[FLINK-6851] Improve built-in scalar function when upgrade calcite version to 1.13

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Abandoned
• Component/s: Table SQL / API
• Assignee: sunjincheng
• Reporter: sunjincheng
• Created: 06/Jun/17 00:32
• Updated: 30/Jan/19 10:31
• Resolved: 30/Jan/19 10:31

Description

I found that the difference between calcite 1.12 and calcite master affects the semantics of the build-in scalar function, for example:
The SQL:

select 
timestampadd(MONTH, 1,  date '2016-05-31'), 
timestampadd(MONTH, 5,  date '2016-01-31'), 
timestampadd(MONTH, -1, date '2016-03-31') from depts;

Calcite 1.12 Result:

+--------+--------+--------+
| EXPR$0 | EXPR$1 | EXPR$2 |
+--------+--------+--------+
| 2016-07-01 | 2016-07-01 | 2016-03-01 |
+--------+--------+--------+
3 rows selected (1.06 seconds)

Calcite Master result:

+--------+--------+--------+
| EXPR$0 | EXPR$1 | EXPR$2 |
+--------+--------+--------+
| 2016-06-30 | 2016-06-30 | 2016-02-29 |
+--------+--------+--------+
3 rows selected (1.047 seconds)

The MSSQL result:

select
  dateadd(MONTH, 1, '2016-05-31'),
  dateadd(MONTH, 5, '2016-01-31'),
  dateadd(MONTH, -1,'2016-03-31') from pvt;
2016-06-30 00:00:00 2016-06-30 00:00:00 2016-02-29 00:00:00

So, when we flink upgrade the calcite version to 1.13. we can open some useful feature. e.g: TIMESTAMPADD(SqlTypeFamily.ANY, SqlTypeFamily.INTEGER, SqlTypeFamily.DATE).

So, In this JIRA. will create some subtask to improve flink according to calcite 1.13.
I appreciated if you can give me some feedback?

Issue Links

• is blocked by: FLINK-6429 Bump up Calcite version to 1.13