commit_dataset does not work with nested partitions

[JoergRittinger]

The following code example

from kartothek.io.eager import commit_dataset, write_single_partition, create_empty_dataset_header
from kartothek.core.common_metadata import make_meta
import pandas as pd
import storefact


store = storefact.get_store_from_url("hfs:///tmp/some_directory")
for x in store.keys():
    store.delete(x)

df = pd.DataFrame({"a": [1, 1, 2, 2], "b": [3, 4, 5, 6]})

create_empty_dataset_header(
    store=store,
    dataset_uuid="uuid",
    table_meta={"table": make_meta(df, "table", ["a"])},
    partition_on=["a"],
)

partitions = []
for x in range(2):
    partitions.append(
        write_single_partition(
            store=store,
            dataset_uuid="uuid",
            data={"data": {"table": df}},
            partition_on=["a"],
        )
    )

commit_dataset(
    store=store,
    dataset_uuid="uuid",
    new_partitions=partitions,
    partition_on=["a"],
)

Will create the error message AttributeError: Accessing label attribute is not allowed while nested.

Additionally you can observe the behaviour that after commit_dataset the partitions list contains one less element. In my understanding commit_dataset should not manipulate input objects.