❓ Is there an extra step required to run Pytest with interpolated variables in the parameters (from

globals.yml

)? The error is

omegaconf.errors.InterpolationResolutionError: Globals key 'dim_product_filepath' not found and no default value provided.

My set up is:

#conf/base/globals.yml
dim_product_filepath: "data/01_raw/dim_product.parquet"

Clearly

dim_product_filepath

is in

globals.yml

#conf/base/catalog.yml
dim_product:
  type: pandas.ParquetDataset
  filepath: ${globals:dim_product_filepath}

#conf/base/parameters.yml
dim_product_filepath: ${globals:dim_product_filepath}

Strangely, running the node works with no problem.

#src/my_project/pipelines/data_processing/pipeline.py
def create_pipeline(**kwargs) -> Pipeline:
    return pipeline(
        [
            node(
                func=load_dim_product,
                inputs=[
                    "params:dim_product_sql",
                    "params:project_id",
                    "params:dim_product_filepath",
                    "params:env",
                ],
                outputs="dim_product",
                name="load_dim_product",
            ),
            ...,
        ]
    )

Hi @Roger Yu, have you read this already? https://docs.kedro.org/en/stable/tutorial/test_a_project.html. It's ususally advised to create the Kedro class directly for testing, i.e.

DataCatalog

OmegaConfigLoader

. These components need to know where is your root directory. when you do

kedro run

we automatically figure it out for you, when you put them in

pytest

, you may need to specific the location otherwise it's looking ifor config in the wrong place.

> https://docs.kedro.org/en/stable/tutorial/test_a_project.html I just read this. 1. There's no mention of

OmegaConfigLoader

2. The example with

DataCatalog

hardcodes the catalog keys/values and parameter keys/values. Is this what's expected in unit testing in Kedro? e.g.

catalog.add_feed_dict(
            {
                "model_input_table" : dummy_data,
                "params:model_options": dummy_parameters["model_options"],
            }
        )

3. There's no example of telling Pytest how to address the

globals.yml

. Is there another page that explains this?

@Roger Yu Sorry for the late reply. I thought the documentation mentions it but it was my mistake. I raised an issue regarding to the default test example. In short, you can create the config loader as follow:

@pytest.fixture
def config_loader():
    return OmegaConfigLoader(conf_source="conf",base_env="base", default_run_env="local")

I cannot find one specific section answering your questions, but there are few concept that are related to this in https://docs.kedro.org/en/stable/configuration/configuration_basics.html#omegaconfigloader 1.

base_env

and

default_run_env

need to be set correctly when you create

OmegaConfigLoader

manually. This isn't needed when you do

kedro run

because the framework will read

settings.py

and apply the default automatically. The reason for this is that we expect people to use

ConfigLoader

outside of framework (decoupled). 2. ^This is related to globals, because the config loader need to know where to locate the config files, the reason why your previous test fail to locate the file is because the

conf_source

and env is set incorrectly. The config loader use default config_pattern （customisable)

Yes, that worked thank you!