We plan to reimplement the CUBE operator in PostgreSQL after
errors in our approach were pointed out to us during the submission of our
patch last time. This time, we would like to follow the approach given
below. Kindly give your opinion/comments.
As opposed to our earlier approach in which we implemented the
operator at the top level, this time we plan to create a separate
parsetree for the CUBE query so that this can be taken in as another
parseitem on the outer tree in the case of nested queries. This would in
turn, solve the problems faced earlier regarding non-functioning in the
case of VIEWS, UNION et al.
Please send in your suggestions regarding this approach so that we
can start work on it.
Srikanth
Sumit