Can you query Delta Lake with SQL without Spark?

Yes. A SQL engine that understands the Delta transaction log can resolve the current table version and expose it through SQL without running Spark.

Should I use ODBC or ADBC?

Use ODBC for established BI tools such as Power BI, Tableau, and Excel. Use ADBC for Arrow-native applications that benefit from columnar result transfer.

Why not query the Parquet files directly?

Direct Parquet reads bypass the Delta transaction log and can include files that are no longer part of the current table version.

Query Delta Lake with SQL, ODBC and ADBC

The architecture

A SQL engine reads _delta_log and identifies the active Parquet files.
The engine applies filters, joins, and aggregations close to the data.
The client receives a normal SQL result through ODBC or ADBC.

DeltaForge is the worked example. It is a commercial, self-hosted engine that reads and writes Delta Lake and Apache Iceberg without Spark or a JVM.

ODBC for BI tools

ODBC remains the practical default for Power BI, Tableau, Excel, and many reporting platforms. Microsoft documents both the standard Power Query ODBC connection flow and the requirements for DirectQuery-capable ODBC connectors.

Use Import mode for a cached model. Use DirectQuery when the report should query the current source at interaction time. The Power BI-specific setup is covered in Power BI to Delta Lake without Databricks.

ADBC for Arrow-native clients

Apache Arrow ADBC defines a database API built around Arrow data. Instead of converting query results into a row-oriented interchange format, an ADBC driver can return Arrow record batches directly.

That makes ADBC a good fit for Python, notebooks, and applications that already use Arrow internally. It complements ODBC rather than replacing it: choose the driver your client ecosystem understands best.

A normal SQL surface

SELECT region, product_category, SUM(net_revenue) AS revenue
FROM lakehouse.sales_fact
WHERE order_date >= DATE '2026-01-01'
GROUP BY region, product_category
ORDER BY revenue DESC;

The engine resolves the current Delta snapshot before executing the SQL.

The same endpoint can expose time travel and tables maintained with UPDATE, DELETE, and MERGE without Spark.

FAQ

Can this work without Databricks?

Yes. The requirement is a Delta-aware query engine, not a specific managed platform.

Does the data leave my environment?

Not with a self-hosted engine. DeltaForge runs in your cloud, on-premises, or air-gapped network and queries the tables in place.

Will other engines still read the tables?

Yes. DeltaForge reads and writes standard Delta Lake tables, so compatible engines can continue using the same storage.

Run it yourself

Install DeltaForge, register a Delta table, and connect through the driver your client supports. Start with the install guide, ODBC driver, or ADBC driver.