[HIVE-7653] Hive AvroSerDe does not support circular references in Schema

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.13.1
• Assignee: Sachin Goyal
• Reporter: Sachin Goyal
• Created: 07/Aug/14 21:48
• Updated: 06/Jul/15 16:34
• Resolved: 19/Feb/15 08:56

Description

Avro allows nullable circular references but Hive AvroSerDe does not.

Example of circular references (passing in Avro but failing in AvroSerDe):

class AvroCycleParent {
  AvroCycleChild child;
  public AvroCycleChild getChild () {return child;}
  public void setChild (AvroCycleChild child) {this.child = child;}
}

class AvroCycleChild {
  AvroCycleParent parent;
  public AvroCycleParent getParent () {return parent;}
  public void setParent (AvroCycleParent parent) {this.parent = parent;}
}

Due to this discrepancy, Hive is unable to read Avro records having circular-references. For some third-party code with such references, it becomes very hard to directly serialize it with Avro and use in Hive.

I have a patch for this with a unit-test and I will submit it shortly.

Issue Links

• relates to: HIVE-10187 Avro backed tables don't handle cyclical or recursive records
• links to: https://reviews.apache.org/r/24609/