MlflowArtifactDataSet does not work with PartitionedDataSet

[Galileo-Galilei]

Description

It is not possible to store a PartitionedDataSet as an mlflow artifact with the MlflowArtifactDataSet.

Context

I had a use case where I need to save a dict with many small result tables to mlflow, and I tried to use PartitionedDataSet for this.

Steps to Reproduce

# catalog.yml

my_dataset:
    type: kedro_mlflow.io.artifacts.MlflowArtifactDataSet
    data_set:
        type: PartitionedDataSet  # or any valid kedro DataSet
        path: /path/to/a/local/folder # the attribute is "path", and not "filepath"!
        dataset: "pandas.CSVDataSet"

then save a dict using this dataset:

catalog.save("my_dataset", dict("a": pd.DataFrame(data=[1,2,3], columns=["a"], "b": pd.DataFrame(data=[1,2,3], columns=["b"])

Expected Result

The 2 Dataframes should be logged as artifacts in the current mlflow run.

Actual Result

An error dataset has not attribute "_filepath" is raised.

Does the bug also happen with the last version on master?

Yes

Potential solution

The error comes from this line:

kedro-mlflow/kedro_mlflow/io/artifacts/mlflow_artifact_dataset.py

Line 53 in 904207a

else self._filepath

maybe we can add a better condition here to default to "path" if there is no "filepath" attribute.