[FLINK-22586] Improve precision derivation for decimal arithmetics

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.13.0
• Fix Version/s: 1.14.0
• Component/s: Table SQL / Runtime
• Labels: pull-request-available
• Assignee: Shuo Cheng
• Reporter: Shuo Cheng
• Created: 06/May/21 11:50
• Updated: 28/Aug/21 11:20
• Resolved: 11/May/21 10:59

Description

Currently the precision and scale derivation is not properly for decimal data arithmetics, e.g,

considering the following example:

select cast('10.1' as decimal(38, 19)) * cast('10.2' as decimal(38, 19)) from T

the result is `null`, which may confuses use a lot, because the result is actually not that big.

The root cause is the precision derivation for the above multiplication is:

(38, 19) * (38, 19) -> (38, 38)

So there is no space for integral digits, which leads to null results. 

Issue Links

• duplicates: FLINK-20471 flink1.10SQL Precision overflow return null when using decimal
• is related to: FLINK-22182 Incorrect round when dividing decimals
• is related to: FLINK-20471 flink1.10SQL Precision overflow return null when using decimal
• links to: GitHub Pull Request #15848