Re: Interval literal not ANSI compliant

Ed Smith <edsmithed(at)yahoo(dot)com> writes:
> The Postgres INTERVAL literal is not compliant with
> the ANSI 2003 SQL Spec.

Yup.

> Are there plans to fix this?

Step right up and have at it. Tom Lockhart was working on migrating the
datetime support to be more like the (IMHO quite bizarre) spec syntax,
but he lost interest and dropped out of the project awhile back.
I don't think any of the rest of the current developers care much about
this point. But we'd accept a patch, as long as it was reasonably
cleanly coded (ie, supportable into the future).

The interval datatype needs love in other ways --- for instance, in my
opinion it really ought to store months/days/seconds internally not
just months/seconds, so as to avoid surprising behavior at DST
transitions. (The existing representation effectively assumes that a
day is always the same number of seconds, which is wrong on DST
transition days.) But this area is not high on the list of interests
of any active PG developers. We need somebody to take ownership of the
problem.

regards, tom lane