Repository navigation
Follow-up to #341: next readers set from DB-Engines ranking #570
Description
Activity
We have ADBC drivers available for most of these:
System Compatible ADBC driver Notes Microsoft Azure SQL Database https://adbc-drivers.org/drivers/mssql/ Presto https://adbc-drivers.org/drivers/presto/ Microsoft Fabric / Azure Synapse https://adbc-drivers.org/drivers/mssql/ Apache Spark SQL https://adbc-drivers.org/drivers/spark/ Aurora, TimescaleDB, Greenplum, CockroachDB, YugabyteDB https://arrow.apache.org/adbc/current/driver/postgresql.html Tested with TimescaleDB, CockroachDB, YugabyteDB; see https://github.com/columnar-tech/adbc-quickstarts/tree/by-database and [1] TiDB, SingleStore, StarRocks https://adbc-drivers.org/drivers/mysql/ Tested with TiDB and Vitess; see https://github.com/columnar-tech/adbc-quickstarts and [2] QuestDB coming soon Ask Ian for details IBM Db2 IBM has been nonresponsive recently Apache Hive Is there actually demand for this? If so we might have a solution SAP HANA https://docs.columnar.tech/drivers/sap-hana/ We're prepared to make this free and open source if someone at SAP could please to talk to us Teradata https://docs.columnar.tech/drivers/teradata/ We're prepared to make this free and open source if someone at Teradata could please to talk to us SAP Adaptive Server (Sybase) Apache Impala Is there actually demand for this? If so we might have a solution Vertica coming soon Ask Ian for details Informix / Netezza [1] For SingleStore, use the SingleStore ADBC driver: https://adbc-drivers.org/drivers/singlestore/
[2] For Doris, use the Flight SQL ADBC driver: https://arrow.apache.org/adbc/current/driver/flight_sql.htmlAlso chDB to complete the ClickHouse story
@ianmcook as for the question of demand, the first round just merged in included all the backends that we have received so far so everything on top of that is just icing for now. That being said we uncovered so many portability issues in our current model through the expansion so it can be good to expand into some weird dialects just for the sake of it
Reacted by Ian CookAlso chDB to complete the ClickHouse story
chDB is available as an ADBC driver: https://clickhouse.com/docs/chdb/install/adbc
Yup - that was the plan. It'll join DataFusion as an ADBC-backed, in-process engine
Reacted by Ian Cook
Cross-referencing the current 17 dialects against the DB-Engines ranking (438 systems, Sept 2026) shows the remaining gaps fall into two cost classes. All relational systems in the top 10 are already covered (Oracle, MySQL, MSSQL, PostgreSQL, Snowflake, Databricks, SQLite).
Level 1 — Wire-protocol aliases (no new dialect module)
These systems speak an existing wire protocol, so support is a scheme mapping in
detect_dialect/dialect_for_schemeplus golden tests — no new dialect module, no new driver machinery. Several outrank systems we already support.Suggested acceptance criteria: scheme aliases resolve to the right dialect, golden SQL tests pass for each alias, and each gets at least a smoke-test row in the Tier 2/3 live matrix where a container exists.
Level 2 — New dialect modules
Ranked by popularity. None have ADBC drivers, but all ship vendor ODBC drivers, so each rides the existing ODBC fallback: dialect module + golden tests + live config, no new driver infrastructure.
FETCH FIRST n ROWS ONLYlimit syntax, ANSI double-quote identsSAMPLE,TOPvariant)Db2 is the obvious first pick. A reasonable sequencing: Db2 alone first (validates the ODBC-fallback dialect workflow end-to-end), then the Hive/Impala pair (shared dialect groundwork), then HANA/Teradata/Vertica as demand warrants.