Re: simultaneously reducing both memory usage and runtime for a query

Hi again Andy,

On Sun, 28 Mar 2010 20:59:24 -0500, Andy Colson <andy(at)squeakycode(dot)net> wrote:
> On 03/28/2010 07:43 PM, Andy Colson wrote:
>> On 03/28/2010 03:05 PM, Faheem Mitha wrote:
>>>
>>>
>>
>> Wait... Your saying your question is so complex it needs 41 pages
>> (including graphs) to ask? I didn't bother before, but now I'm curious,
>> I'll have to go take a look.
>>
>> -Andy

> Faheem, you seem to be incredibly detail oriented. We probably on
> need 10% of whats in diag.pdf:

Most of this is stuff I've been asked on #postgresql. Admittedly, in
different contexts than this query.

[snippage]

> The rest is irc chat about getting the data imported into PG, and
> other than slowness problems, does not seem relevant to the sql in
> opt.pdf.

True, the IRC sessions should go. I'll take them out.

> As for opt.pdf, I dont think, again, we need all that detail. And
> the important parts got cut off. The explain analyze output is
> needed, but its cut off.

> I'd recommend you paste the output here:
>
> http://explain.depesz.com/

> And give us links. The explain analyze will have counts and info
> that we (ok not me, but Tom and others) can use to help you.

That's one way to go. I was going to paste the entirety of opt.tex
into an email. That would include all the EXPLAIN ANALYZE STUFF, but
not the graphs, and thus would be relatively self-contained. For the
graphs you'd have to look at a pdf (unless a ps.gz is preferred).

> You also seem to have gone through several revisions of the sql (I
> admit, I just skimmed the pdf's), it would be great if you could
> drop the ones you are sure are not useful, and we concentrate on
> just one or two.

Ok, I'll trim it down a bit. At least the initial queries in both
sections are not relevant. Thanks for the feedback.

Regards,Faheem.