Spark SQL

• Apache Arrow in PySpark
  • Ensure PyArrow Installed
  • Conversion to/from Arrow Table
  • Enabling for Conversion to/from Pandas
  • Pandas UDFs (a.k.a. Vectorized UDFs)
  • Pandas Function APIs
  • Arrow Python UDFs
  • Usage Notes
• Vectorized Python User-defined Table Functions (UDTFs)
  • Vectorized Python UDTF Interface
  • Defining the Output Schema
  • Emitting Output Rows
  • Usage Examples
  • TABLE Argument
  • PARTITION BY and ORDER BY
  • Best Practices
  • More Examples
• Python User-defined Table Functions (UDTFs)
  • Implementing a Python UDTF
  • Defining the Output Schema
  • Emitting Output Rows
  • Registering and Using Python UDTFs in SQL
  • Arrow Optimization
  • UDTF Examples with Scalar Arguments
  • Accepting an Input Table Argument
• Python Data Source API
  • Overview
  • Simple Example: Data Source with Batch Reader
  • Comprehensive Example: Data Source with Batch and Streaming Readers and Writers
  • Serialization Requirement
  • Using a Python Data Source
  • Python Data Source Reader with direct Arrow Batch support for improved performance
  • Usage Notes
• Python to Spark Type Conversions
  • Browsing Type Conversions
  • Configuration
  • All Conversions
  • Conversions in Practice - UDFs
  • Conversions in Practice - Creating DataFrames
  • Conversions in Practice - Nested Data Types