Configuration Reference¶

This page lists every option the driver understands, grouped by the ADBC level at which it is set (database, connection, or statement). Spark-specific options use the adbc.spark. prefix; standard ADBC options use the adbc. prefix.

Options can be supplied as a map[string]string (Go), through db_kwargs / conn_kwargs (Python), or inline in the sc:// connection string where a matching string field exists.

Database options¶

Set these when creating the database (Go NewDatabase, Python adbc_driver_spark.connect(..., db_kwargs=...)).

Option key	Connection-string field	Meaning
`uri`	(the whole string)	Spark Connect connection string. Required. Defaults to `sc://localhost:15002`.
`adbc.spark.token`	`token`	Bearer token, sent as `Authorization: Bearer`. Implies TLS.
`adbc.spark.tls.enabled`	`use_ssl` (or `DatabaseOptions.TLS_ENABLED`)	Enable TLS for the gRPC channel: `true` or `false`. Inferred `true` when a token is set.
`adbc.spark.user_id`	`user_id`	Spark Connect user id for the session.
`adbc.spark.user_agent`	`user_agent`	User agent advertised to the server. Defaults to `adbc-driver-spark/<version>`.
`adbc.spark.session_id`	`session_id`	Reuse an existing session id (a UUID). A new session is created when omitted.
`adbc.spark.headers.<NAME>`	(option only)	Arbitrary gRPC metadata header `<NAME>`. Repeat the prefix per header.

Connection options¶

The driver does not implement connection-level option setters; calling SetOption on a connection returns ADBC_STATUS_NOT_IMPLEMENTED. To select a catalog or schema, run it as SQL on a statement (USE CATALOG <c>, USE <db>), or pass the catalog and schema arguments to the metadata methods (GetObjects, GetTableSchema).

Autocommit and transactions¶

Spark Connect has no multi-statement transactions, so the driver always operates in autocommit mode. Commit and Rollback report ADBC_STATUS_NOT_IMPLEMENTED. The Python DBAPI connect(..., autocommit=False) is accepted for API symmetry but transactions remain unavailable. See Compatibility.

Statement options¶

The driver defines no statement-level options; calling SetOption on a statement returns ADBC_STATUS_NOT_IMPLEMENTED. Results are streamed one batch at a time with no tunable prefetch (see Architecture).

Standard ADBC option keys¶

The driver honors these standard ADBC database keys (all set at the database level), in addition to the adbc.spark. keys above:

Option key	Applies to	Meaning
`uri`	database	Spark Connect connection string.
`username`	database	Mapped to the Spark Connect user id.
`password`	database	Mapped to the bearer token.

Examples¶

PythonGo

import adbc_driver_spark.dbapi as dbapi

conn = dbapi.connect(
    "sc://spark.example.com:443",
    token="eyJhbGci...",                                  # implies use_ssl=True
    db_kwargs={
        "adbc.spark.user_id": "analyst",
        "adbc.spark.headers.x-request-source": "etl",
    },
)

# Select a catalog or schema with SQL rather than a connection option.
with conn.cursor() as cur:
    cur.execute("USE CATALOG spark_catalog")
    cur.execute("USE default")

db, err := drv.NewDatabase(map[string]string{
    "uri":                                  "sc://spark.example.com:443",
    "adbc.spark.token":                     "eyJhbGci...",
    "adbc.spark.tls.enabled":               "true",
    "adbc.spark.user_id":                   "analyst",
    "adbc.spark.headers.x-request-source":  "etl",
})

In Python, the Spark-specific keys are also available as enum values on adbc_driver_spark.DatabaseOptions, so you can reference DatabaseOptions.TOKEN.value instead of the raw string.