Re: A new proof of the superiority of set oriented approaches: numerical/time serie linear interpolation

> Cimode you rewrote some queries on that particular database and gained
> a performance improvement, I'm sure that makes you feel good but you
> cannot conclude a proof.

This one added to the hundred others taking out cursors.

> Your "proof" would have to be true across different queries, different
> databases, different locking mechanisms, languages, developers,
> schemas ...

Fair enough.

> I also have plenty of cases where a procedural approach yielded a 200
> fold improvement but at other times a set based one did the same. I
> cannot prove apples are better fruit than oranges.
What performance was improved? I am curious.

> BTW locking is not driven by procedural versus set based processing
> and each database vendor implements different mechanisms. For
> instance, many times Oracle will not have to lock anything and yet
> it'll get the correct results while other vendors must lock.
>
> As for missing data, that depends on each domain/type. The RM allows
> for domain/type specific operators that make sense for a given domain,
> so you could have interpolation operators and they could even be
> implemented inside the db, will your solution be faster then? Maybe,
> maybe not.
>
> Besides what does superior mean, performance, developer time,
> maintainability, higher concurrency...?
If you dig down in the subject, you will realize I have already provided that information throught exposing the results of the modification.

> Cheers.
Received on Fri May 04 2007 - 01:40:45 CDT