Re: 9/18 Visual Planner Meeting Wrapup

On Tue, 2008-10-21 at 10:07 -0700, Jeff Davis wrote:
> I think the only correct answer to your problem involves functional
> relationships that cannot be (sanely) materialized (as in my example,
> the functional relationship between a specific timestamp and the hour in
> which it occurs). To be useful, the function must not be one-to-one (as
> in my example), otherwise you could just join using the original value.

[ replying to myself ]

On second thought, there's no reason the function must not be
one-to-one, although many practical examples involve functional
relationships that are not one-to-one.

An example that uses a one-to-one function would be something like:
"give me all events that had a lower turnout than the same event at the
same time on the previous day". This is a self-join using date math
(subtracting one day), so there can't possibly be a FK relationship
there.

Also, I am not 100% sure that correct answers need to use a functional
relationship like that, but they are the only examples I can think of.
This is because a functional relationship is basically like a relation,
except that it can't be materialized within reason (e.g. a relation
representing the + operator would be huge).

The inability to materialize the relation means that it can't be stored
in a variable, and therefore there can be no FK or constraint of any
kind (you can only constrain variables, not values).

Regards,
Jeff Davis