steve wrote:
> Given the idea that Rank is a primary concept I'm suggesting that it's
> understandable based on more primitive ones. One of which a cursor
> (type). It's where the order in the ranking comes from. Rather than
> shoehorning it in it seems to be a case of - if the shoe fits wear
> it :) I'm not sure I understand your suggestion of introducing
> complexity. . The idea of a cursor seems very basic. One of the themes
> of a relational database is to simplify understanding. That's what I
> was trying to do though I'm not sure I succeeded:) Behind every
> function is an idea. Ranking is no exception.
Replacing this:
select c1, c2, rank() over (order by c1) r
from t
order by c1, c2
with this:
select ToTable(
ToList(
cursor(t order by c1)
)
)
order by c1, c2
is hardly a simplification.
>> But this is arguably not a rank; it's an understandable concept, but it
>> deserves a different name. More to the point, it deserves a practical
>> example in which you would actually want to do something like this; I
>> can't think of one offhand.
>
> I'm taking the position that a rose is a rose is a rose. Because the
> data doesn't fit within the parameters of the dense rank function is
> that reason to call it something else? :)
I suppose it's a rank over something that isn't a simple ORDER BY. I
still can't think of a practical situation where you'd actually want
this, though, and this:
> You may find this post interesting on the subject:
>
> microsoft.public.sqlserver.programming
> Oct 30 2007
> 'Count Occurances in Select Statement'
> http://www.themssforum.com/SQLServerDev/Count-Occurances-726289/
is an equally artificial example and thus doesn't help.
Assuming that such a practical situation does exist and I just don't
know that it is (i.e. that Josh is not a sock puppet of yours), I
would still prefer to see it expressed without explicit reference to
a cursor:
select c1, c2, c3, rank() over (order by c1, change(c2)) r
from t
order by c1, c2, c3
Some of your ideas (this one among them) strike me as analogous to
CISC, which has been largely superseded by RISC for a reason: making
common simple cases harder in return for making uncommon complex cases
easier is generally not a net win.
|
