[HIVE-14085] Allow type widening primitive conversion on hive/parquet tables

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.3.0
• Component/s: File Formats
• Assignee: Vihang Karajgaonkar
• Reporter: Vihang Karajgaonkar
• Created: 23/Jun/16 23:08
• Updated: 21/Jul/17 17:55
• Resolved: 27/Jun/16 19:56

Description

There is a JIRA ticket on upstream that brought this usability improvement in Hive to support auto type widening for Parquet tables. See https://issues.apache.org/jira/browse/HIVE-12080
This improvement is very useful for users who have schema evolution on their tables. For example, a Hive table with a "bigint" can read parquet files with "int32" and "int64" types.
The patch only supports widening conversions from int->bigint and float->double. We should support more types to allow users read their changed parquet schema.
Here's a list of widening conversions we should support:

tinyint     ->  smallint,int,bigint,float,double
smallint  -> int,bigint,float,double
int          -> bigint,float,double
bigint     -> float,double
float       ->  double
double   ->  --