On Tue, Jan 16, 2024 at 11:46 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> They're by no means independent. What would it mean to have a
> multirange without the underlying range type?
It would mean just that - no more, and no less. If it's possible to
imagine a data type that stores pairs of values from the underlying
data type with the constraint that the first is less than the second,
plus the ability to specify inclusive or exclusive bounds and the
ability to have infinite bounds, then it's equally possible to imagine
a data type that represents a set of such ranges such that no two
ranges in the set overlap. And you need not imagine that the former
data type must exist in order for the latter to exist. Theoretically,
they're just two different data types that somebody could decide to
create.
> Also, we already
> treat the multirange as dependent for some things:
But this seems like an entirely valid point.
--
Robert Haas
EDB: http://www.enterprisedb.com