Re: Dreaming About Redesigning SQL

Christopher Browne wrote:

>Lauri Pietarinen <lauri.pietarinen_at_atbusiness.com> wrote:
>
>
>>Well, however you want to call it. SQL (or relational) represents a
>>different way of thinking.
>>
>>
>
>Yes, it does, and that is essential to _what is being done_.
>
>If you are navigating a very large set of data and want to do so both
>flexibly and efficiently, you *MUST* do that "different way of
>thinking."
>
>If you stay in the "other procedural language," then you risk making
>really dramatic mistakes.
>
>Our "Perl guy" tried staying in his procedural language; when he
>started processing a million records, this mapped to a million
>"procedural queries" that ran for days. (And he wasn't even doing
>amortizations on the _big_ set of data; that one would have taken
>WEEKS :-).) I set up a stored procedure (to do the amortization
>calculation) and showed him a SQL statement that did the million
>records in a few minutes by doing it as a single query applied to the
>whole million objects.
>

Agree 100%. I have had exactly the same experiences, and yes, the guy was using Perl!

>I am quite sure you are also conflating the fact that SQL is a bit
>clumsy ("not everyone's ideal relational language") with the necessity
>to shift thinking.
>

Yes I am. Just go thru my previous postings!

> There are several other would-be alternatives,
>whether R12 (from System R, the "original" relational database),
>Tutorial D (the Darwen/Date thing),
>

or QUEL/PostQUEL (used with Ingres

>and Postgres). They all have quite different syntax from SQL,
>arguably better/worse in their own ways, but all have that very same
>necessity of shifting thinking from a procedural mode to a "looking at
>sets of data" mode.
>

>When people _don't_ do that "thinking differently," we are certain to
>see hideous performance, and that is neither a SQL issue nor a
>"relational" issue. The point is that if they are accessing a big
>pile of data, they have to think carefully [jumping to that "different
>way of thinking"] irrespective of what specific language(s),
>libraries, or other tools they are using.
>

Yes, that's true. Even though the (SQL-) DBMS does the "how", you have to make sure it knows
what it's doing.

But my experience is that people who understand relational also understand large databases.

regards,
Lauri Received on Sat Oct 18 2003 - 14:27:17 CDT