Jeff Eckermann <jeff_eckermann@yahoo.com> writes:
> According to posts in the archives, the following used
> to work:
> jeff=# select timestamp(1019571752);
> ERROR: TIMESTAMP(1019571752) WITH TIME ZONE precision
> must be between 0 and 13
> jeff=#
Still works if you quote it:
regression=# select "timestamp"(1019571752);
timestamp
---------------------
2002-04-23 10:22:32
(1 row)
regression=# select "timestamptz"(1019571752);
timestamptz
------------------------
2002-04-23 10:22:32-04
(1 row)
However, the cast to abstime may be preferable, since Thomas is unlikely
to keep fiddling with the syntax and semantics of that type ;-) ;-).
Eventually there should be an exact converse of the extract(epoch)
functionality, rather than these various kluges relying on
abstime-to-integer equivalence.
We'll probably start thinking about removing abstime when Unix systems
start to migrate away from 32-bit time_t, which one hopes will happen
well before the year 2038. So eventually something's got to be done.
regards, tom lane