[SPARK-40307] Introduce Arrow Python UDFs

• Type: Umbrella
• Status: Resolved
• Priority: Major
• Resolution: Done
• Affects Version/s: 3.5.0
• Fix Version/s: 3.5.0
• Component/s: Connect, PySpark
• Assignee: Xinrong Meng
• Reporter: Xinrong Meng
• Created: 02/Sep/22 00:32
• Updated: 25/Aug/23 05:48
• Resolved: 25/Aug/23 05:48

Description

Python user-defined function (UDF) enables users to run arbitrary code against PySpark columns. It uses Pickle for (de)serialization and executes row by row.

One major performance bottleneck of Python UDFs is (de)serialization, that is, the data interchanging between the worker JVM and the spawned Python subprocess which actually executes the UDF. We should seek an alternative to handle the (de)serialization: Arrow, which is used in the (de)serialization of Pandas UDF already.

There should be two ways to enable/disable the Arrow optimization for Python UDFs:

• the Spark configuration `spark.sql.execution.pythonUDF.arrow.enabled`, disabled by default.
• the `useArrow` parameter of the `udf` function, None by default.

The Spark configuration takes effect only when `useArrow` is None. Otherwise, `useArrow` decides whether a specific user-defined function is optimized by Arrow or not.

The reason why we introduce these two ways is to provide both a convenient, per-Spark-session control and a finer-grained, per-UDF control of the Arrow optimization for Python UDFs.

Issue Links

• relates to: SPARK-43543 Standardize Nested Complex DataTypes Support
• links to: [Github] Pull Request #39384 (xinrong-meng)

Sub-Tasks

• Block Arrow Python UDFs: Resolved
• Arrow Python UDFs in Spark Connect: Resolved
• Introduce `SQL_ARROW_BATCHED_UDF` EvalType for Arrow Python UDFs: Resolved
• Support registration of an Arrow Python UDF: Resolved
• Non-atomic data type support in Arrow Python UDF: Resolved
• Improve ArrayType input support in Arrow Python UDF: Resolved
• Explicit Arrow casting for mismatched return type in Arrow Python UDF: Resolved
• Import SparkSession in Python UDF only when useArrow is None: Resolved
• Arrow Python UDF Use Guide: Resolved
• Improve tests and documentation for Arrow Python UDF: Resolved
• Arrow python UDFS didn't support UDT as output type: Open