Configuration¶

All configuration surfaces for the plugin.

Settings¶

Setting	Type	Default	Description
`--adbc-record`	CLI flag	`none`	Record mode for this run. Overrides `adbc_record_mode`. Choices: `none`, `once`, `new_episodes`, `all`.
`adbc_cassette_dir`	ini key (str)	`tests/cassettes`	Directory where cassette subdirectories are stored. Relative to the pytest rootdir.
`adbc_record_mode`	ini key (str)	`none`	Default record mode. Overridden by `--adbc-record` for a single run.
`adbc_dialect`	ini key (linelist)	`[]`	SQL dialect for normalisation passed to sqlglot. Bare value = global fallback. Per-driver form: `driver_name: dialect`. Empty = auto-detect.
`adbc_auto_patch`	ini key (linelist)	`[]`	List of ADBC driver module names whose `connect()` function is intercepted automatically. Only active for tests with `@pytest.mark.adbc_cassette`.
`adbc_scrub_keys`	ini key (linelist)	`[]`	Parameter key names to redact from `.json` cassette files. Global form: space-separated keys. Per-driver form: `driver_name: key1 key2`.

pyproject.toml¶

[tool.pytest.ini_options]
adbc_cassette_dir = "tests/cassettes"
adbc_record_mode = "none"
adbc_dialect = []  # e.g. ["adbc_driver_snowflake.dbapi: snowflake", "adbc_driver_duckdb.dbapi: duckdb"]
adbc_auto_patch = []  # e.g. ["adbc_driver_duckdb.dbapi", "adbc_driver_snowflake.dbapi"]
adbc_scrub_keys = []  # e.g. ["token password", "adbc_driver_snowflake: account_id"]

pytest.ini¶

[pytest]
adbc_cassette_dir = tests/cassettes
adbc_record_mode = none
adbc_dialect =
    # adbc_driver_snowflake.dbapi: snowflake
    # adbc_driver_duckdb.dbapi: duckdb
adbc_auto_patch =
adbc_scrub_keys =

`adbc_dialect`¶

adbc_dialect is a linelist ini key. Each line is either a bare value (global fallback) or a per-driver line with the driver module name followed by a colon:

[tool.pytest.ini_options]
adbc_dialect = [
    # Global fallback for any driver not explicitly listed
    "snowflake",
    # Per-driver override
    "adbc_driver_duckdb.dbapi: duckdb",
]

In pytest.ini:

adbc_dialect =
    snowflake
    adbc_driver_duckdb.dbapi: duckdb

Dialect resolution priority chain:

Explicit dialect= argument passed to adbc_replay.wrap() or wrap_from_item()
dialect= argument on @pytest.mark.adbc_cassette for the current test
Per-driver ini entry matching the driver module name
Global ini fallback (bare value)
Auto-detect (no dialect configured)
A bare value sets the global fallback. When multiple bare lines are present, the last one wins.
Per-driver lines use the form driver_module_name: dialect_string. The driver module name must match what is passed to adbc_replay.wrap() (e.g. adbc_driver_snowflake.dbapi).
Accepted dialect strings are any value sqlglot recognises ("snowflake", "bigquery", "duckdb", etc.).
An empty list (the default) triggers sqlglot auto-detect, which works for standard SQL.

For most projects, per-driver ini config is the right path. The dialect= marker argument is an edge-case override for individual tests that need a different dialect than their driver's configured value. See Setting the SQL dialect for usage examples.

`adbc_scrub_keys`¶

adbc_scrub_keys is a linelist ini key. Each line is either a space-separated list of global key names, or a per-driver line with the driver module name followed by a colon:

[tool.pytest.ini_options]
adbc_scrub_keys = [
    # Global: redact these keys for all drivers
    "token password api_key",
    # Per-driver: only for adbc_driver_snowflake
    "adbc_driver_snowflake: account_id warehouse",
]

Global keys and per-driver keys are combined (unioned) when scrubbing for a given driver.
Only dictionary params are affected. List params have no key names and are written unchanged.
Keys not present in the params dict are silently ignored.
Matched values are replaced with the sentinel string REDACTED.

Multiple global lines are merged into a single flat list. Multiple per-driver lines for the same driver are accumulated. See Scrub sensitive values for a how-to guide.

Notes¶

The --adbc-record CLI flag takes precedence over adbc_record_mode for the duration of that pytest run. Neither modifies the stored configuration.

Precedence¶

--adbc-record (CLI flag) > adbc_record_mode (ini key)

When both are set, the CLI flag wins for that session only.

Automatic patching¶

adbc_auto_patch is a linelist ini key. Each line is a Python module name that exposes connect() at its top level — for ADBC drivers this is always adbc_driver_<name>.dbapi.

pyproject.tomlpytest.ini

[tool.pytest.ini_options]
adbc_auto_patch = [
    "adbc_driver_duckdb.dbapi",
    "adbc_driver_snowflake.dbapi",
]

[pytest]
adbc_auto_patch =
    adbc_driver_duckdb.dbapi
    adbc_driver_snowflake.dbapi

See Finding the right module path.

Key behaviours:

Only drivers in this list have their connect() intercepted.
Only tests with @pytest.mark.adbc_cassette are intercepted — unmarked tests receive the real connect() unchanged.
If a listed driver module is not installed, the plugin skips it silently. This means replay-only CI environments do not need the driver installed.
Session-scoped and module-scoped connections cannot use auto-patching (the monkeypatch fires once per test). Use the adbc_connect fixture or adbc_replay.wrap() for those cases.

Rootdir¶

adbc_cassette_dir is resolved relative to the pytest rootdir, not the current working directory. Pytest determines rootdir from the location of pyproject.toml, pytest.ini, setup.py, or setup.cfg.

Record Modes — behaviour of each mode value
Markers — per-test dialect override via @pytest.mark.adbc_cassette