David Lee created ARROW-6001:
--------------------------------
Summary: Add from_pydict(), from_pylist() and to_pylist() to
pyarrow.Table + improve pandas.to_dict()
Key: ARROW-6001
URL: https://issues.apache.org/jira/browse/ARROW-6001
Project: Apache Arrow
Issue Type: Improvement
Reporter: David Lee
I noticed that pyarrow.Table.to_pydict() exists, but there is no
pyarrow.Table.from_pydict() doesn't exist. There is a proposed ticket to create
one, but it doesn't take into account potential mismatches between column order
and number of columns.
I'm attached some code I've written which I've been using to handle arrow to
ordered dictionaries and arrow to lists of dictionaries.. I've also included an
example where this can be used to speed up pandas.to_dict() by a factor of 20x.
{code:java}
def from_pylist(pylist, names=None, schema=None, safe=True):
"""
Converts a python list of dictionaries to a pyarrow table
:param pylist: pylist list of dictionaries
:param names: list of column names
:param schema: pyarrow schema
:param safe: True or False
:return: arrow table
"""
arrow_columns = list()
if schema:
for column in schema.names:
arrow_columns.append(pa.array([v[column] if column in v else None
for v in pylist], safe=safe, type=schema.types[schema.get_field_index(column)]))
arrow_table = pa.Table.from_arrays(arrow_columns, schema.names)
else:
for column in names:
arrow_columns.append(pa.array([v[column] if column in v else None
for v in pylist], safe=safe))
arrow_table = pa.Table.from_arrays(arrow_columns, names)
return arrow_table
def to_pylist(arrow_table, index_columns=None):
"""
Converts a pyarrow table to a python list of dictionaries
:param arrow_table: arrow table
:param index_columns: columns to index
:return: python list of dictionaries
"""
pydict = arrow_table.to_pydict()
if index_columns:
columns = arrow_table.schema.names
columns.append("_index")
pylist = [{column: tuple([pydict[index_column][row] for index_column in
index_columns]) if column == '_index' else pydict[column][row] for column in
columns} for row in range(arrow_table.num_rows)]
else:
pylist = [{column: pydict[column][row] for column in
arrow_table.schema.names} for row in range(arrow_table.num_rows)]
return pylist
def from_pydict(pydict, names=None, schema=None, safe=True):
"""
Converts a pyarrow table to a python ordered dictionary
:param pydict: ordered dictionary
:param names: list of column names
:param schema: pyarrow schema
:param safe: True or False
:return: arrow table
"""
arrow_columns = list()
dict_columns = list(pydict.keys())
if schema:
for column in schema.names:
if column in pydict:
arrow_columns.append(pa.array(pydict[column], safe=safe,
type=schema.types[schema.get_field_index(column)]))
else:
arrow_columns.append(pa.array([None] *
len(pydict[dict_columns[0]]), safe=safe,
type=schema.types[schema.get_field_index(column)]))
arrow_table = pa.Table.from_arrays(arrow_columns, schema.names)
else:
if not names:
names = dict_columns
for column in names:
if column in dict_columns:
arrow_columns.append(pa.array(pydict[column], safe=safe))
else:
arrow_columns.append(pa.array([None] *
len(pydict[dict_columns[0]]), safe=safe))
arrow_table = pa.Table.from_arrays(arrow_columns, names)
return arrow_table
def get_indexed_values(arrow_table, index_columns):
"""
returns back a set of unique values for a list of columns.
:param arrow_table: arrow_table
:param index_columns: list of column names
:return: set of tuples
"""
pydict = arrow_table.to_pydict()
index_set = set([tuple([pydict[index_column][row] for index_column in
index_columns]) for row in range(arrow_table.num_rows)])
return index_set
{code}
Here are my benchmarks using pandas to arrow to python vs of pandas.to_dict()
{code:java}
# benchmark panda conversion to python objects.
start_time = time.time()
python_df1 = panda_df1.to_dict(orient='records')
total_time = time.time() - start_time
print("pandas to python - 1 million rows - " + str(total_time))
start_time = time.time()
python_df4 = panda_df4.to_dict(orient='records')
total_time = time.time() - start_time
print("pandas to python - 4 million rows - " + str(total_time))
start_time = time.time()
arrow_df1 = pa.Table.from_pandas(panda_df1)
pydict = arrow_df1.to_pydict()
python_df1 = [{column: pydict[column][row] for column in
arrow_df1.schema.names} for row in range(arrow_df1.num_rows)]
total_time = time.time() - start_time
print("pandas to arrow to python - 1 million rows - " + str(total_time))
start_time = time.time()
arrow_df4 = pa.Table.from_pandas(panda_df1)
pydict = arrow_df4.to_pydict()
python_df4 = [{column: pydict[column][row] for column in
arrow_df1.schema.names} for row in range(arrow_df1.num_rows)]
total_time = time.time() - start_time
print("pandas to arrow to python - 4 million rows - " + str(total_time))
{code}
{code:java}
pandas to python - 1 million rows - 13.03793740272522
pandas to python - 4 million rows - 51.73868107795715
pandas to arrow to python - 1 million rows - 2.2239842414855957
pandas to arrow to python - 4 million rows - 2.6550424098968506{code}
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
