Re: A new proof of the superiority of set oriented approaches: numerical/time serie linear interpolation
Date: 3 May 2007 04:53:07 -0700
Message-ID: <1178193187.891144.135920_at_q75g2000hsh.googlegroups.com>
On May 3, 1:41 pm, Kevin Kirkpatrick <kvnkrkpt..._at_gmail.com> wrote:
> Not sure if it advances the conversation, but I thought I'd look into
> this a bit further.
> I decided to look into this a bit further. Obviously, no by-hand
> iteration-based code is going to beat a declarative solution executed
> by a "sufficiently intelligent" optimizer. Yet, I think the query can
> be expressed much more concisely (allowing the optimizer a much better
> chance at an efficient execution) than what Cimode posted.
The query can probably be seriously improved while remaining in set
oriented approach.
[Snipped]
Kevin,
It seems you diverted from the initial example given.
I suggest to redo the initial experience without changing the initial
facts (I purposely posted it for that and without changing the
incorrect design of it - that's how it was given to me), it would
beneficial to retake the exact same canvas than the one proposed.
Ideally I will try to adapt Vadim's regressive solution. Finally I
believe it would be difficult to consider such experiment as
conclusive without some kind of stress tool simulating concurrent
random activity while running the query.
Then and only then we could imagine some experiment to see how current
SQL DBMS's behave...(ORACLE, SQL Server,...)..
Regards...