[HUDI-631] HoodieAvroUtils.rewrite does not handle schema change such as optional fields removal

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Common Core
• Assignee: Unassigned
• Reporter: Yixue Zhu
• Created: 23/Feb/20 18:40
• Updated: 04/Nov/21 03:26

Description

The usage of utility function https://github.com/apache/hudi/blob/6a0aa9a645d11ed7b50e18aa0563dafcd9d145f7/hudi-common/src/main/java/org/apache/hudi/avro/HoodieAvroUtils.java#L258 does not handle schema evolution such as optional fields removal (from new schema).

There are example usage cases:
1. A typical schema evolution like changing a field type would require one to add a new field with new type, then once old data is no longer used, or new field is backfilled to old records, old field can be dropped.
2. A field which is no longer needed.

The schema evolution check, which was implemented in HUDI-741, would hinder this kind of usage.