From: | Roberto Cornacchia <rcorna(at)tin(dot)it> |
---|---|
To: | Bruce Momjian <maillist(at)candle(dot)pha(dot)pa(dot)us> |
Cc: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, pgsql-hackers(at)postgreSQL(dot)org |
Subject: | Re: [HACKERS] Re: Top N queries and disbursion |
Date: | 1999-10-08 13:18:42 |
Message-ID: | 37FDEF32.DDE759AC@tin.it |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Bruce Momjian wrote:
>
> > No, it's certainly not the right thing. To my understanding, disbursion
> > is a measure of the frequency of the most common value of an attribute;
> > but that tells you very little about how many other values there are.
> > 1/disbursion is a lower bound on the number of values, but it wouldn't
> > be a good estimate unless you had reason to think that the values were
> > pretty evenly distributed. There could be a *lot* of very-infrequent
> > values.
> >
> > > with 100 distinct values of an attribute uniformly distribuited in a
> > > relation of 10000 tuples, disbursion was estimated as 0.002275, giving
> > > us 440 distinct values.
> >
> > This is an illustration of the fact that Postgres' disbursion-estimator
> > is pretty bad :-(. It usually underestimates the frequency of the most
> > common value, unless the most common value is really frequent
> > (probability > 0.2 or so). I've been trying to think of a more accurate
> > way of figuring the statistic that wouldn't be unreasonably slow.
> > Or, perhaps, we should forget all about disbursion and adopt some other
> > statistic(s).
>
> Yes, you have the crux of the issue. I wrote it because it was the best
> thing I could think of, but it is non-optimimal. Because all the
> optimal solutions seemed too slow to me, I couldn't think of a better
> one.
Thank you, Tom and Bruce.
This is not a good news for us :-(. In any case, is 1/disbursion the
best estimate we can have by now, even if not optimal?
Roberto Cornacchia
Andrea Ghidini
From | Date | Subject | |
---|---|---|---|
Next Message | Ansley, Michael | 1999-10-08 13:38:07 | RE: [HACKERS] Re: Top N queries and disbursion |
Previous Message | Jan Wieck | 1999-10-08 12:32:32 | RI status report #4 (come and join) |