`DirectOpExecutionContext` uses wrong attribute when using `context.asset_partition_key_range_for_output()`

[mgierada]

Dagster version

1.7.4

What's the issue?

I have the following unit test to test the daily partition asset

def test_asset_to_be_tested(
    mock1 pd.DataFrame,
    mock2 pd.DataFrame,
    expected_outlier_treated_instrument_returns: pd.DataFrame,
) -> None:
    from some_location import asset_to_be_tested

    partition_key = "2022-04-19"

    context = build_op_context(
        partition_key=partition_key,
        op_config={
            "sample_size": 5,
            "percentile": 0.02,
        },
    )

    results = asset_to_be_tested(
        context, mock1, mock2
    )

    pd.testing.assert_frame_equal(
        results.reset_index(drop=True),
        expected.reset_index(drop=True),
    )

When I run it, I got the following error

AttributeError: 'DirectOpExecutionContext' object has no attribute '_step_execution_context'. Did you mean: 'get_step_execution_context'?

What did you expect to happen?

No exceptions are raised.

How to reproduce?

Use the pseudo-code of the unit test I added.

@asset(
    partitions_def=DailyPartitionsDefinition("some_starting_date"),
    metadata={"partition_expr": "obs_date"},
    group_name="market",
    io_manager_key="my_manager",
    dagster_type=pandera_schema_to_dagster_type(SomePandera)
    ins={
        "instrument_prices": AssetIn(
            "whatever",
        ),
        "security_master": AssetIn(
            "whatever2",
        ),
    },
    config_schema={
         "sample_size": Field(
            int,
            is_required=False,
            default_value=10000,
        ),
        "percentile": Field(
            float,
            is_required=False,
            default_value=0.02,
        ),
    },
    key_prefix="indicator",
    code_version="20240430",
    auto_materialize_policy=policy...
)
def asset_to_be_tested(
    context: AssetExecutionContext,
    df1: pd.DataFrame,
    df2: pd.DataFrame,
) -> pd.DataFrame:

    start_date_str = context.asset_partition_key_range_for_output().start
    (some pandas transformation)...
    return pandasdataframe

Deployment type

Other Docker-based deployment

Deployment details

No response

Additional information

No response

Message from the maintainers

Impacted by this issue? Give it a 👍! We factor engagement into prioritization.

[OwenKephart]

Hi @mgierada, thanks for filing. Looks like the underlying issue is that the DirectOpExecutionContext does not override the implementation for context.asset_partition_key_range_for_output(). This bug should get fixed, but until then accessing the property context.partition_key_range should work as an alternative, as that is properly overridden.

[mgierada]

Thanks @OwenKephart I confirm that context.partition_key_range does work as an alternative. Thanks for digging in!