: 0.21.0
  • Java: 0.21.0
  • R: 0.21.0
  • Rust: 0.21.0

The release notes below are not exhaustive and only expose selected highlights of the release. Many other bugfixes and improvements have been made: we refer you to the complete changelog.

Release Highlights

Language bindings:

  • The driver manager reports better errors when it fails to find a driver (#3646).
  • The Python driver manager now searches for manifests in the proper location when inside a Python virtual environment (#3490).
  • Added convenience methods to the driver manager on top of the standard DBAPI-2.0/PEP249 APIs (#3539).
  • The signature for connect has been simplified, so you can connect without having to repeat the driver name or specify explicit keyword arguments depending on the driver (#3537).
  • Support for Python 3.9 has been dropped (#3573, #3663).
  • Support for Python 3.14 (including the free-threading variant) has been added (#3575, #3620, #3663).
  • The R bindings now support replace and create_append ingest modes.

Drivers:

  • The Go BigQuery driver now supports service account impersonation (#3488) and setting a quota project (#3622).
  • The Go BigQuery driver now returns more detailed type metadata in result sets (#3604).
  • The C# BigQuery driver allows setting a location (#3494).
  • The C# Databricks driver adjusted default settings to make small queries faster (#3489).
  • Memory usage of the C# Databricks driver was improved (#3652, #3656).
  • All C# HiveServer2-based drivers (Hive, Impala, Spark, Databricks) throw Unauthorized exceptions when appropriate (#3551).
  • JNI bindings to the C++ driver manager are now released, making it possible to use non-Java drivers from a Java application (in a very limited fashion) (#3429). Binary artifacts are available for amd64/arm64 Linux, arm64 macOS, and amd64 Windows.
  • The PostgreSQL driver can now return the schema of any bind parameters in a prepared query ([#3579[(https://github.com/apache/arrow-adbc/pull/3579)]]).
  • The PostgreSQL driver properly batches result sets with large string/binary values now (#3616).
  • The Snowflake driver now returns float64 for numeric columns when use_high_precision is false and scale is nonzero (#3295). (Previously it incorrectly truncated to int64.)
  • Added an option to disable the "vectorized" scanner when ingesting data into Snowflake, which sometimes appeared to cause performance issues (#3555).

Packaging:

  • Added AlmaLinux 10 (#3514).
  • Added Debian Trixie (#3513).

Contributors

$ git shortlog --perl-regexp --author='^((?!dependabot\[bot\]).*)$' -sn apache-arrow-adbc-20..apache-arrow-adbc-21
    28	David Li
     9	Bruce Irschick
     7	eric-wang-1990
     6	Curt Hagenlocher
     5	eitsupi
     4	msrathore-db
     3	Bryce Mecum
     3	Mandukhai Alimaa
     3	Sutou Kouhei
     3	davidhcoe
     2	Anna Lee
     2	Jason Lin
     2	Kevin Liu
     1	Dewey Dunnington
     1	Ian Cook
     1	Jacky Hu
     1	Jade Wang
     1	Kristin Cowalcijk
     1	Lucas Valente
     1	Matthijs Brobbel
     1	bruceNu1l
     1	praveentandra
     1	rnowacoski

Roadmap

We are starting work on async interfaces and other API enhancements, and welcome comments or contributions from anyone interested. See the initial pull requests:

Getting Involved

We welcome questions and contributions from all interested. Issues can be filed on GitHub, and questions can be directed to GitHub or the Arrow mailing lists.