Hello,
I have logged a CALCITE-5024
<https://issues.apache.org/jira/browse/CALCITE-5024> and submitted a pr-2732
<https://github.com/apache/calcite/pull/2732>, welcome suggestions and
review.
Yanjing Wang <zhuangzixiao...@gmail.com> 于2022年2月25日周五 18:16写道：
> When I override the RelDataTypeSystemImpl#getDefaultPrecision method to
> force INTERVAL_HOUR type having precision 10, the validation passes, but
> converting to rel fails.
>
> The problem is SqlLiteral.getValueAs(Long.class) will
> invoke intervalToMillis and which invokes
> intervalQualifier.evaluateIntervalLiteral(literal,
> intervalQualifier.getParserPosition(),
> RelDataTypeSystem.DEFAULT).
>
> and you can find that RelDataTypeSystem.DEFAULT has been hard coded.
>
> Yanjing Wang <zhuangzixiao...@gmail.com> 于2022年2月25日周五 16:39写道：
>
>> Hello,
>>
>> I found the calcite INTERVAL HOUR default type is HOUR(2), and the range
>> is 0-99.
>> for example:
>> @Test void testIntervalHourRange() {
>> String expr = "interval '100' hour";
>> expr(expr).ok();
>> }
>> This test will throw CalciteContextException: From line 1, column 1 to
>> line 1, column 19: Interval field value 100 exceeds precision of HOUR(2)
>> field.
>>
>> I have tried this test on MYSQL, PG, BigQuery, SparkSql, all of them
>> succeed and I think they may support out of range INTERVAL HOUR.
>>
>> In the PG doc, I found the piece "Functions justify_days and
>> justify_hours are available for adjusting days and hours that overflow
>> their normal ranges.".
>>
>> Would calcite support this automatic precision increase?
>>
>>
>>
>>
>>
>>
