Olav Sandstaa wrote:
> Hi Jørgen,
>
> I think both your code change and test case look fine. OK to push.
>
> Related question: How much work do you think it would be to do a
> proper/correct optimization for XOR?
>
I don't have any good time estimates, but the comparable Item_cond_and
is treated specially (if/switch with COND_AND_FUNC) around 20 places in
the optimizer. In addition, NDB is affected. Further, for it to be
useful we would also need to implement e.g. index_merge for it. We're
probably looking at (a few) month time-units.
I think it is correctly set as low priority (see comment for the xor
item); my guess is that it will never be optimized because there will
always be better things to do.
--
Jørgen Løland
