Recently, we encounter a problem caused by some inconsistency query results.
After some investigations, I find the GROUP_ID()'s implementation in
Calcite is wrong [1].

However, if we want to fix it, another field is required to be added into
*Aggregate*, bringing huge impacts on almost every phases.

Therefore, can we just remove it?
On one hand, "not support" message is better than "wrong result". On the
other hand, GROUP_ID() is not in the SQL standard. The oracle function is
rarely used.

[1] https://issues.apache.org/jira/browse/CALCITE-1824

Best,
Feng

Reply via email to