[SPARK-38073] Update atexit function to avoid issues with late binding

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.2.0, 3.3.0
• Fix Version/s: 3.3.0, 3.2.2
• Component/s: PySpark, Spark Shell
• Assignee: Maciej Szymkiewicz
• Reporter: Maciej Szymkiewicz
• Created: 30/Jan/22 15:25
• Updated: 12/Dec/22 18:10
• Resolved: 05/Feb/22 04:21

Description

When PYSPARK_DRIVER_PYTHON=$(which ipython) bin/pyspark is executed with Python >= 3.8, function registered wiht atexit seems to be executed in different scope than in Python 3.7.

It result in NameError: name 'sc' is not defined on exit:

Welcome to
      ____              __
     / __/__  ___ _____/ /__
    _\ \/ _ \/ _ `/ __/  '_/
   /__ / .__/\_,_/_/ /_/\_\   version 3.3.0-SNAPSHOT
      /_/

Using Python version 3.8.12 (default, Oct 12 2021 21:57:06)
Spark context Web UI available at http://192.168.0.198:4040
Spark context available as 'sc' (master = local[*], app id = local-1643555855409).
SparkSession available as 'spark'.

In [1]:                                                                                                                                                                                 
Do you really want to exit ([y]/n)? y
Error in atexit._run_exitfuncs:
Traceback (most recent call last):
  File "/path/to/spark/python/pyspark/shell.py", line 49, in <lambda>
    atexit.register(lambda: sc.stop())
NameError: name 'sc' is not defined

This could be easily fixed by capturing `sc` instance

diff --git a/python/pyspark/shell.py b/python/pyspark/shell.py
index f0c487877a..4164e3ab0c 100644
--- a/python/pyspark/shell.py
+++ b/python/pyspark/shell.py
@@ -46,7 +46,7 @@ except Exception:
 
 sc = spark.sparkContext
 sql = spark.sql
-atexit.register(lambda: sc.stop())
+atexit.register((lambda sc: lambda: sc.stop())(sc))
 
 # for compatibility
 sqlContext = spark._wrapped

Issue Links

• links to: [Github] Pull Request #35396 (zero323)