Re: Extending my question. Was: The relational model and relational algebra - why did SQL become the industry standard?

Bob Badour wrote:
>"Jan Hidders" <jan.hidders_at_REMOVE.THIS.ua.ac.be> wrote in message
>news:3e6da66b.0_at_news.ruca.ua.ac.be...
>>
>> I'm not convinced that bags are never needed. [...] And if you are then
>> going to simulate them with sets then the cost might become negative
>> because the set-based optimizer might miss certain optimizations that
>> would have been easier to spot for a bag-based optimizer.
>
>Such as?

Such as combining two iterations over the same bag into one. It's pretty easy to see that

  SELECT f(x)
  FROM x IN
    SELECT g(y)
    FROM y IN Y

is the same as

  SELECT f(g(y))
  FROM y in Y

which is a simple straightforward rule in bag algebra. However, if you simulate that with sets the optimizer has to see that

  SELECT f(x) AS fx, COUNT(*) AS cnt
  FROM x IN
    SELECT g(y) AS fy, COUNT(*) AS cnt
    FROM y IN Y
    GROUP BY fy
  GROUP BY fx

is the same as

  SELECT f(g(y)) AS fx, COUNT(*) AS cnt
  FROM y IN Y
  GROUP BY fx

which is a lot less straightfoward.

• Jan Hidders