[HIVE-25343] Create or replace view should clean the old table properties

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.1.3, 3.2.0
• Component/s: SQL
• Labels: pull-request-available
• Assignee: Lantao Jin
• Reporter: Lantao Jin
• Created: 19/Jul/21 07:41
• Updated: 30/Sep/21 00:09

Description

In many cases, users use Spark and Hive together. When a user creates a view via Spark, the table output columns will store in table properties, such as

After that, if the user runs the command "create or replace view" via Hive, to change the schema. The old table properties added by Spark are not cleaned by Hive. Then users read the table via Spark. The schema didn't change. It very confused users.

How to reproduce:

spark-sql>create table lajin_table (a int, b int) stored as parquet;
spark-sql>create view lajin_view as select * from lajin_table;
spark-sql> desc lajin_view;
a       int     NULL    NULL
b       int     NULL    NULL

hive>desc lajin_view;
a                       int                                         
b                       int
hive>create or replace view lajin_view as select a, b, 3 as c from lajin_table;
hive>desc lajin_view;
a                       int                                         
b                       int                                         
c                       int

spark-sql> desc lajin_view; -- not changed
a       int     NULL    NULL
b       int     NULL    NULL

Issue Links

• links to: GitHub Pull Request #2492