Re: Returning 'Infinity'::TIMESTAMPTZ from "to_timestamp" function

Vitaly Burovoy <vitaly(dot)burovoy(at)gmail(dot)com> writes:
> To add an ability to construct 'Infinity' TIMESTAMPTZ via
> "to_timestamp" call, there are two ways:

> 1. Rewrite the function "pg_catalog.to_timestamp(double precision)" as
> an internal one. It's the easiest way, because it allows to avoid
> usage of INTERVAL as a helper (of course, there is still possible to
> use intervals as shown above in user's scripts, but without "Infinity"
> support).

> 2. Add support of infinite intervals. It is harder, because it touches
> a lot of functions. I can add that support if it is in demand.

> Which way is preferred?

I think you should stay away from infinite intervals; that seems like
there would be a lot of definitional questions to be resolved. Even
if we decide we want to deal with that someday, it shouldn't be a blocking
issue for conversion between infinite floats and infinite timestamps.

regards, tom lane