def _dataframe_from_json(path_or_str,
schema: Schema = None,
pandas_orient: str = "split") -> pd.DataFrame:
"""
Parse json into pandas.DataFrame. User can pass schema to ensure correct type parsing and to
make any necessary conversions (e.g. string -> binary for binary columns).
:param path_or_str: Path to a json file or a json string.
:param schema: Mlflow schema used when parsing the data.
:param pandas_orient: pandas data frame convention used to store the data.
:return: pandas.DataFrame.
"""
if schema is not None:
dtypes = dict(zip(schema.column_names(), schema.column_types()))
df = pd.read_json(path_or_str, orient=pandas_orient,
dtype=dtypes)[schema.column_names()]
binary_cols = [
i for i, x in enumerate(schema.column_types())
if x == DataType.binary
]
for i in binary_cols:
col = df.columns[i]
df[col] = np.array(df[col].map(_base64decode), dtype=np.bytes_)
return df
else:
return pd.read_json(path_or_str, orient=pandas_orient, dtype=False)
def _dataframe_from_json(path_or_str,
schema: Schema = None,
pandas_orient: str = "split",
precise_float=False) -> pd.DataFrame:
"""
Parse json into pandas.DataFrame. User can pass schema to ensure correct type parsing and to
make any necessary conversions (e.g. string -> binary for binary columns).
:param path_or_str: Path to a json file or a json string.
:param schema: Mlflow schema used when parsing the data.
:param pandas_orient: pandas data frame convention used to store the data.
:return: pandas.DataFrame.
"""
if schema is not None:
dtypes = dict(zip(schema.column_names(), schema.pandas_types()))
df = pd.read_json(path_or_str,
orient=pandas_orient,
dtype=dtypes,
precise_float=precise_float)
actual_cols = set(df.columns)
for type_, name in zip(schema.column_types(), schema.column_names()):
if type_ == DataType.binary and name in actual_cols:
df[name] = df[name].map(
lambda x: base64.decodebytes(bytes(x, "utf8")))
return df
else:
return pd.read_json(path_or_str,
orient=pandas_orient,
dtype=False,
precise_float=precise_float)
