Re: Backend misfeasance for DEFAULT NULL

Simon Riggs <simon(at)2ndquadrant(dot)com> writes:
> Seems more like an unwanted looseness in the meaning of an ALTER
> TABLE .. INHERIT to me. I'd prefer it if we added some extra clauses to
> ALTER TABLE:

> [ { INCLUDING | EXCLUDING } { DEFAULTS | CONSTRAINTS | INDEXES } ]

I think you're confusing this with a CREATE TABLE operation.

"Excluding constraints" is not sensible in any case: failing to inherit
check constraints should be disallowed IMHO. (There's already a TODO to
add inheritance info to pg_constraint so that that can be enforced in a
more bulletproof fashion.)

The other two categories of things are explicitly allowed to be
different between a child and a parent, and so I'm not convinced that
ALTER INHERIT has any business touching them.

But even if it's decided that the above is a sensible future feature,
it's certainly not something we can do as a backpatchable bug fix.

regards, tom lane