Re: Rethinking opclass member checks and dependency strength

 

PostGIS has some rarely used box operations as part of GiST opclass, like "overabove". 

These are source of misunderstanding, as it hinges on the fact that non-square geometry will be coerced into a box on a call, which is not obvious when you call it on something like diagonal linestrings.
It may happen that we will decide to remove them. On such circumstances, I expect that ALTER OPERATOR CLASS DROP OPERATOR will work. 

Other thing that I would expect is that DROP FUNCTION ... CASCADE will remove the operator and unregister the operator from operator class without dropping operator class itself.