Andres Freund <andres@2ndquadrant.com> writes:
> ... So we'd get approximately one line further unless we fix this for DROP
> DEFAULT and DROP CONSTRAINT as well.
True. As far as pg_dump --clean is concerned, it'd undoubtedly be easier
if we did what you suggest and just eliminate the emitted DROP commands
for table components, relying on the assumption that there'll never be
a partial-dump mode that would allow dumping a table's components without
the table. However, the server-side approach has the benefit that it'll
likely make life easier for other applications besides pg_dump.
regards, tom lane
