[DB-SIG] Help on this query
Andy Todd
andy47@halfcooked.com
Wed, 05 Jun 2002 14:36:26 +1000
Chris Cogdon wrote:
> On Tuesday 04 June 2002 20:53, Andy Todd wrote:
>
>
>>You are absolutely right, I am an idiot. Mind you, thats some pretty
>>scary SQL you have there and, as you say, it won't work in every
>>database. I've been sitting here scratching my head trying to come up
>>with some ANSI standard SQL that will do the job and failing terribly.
>
>
> Of course, there's lots of stuff you can do in python. However, the fewer
> queries you make to the server, and the less information that has to come
> back for processing within python, the faster things will go.
>
>
Again true, but beware premature optimisation. For all we know two
simple queries and a little bit of Python code are just as efficient as
the pure SQL you posted.
In fact, in Oracle that would end up in two queries at the server anyway
and would probably result in more process cycles to complete (as the
intermediate result sets are cached). Contrast that with the overhead of
storing two numbers at the client end. The cost of parsing not
withstanding, of course.
Anyway, its a purely theoretical argument, I just try and follow the
tenets of the Pragmatic Programmers (and Uncle Tim) and keep it simple.
Ninety nine times out of a hundred the extra cost of having multiple
simple SQL statements versus one complex statement are worth it in the
time saved developing, debugging and testing. If the simple approach
causes a hot spot in your system then by all means optimise, but we
shouldn't fuss about it until it *is* a problem.
Besides, 99.9997% of database performance problems can be removed by
adding an index <0.5 wink>
Regards,
Andy
--
----------------------------------------------------------------------
From the desk of Andrew J Todd esq - http://www.halfcooked.com