Pandas dataframes with array column values are not correctly persisted as AzureML datasets

[jarandaf]

Pandas dataframes with arrays as column values seem to be incorrectly persisted. An example:

test_df = pd.DataFrame({'x': [np.random.rand(1000) for _ in range(1000)]})
ds = Datastore.get_default(ws)
Dataset.Tabular.register_pandas_dataframe(test_df, ds, 'test_dataset')

test_df.head()
###
	x
0	[0.5044850335733219, 0.6054305053424696, 0.669...
1	[0.41759815476145723, 0.266477750018155, 0.511...
2	[0.6777708610872593, 0.16925324567267985, 0.16...
3	[0.4268294269387616, 0.6540643485117185, 0.033...
4	[0.6560106490417036, 0.5804652379458484, 0.582...

Dataset.get_by_name(ws, 'test_dataset').to_pandas_dataframe().head()
###
x
0	ERROR
1	ERROR
2	ERROR
3	ERROR
4	ERROR