I have the same feeling with Stamatis.
I completely understand that Vladimir is trying to keep it working correctly.
However, I know some downstream project already updated Calcite version and
changed their code to adapt to the Sarg/SEARCH operator. Reverting it and in
case we failed to fix
Probably RelRule should not change nullability. But we have done so in
the past, and I suspect we continue to do so today. It would be
interesting to find a list of rules that do this.
Deducing that a field is never null is very useful. With that
knowledge, further optimizations become possible.
Danny>-1 for the revert, we should fix the issues we encountered instead of
Danny>reverting the code brainless for a whole release.
Hi Danny, did you just veto the code change? (see
https://www.apache.org/foundation/voting.html#Veto )
I am afraid I fail to find a technical justification behind