Accept H2OFrame as input to H2OFrame itself

[bkowshik]

Came across this by accident where I assumed a DataFrame was a Pandas DataFrame and I wanted to convert to a H2O DataFrame. But, since it was already a H2O DataFrame it failed instead.

H2O 3.44.0.1 on Kaggle notebooks

Actual behavior

---------------------------------------------------------------------------
H2OTypeError                              Traceback (most recent call last)
Cell In[9], line 2
      1 train_df = h2o.import_file('/kaggle/input/playground-series-s3e24/train.csv')
----> 2 h2o.H2OFrame(train_df)

File /opt/conda/lib/python3.10/site-packages/h2o/frame.py:97, in H2OFrame.__init__(self, python_obj, destination_frame, header, separator, column_names, column_types, na_strings, skipped_columns, force_col_types)
     92 def __init__(self, python_obj=None, destination_frame=None, header=0, separator=",",
     93              column_names=None, column_types=None, na_strings=None, skipped_columns=None, force_col_types=False):
     95     coltype = U(None, "unknown", "uuid", "string", "float", "real", "double", "int", "long", "numeric",
     96                 "categorical", "factor", "enum", "time")
---> 97     assert_is_type(python_obj, None, list, tuple, dict, numpy_ndarray, pandas_dataframe, scipy_sparse)
     98     assert_is_type(destination_frame, None, str)
     99     assert_is_type(header, -1, 0, 1)

File /opt/conda/lib/python3.10/site-packages/h2o/utils/typechecks.py:444, in assert_is_type(var, *types, **kwargs)
    442 etn = _get_type_name(expected_type, dump=", ".join(args[1:]))
    443 vtn = _get_type_name(type(var))
--> 444 raise H2OTypeError(var_name=vname, var_value=var, var_type_name=vtn, exp_type_name=etn, message=message,
    445                    skip_frames=skip_frames)

H2OTypeError: Argument `python_obj` should be a None | list | tuple | dict | numpy.ndarray | pandas.DataFrame | scipy.sparse.issparse, got H2OFrame

Expected behavior

Converting a Pandas DataFrame, to a H2O DataFrame.

train_df = pd.read_csv('/kaggle/input/playground-series-s3e24/train.csv')
h2o.H2OFrame(train_df)

The following code snippet should work too without throwing an error.

train_df = h2o.import_file('/kaggle/input/playground-series-s3e24/train.csv')
h2o.H2OFrame(train_df)

[sebhrusen]

precising expected behaviour: the wrapped frame should share same data with the original one, but have a different id, this means that they are represented by different objects on the backend—we explicitely created a new frame using the Py client after all—, even if they may share the data content (no deep cloning).

fr_ori = h2o.import_file('h2o://iris')
fr_new = h2o.H2OFrame(fr_ori)

assert id(fr_ori) != id(fr_new)
assert fr_ori.key != fr_new.key
pd.testing.assert_frame_equal(fr_ori.as_data_frame(),  fr_new.as_data_frame())

[sebhrusen]

will be in the next 3.44.0.3 release