[SPARK-44042] SPIP: PySpark Test Framework

• Type: New Feature
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.5.0
• Component/s: PySpark
• Assignee: Unassigned
• Reporter: Amanda Liu
• Shepherd: Holden Karau
• Created: 13/Jun/23 14:26
• Updated: 21/Jun/23 14:18

Description

Currently, there's no official PySpark test framework, but only various open-source repos and blog posts. Many of these open-source resources are very popular, which demonstrates user-demand for PySpark testing capabilities. spark-testing-base has 1.4k stars, and chispa has 532k downloads/month. However, it can be confusing for users to piece together disparate resources to write their own PySpark tests (see The Elephant in the Room: How to Write PySpark Tests). We can streamline and simplify the testing process by incorporating test features, such as a PySpark Test Base class (which allows tests to share Spark sessions) and test util functions (for example, asserting dataframe and schema equality). Please see the full SPIP document attached: https://docs.google.com/document/d/1OkyBn3JbEHkkQgSQ45Lq82esXjr9rm2Vj7Ih_4zycRc/edit#heading=h.f5f0u2riv07v.

Sub-Tasks

• Add assertDataFrameEqual util function: Resolved
• Display percent of unequal rows in DataFrame comparison: Resolved
• Add pyspark_testing module for GHA tests: Resolved
• Expose assertDataFrameEqual in pyspark.testing.utils: Resolved
• Make assertSchemaEqual API public: Resolved
• Allow custom precision for fp approx equality: Resolved
• Support List[Row] data type for expected DataFrame argument: Resolved
• Add checks for expected list type special cases: Resolved
• Use difflib to display errors in assertDataFrameEqual: Resolved
• Clarify error for unsupported arg data type in assertDataFrameEqual: Resolved
• Add support for pandas-on-Spark DataFrame assertDataFrameEqual: Resolved
• Fix pandas-on-Spark type checks for assertDataFrameEqual: Resolved
• Customize diff log in assertDataFrameEqual error message format: Resolved
• Update assertDataFrameEqual docs error example output: Resolved
• Add PySparkTestBase unit test class: Open
• Add pyspark.testing to setup.py: Resolved
• Support comparison between lists of Rows: Resolved
• Publish PySpark Test Guidelines webpage: Resolved
• Raise error when only one df is None: Resolved
• Add support for pandas DataFrame assertDataFrameEqual: Resolved
• Make pandas error class message_parameters strings: Resolved
• Fix PySpark testing guide links: Resolved