[AVRO-3789] Comparing maps in GenericData is wrong for certain combinations and fails for empty maps

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.11.2
• Fix Version/s: 1.11.3
• Component/s: java
• Labels: pull-request-available
• Assignee: Felix Krull
• Reporter: Felix Krull
• Created: 05/Jul/23 14:44
• Updated: 12/Jul/23 16:21
• Resolved: 12/Jul/23 11:26

Description

The rewritten map comparison implementation in GenericData (AVRO-2943 according to the commit message) throws a NoSuchElementException when used to compare empty maps. Partial stacktrace:

Caused by: java.util.NoSuchElementException
	at java.base/java.util.HashMap$HashIterator.nextNode(HashMap.java:1513)
	at java.base/java.util.HashMap$KeyIterator.next(HashMap.java:1534)
	at org.apache.avro.generic.GenericData.compareMaps(GenericData.java:1163)
	at org.apache.avro.generic.GenericData.compare(GenericData.java:1250)
	at org.apache.avro.specific.SpecificData.compare(SpecificData.java:476)
	at org.apache.avro.generic.GenericData.compare(GenericData.java:1229)
	at org.apache.avro.specific.SpecificData.compare(SpecificData.java:476)
	at org.apache.avro.specific.SpecificRecordBase.equals(SpecificRecordBase.java:88)
	at scala.runtime.BoxesRunTime.equals2(BoxesRunTime.java:133)
	at scala.runtime.BoxesRunTime.equals(BoxesRunTime.java:119)
	at org.scalactic.DefaultEquality$.areEqualComparingArraysStructurally(DefaultEquality.scala:70)
	at org.scalactic.DefaultEquality.areEqual(DefaultEquality.scala:37)
	at org.mockito.package$$anon$2.areEqual(mockito.scala:614)
	at org.scalactic.TripleEqualsSupport$Equalizer.$eq$eq$eq(TripleEqualsSupport.scala:117)
	at org.mockito.matchers.EqTo.matches(EqTo.scala:11)
... 

 

---

Also, the check in line 1170 that's intended to shortcircuit for maps of different sizes is incorrect. Because of this, maps will incorrectly compare equal when:

• their sizes are different
• one is a superset of the other map

Issue Links

• is duplicated by: AVRO-3792 GenericData compare fails for Maps size = 0 or sizes not same
• links to: GitHub Pull Request #2318