[SPARK-23774] `Cast` to CHAR/VARCHAR should truncate the values

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Won't Do
• Affects Version/s: 2.1.2, 2.2.1, 2.3.0
• Component/s: SQL
• Assignee: Unassigned
• Reporter: Dongjoon Hyun
• Created: 22/Mar/18 23:15
• Updated: 05/Sep/18 23:44
• Resolved: 05/Sep/18 23:44

Description

This issue aims to fix the following `CAST` behavior on `CHAR/VARCHAR` types.
Since HiveStringType is used only in parsing, this PR is also about parsing.

Spark

scala> sql("SELECT CAST('123' AS CHAR(1)), CAST('123' AS VARCHAR(1))").show
+-------------------+-------------------+
|CAST(123 AS STRING)|CAST(123 AS STRING)|
+-------------------+-------------------+
|                123|                123|
+-------------------+-------------------+

scala> sql("SELECT CAST('123' AS CHAR(0)), CAST('123' AS VARCHAR(0))").show
+-------------------+-------------------+
|CAST(123 AS STRING)|CAST(123 AS STRING)|
+-------------------+-------------------+
|                123|                123|
+-------------------+-------------------+

Hive

hive> SELECT CAST('123' AS CHAR(1)), CAST('123' AS VARCHAR(1));
OK
1	1

hive> SELECT CAST('123' AS CHAR(0)), CAST('123' AS VARCHAR(0));
FAILED: RuntimeException Char length 0 out of allowed range [1, 255]

Issue Links

• relates to: SPARK-19459 ORC tables cannot be read when they contain char/varchar columns
• links to: [Github] Pull Request #20887 (dongjoon-hyun)