[SPARK-41299] OOM when filter pushdown `last_day` function

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.3.1
• Component/s: Optimizer

Environment

Spark 3.3.1

JDK 8 (openjdk version "1.8.0_352")

• Assignee: Unassigned
• Reporter: André F.
• Created: 28/Nov/22 12:42
• Updated: 12/Dec/22 18:11

Description

Using the following transformation on Spark 3.3.1:

df.where($"date" === last_day($"date")) 

Where `df` is a dataframe created from a set Parquet files.  I'm trying to filter dates where they match with the last day of the month of where `date` happened.

Executors are dying with the following error:

java.lang.OutOfMemoryError: GC overhead limit exceeded
at java.util.regex.Pattern.compile(Pattern.java:1722) ~[?:1.8.0_252]
at java.util.regex.Pattern.<init>(Pattern.java:1352) ~[?:1.8.0_252]
at java.util.regex.Pattern.compile(Pattern.java:1028) ~[?:1.8.0_252] 

By disabling the predicate pushdown rule, the job works normally.

 Also, this works normally on Spark 3.3.0. I also couldn't verify other date functions failing on the same way.