[IMPALA-7281] Mixed-format Parquet+Avro tables incorrectly override column comments

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: Impala 3.0
• Component/s: Catalog, Frontend
• Epic Link: avro-metadata-issues
• Epic Color: ghx-label-3
• Assignee: Unassigned
• Reporter: Todd Lipcon
• Created: 11/Jul/18 22:39
• Updated: 11/Mar/24 11:52

Description

Following the following steps:

• create a partitioned table with parquet format
• add a partition
• set the new partition to avro format
• 'show create table' shows no comments on the columns
• refresh the table
• 'show create table' now shows "from deserializer" on all of the columns

This seems to be because, if there is no explicit avro schema, we infer one from the defined FieldSchemas. Then, we replace the FieldSchemas with new objects inferred from the Avro schema.