david@smooth1.co.uk wrote:
> On 19 Jun, 23:05, Fernando Nunes <s...@domus.online.pt> wrote:
>> Jack Parker wrote:
>>> Cheetah has a feature where statistics are gathered while the index is being
>>> created, thus obviating the need to update stats after the index creation.
>> Some statistics are gathered. Not all that a column with an index should have...
>>
>> --
>> Fernando Nunes
>> Portugal
>>
>> http://informix-technology.blogspot.com
>> My email works... but I don't check it frequently...
>
> So this feature that was ported from DB2 is crap then? Typically half-
> baked implmentation from another product.
>
> Why does it not built everything if it has read and sorted the columns
> already?
>

Err... this is hardly a "feature that can be ported". The way statistics are
gathered and represented is completely different. You may say the idea was
ported... I don't know... But in this case it was not necessarily from DB2. I
imagine most databases do this... or should do...

As for your specific question, the definition of what are the statistics that a
column "should" have is not absolute... It really depends on the column, the
data and the queries.. It collects medium distributions for the column heading
and index... And LOW for the index itself...
It does not collect distributions for other columns in the index... And you may
like distributions in high mode for the heading column...

For more information about the modes and best practices please check the
performance guide.

This area will hopefully see more autonomic stuff in the future...

--
Fernando Nunes
Portugal

http://informix-technology.blogspot.com
My email works... but I don't check it frequently...