Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[SPARK-43440][PYTHON][CONNECT] Support registration of an Arrow-optimized Python UDF #41125

Closed
wants to merge 5 commits into from
Closed

[SPARK-43440][PYTHON][CONNECT] Support registration of an Arrow-optimized Python UDF #41125

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4784085
fix
xinrong-meng May 10, 2023
9e1b534
test
xinrong-meng May 10, 2023
4db78a9
restore Connect
xinrong-meng May 11, 2023
d0cd0e5
restore Connect
xinrong-meng May 11, 2023
23e6cba
fix
xinrong-meng May 12, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 8 additions & 7 deletions python/pyspark/sql/connect/udf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -252,31 +252,32 @@ def register(
f = cast("UserDefinedFunctionLike", f)
if f.evalType not in [
PythonEvalType.SQL_BATCHED_UDF,
PythonEvalType.SQL_ARROW_BATCHED_UDF,
PythonEvalType.SQL_SCALAR_PANDAS_UDF,
PythonEvalType.SQL_SCALAR_PANDAS_ITER_UDF,
PythonEvalType.SQL_GROUPED_AGG_PANDAS_UDF,
]:
raise PySparkTypeError(
error_class="INVALID_UDF_EVAL_TYPE",
message_parameters={
"eval_type": "SQL_BATCHED_UDF, SQL_SCALAR_PANDAS_UDF, "
"SQL_SCALAR_PANDAS_ITER_UDF or SQL_GROUPED_AGG_PANDAS_UDF"
"eval_type": "SQL_BATCHED_UDF, SQL_ARROW_BATCHED_UDF, "
"SQL_SCALAR_PANDAS_UDF, SQL_SCALAR_PANDAS_ITER_UDF or "
"SQL_GROUPED_AGG_PANDAS_UDF"
},
)
return_udf = f
self.sparkSession._client.register_udf(
f.func, f.returnType, name, f.evalType, f.deterministic
)
return f
else:
if returnType is None:
returnType = StringType()
return_udf = _create_udf(
py_udf = _create_udf(
f, returnType=returnType, evalType=PythonEvalType.SQL_BATCHED_UDF, name=name
)

self.sparkSession._client.register_udf(f, returnType, name)

return return_udf
self.sparkSession._client.register_udf(py_udf.func, returnType, name)
return py_udf

register.__doc__ = PySparkUDFRegistration.register.__doc__

Expand Down
2 changes: 1 addition & 1 deletion python/pyspark/sql/tests/pandas/test_pandas_grouped_map.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -219,7 +219,7 @@ def test_register_grouped_map_udf(self):
exception=pe.exception,
error_class="INVALID_UDF_EVAL_TYPE",
message_parameters={
"eval_type": "SQL_BATCHED_UDF, SQL_SCALAR_PANDAS_UDF, "
"eval_type": "SQL_BATCHED_UDF, SQL_ARROW_BATCHED_UDF, SQL_SCALAR_PANDAS_UDF, "
"SQL_SCALAR_PANDAS_ITER_UDF or SQL_GROUPED_AGG_PANDAS_UDF"
},
)
Expand Down
18 changes: 18 additions & 0 deletions python/pyspark/sql/tests/test_arrow_python_udf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,6 +119,24 @@ def test_eval_type(self):
udf(lambda x: str(x), useArrow=False).evalType, PythonEvalType.SQL_BATCHED_UDF
)

def test_register(self):
df = self.spark.range(1).selectExpr(
"array(1, 2, 3) as array",
)
str_repr_func = self.spark.udf.register("str_repr", udf(lambda x: str(x), useArrow=True))

# To verify that Arrow optimization is on
self.assertEquals(
df.selectExpr("str_repr(array) AS str_id").first()[0],
"[1 2 3]", # The input is a NumPy array when the Arrow optimization is on
)

# To verify that a UserDefinedFunction is returned
self.assertListEqual(
df.selectExpr("str_repr(array) AS str_id").collect(),
df.select(str_repr_func("array").alias("str_id")).collect(),
)


class PythonUDFArrowTests(PythonUDFArrowTestsMixin, ReusedSQLTestCase):
@classmethod
Expand Down
18 changes: 12 additions & 6 deletions python/pyspark/sql/udf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -623,32 +623,38 @@ def register(
f = cast("UserDefinedFunctionLike", f)
if f.evalType not in [
PythonEvalType.SQL_BATCHED_UDF,
PythonEvalType.SQL_ARROW_BATCHED_UDF,
PythonEvalType.SQL_SCALAR_PANDAS_UDF,
PythonEvalType.SQL_SCALAR_PANDAS_ITER_UDF,
PythonEvalType.SQL_GROUPED_AGG_PANDAS_UDF,
]:
raise PySparkTypeError(
error_class="INVALID_UDF_EVAL_TYPE",
message_parameters={
"eval_type": "SQL_BATCHED_UDF, SQL_SCALAR_PANDAS_UDF, "
"SQL_SCALAR_PANDAS_ITER_UDF or SQL_GROUPED_AGG_PANDAS_UDF"
"eval_type": "SQL_BATCHED_UDF, SQL_ARROW_BATCHED_UDF, "
"SQL_SCALAR_PANDAS_UDF, SQL_SCALAR_PANDAS_ITER_UDF or "
"SQL_GROUPED_AGG_PANDAS_UDF"
},
)
register_udf = _create_udf(
source_udf = _create_udf(
f.func,
returnType=f.returnType,
name=name,
evalType=f.evalType,
deterministic=f.deterministic,
)._unwrapped # type: ignore[attr-defined]
return_udf = f
)
if f.evalType == PythonEvalType.SQL_ARROW_BATCHED_UDF:
register_udf = _create_arrow_py_udf(source_udf)._unwrapped
Copy link
Member Author

@xinrong-meng xinrong-meng May 11, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That's the magic line :).

else:
register_udf = source_udf._unwrapped # type: ignore[attr-defined]
return_udf = register_udf
else:
if returnType is None:
returnType = StringType()
return_udf = _create_udf(
f, returnType=returnType, evalType=PythonEvalType.SQL_BATCHED_UDF, name=name
)
register_udf = return_udf._unwrapped # type: ignore[attr-defined]
register_udf = return_udf._unwrapped
self.sparkSession._jsparkSession.udf().registerPython(name, register_udf._judf)
return return_udf

Expand Down