Saves a dataset.
Saves a dataset. If the path is empty (the first time we call metrics on the schema) then we can write.
If there's already parquet files stored in it, then create a temporary directory to compute on, and flush the path to move updated metrics in it
: dataset to be saved
: Path to save the file at
Partition a dataset using dataset columns.
Partition a dataset using dataset columns. To partition the dataset using the ingestion time, use the reserved column names :
: Input dataset
: list of columns to use for partitioning.
The Spark session used to run this job
Just to force any job to implement its entry point using within the "run" method