Re: [Mged-queries] DW

[<ja...@op...>]

"Ugis Sarkans" <ug...@eb...> writes:

> Performance is exactly the reason why I think something
> like what I described will be needed for real databases if queries
> other than very simple ones are needed. Normal views
> are as (in)efficient as direct complicated queries, so
> some materialization would be helpful. 

I read that sometimes making something a view can help out the query
optimizer so that it can make more efficient queries, but it still has
to be executed each time, so the performance increase must be smaller
than what it would be if things were pre-computed.

> Also, conceptually it would make life easier to think of classes
> that could be queried, and on which attributes they could be
> queried, even if those attributes are not real in MAGE and have to
> be computed in some way; therefore we can split the tasks of
> defining (requirements for) queries and figuring out how to
> implement them. Another task is to understand what functions are
> needed in order to support data-oriented queries (like
> "downRegulated").

Right, I agree. That seems like part of the job of this group then is
to come up with that list.

jas.