lincoln lee created FLINK-38400:
-----------------------------------
Summary: STDDEV/VAR function with FILTER condition may cause wrong
result
Key: FLINK-38400
URL: https://issues.apache.org/jira/browse/FLINK-38400
Project: Flink
Issue Type: Bug
Affects Versions: 2.1.0, 1.20.2
Reporter: lincoln lee
Assignee: lincoln lee
Fix For: 2.2.0, 2.1.1, 1.20.4
As reported in https://issues.apache.org/jira/browse/CALCITE-7192, the
following query may produce wrong result:
{code}
SELECT STDDEV_POP(salary) FILTER (WHERE salary > 1000) FROM employees;
{code}
the decomposition for STDDEV_POP will be:
{code}
SQRT((SUM(x * x) - SUM(x) * SUM(x) / COUNT(x))/ COUNT(x))
{code}
for the above example, the `SUM(salary * salary)` lost the filter `salary >
1000`.
*Impact*
This affects all functions that use the `reduceStddev` method:
-**STDDEV_POP(x) FILTER (WHERE condition)**
- **STDDEV_SAMP(x) FILTER (WHERE condition)**
-**VAR_POP(x) FILTER (WHERE condition)**
- **VAR_SAMP(x) FILTER (WHERE condition)**
Before we bump up to calcite 1.41.0, we should port the fix first.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
