fedetxf wrote:

> I have a theoretical question regarding the way FB handles index
> statistics.
>
> I work with a table that has millions of records and I have an index
> on a column that has only 22 possible different values, but typically
> only 6 or 7 different values appear in the data.
> The typical occurrence of each value is like this
>
> 0: 99% of the records
> greater than 0: 1% of the records
>
> Among that 1% that has value > 0 I want to find which ones have each
> value.
>
> So the typical queries look like this
>
> select *
> from table
> where error > 0
>
> or
>
> select *
> from table
> where error = 3
>
> Some joins are used too.
>
> FB would assign a selectivity of 1/6, so when quering for a specific
> value (error = 3) it would assume it would have to read 166.666 rows
> when in fact they would be like 1.666 and when asking to read with
> error = 0 it would think it needs to read the 166.666 when in fact it
> would be reading 990.000 rows.
>
> Would FB have the risk of using a different index based on the
> statistics it would have in this situation?
>

Yes !

FB does not store a histogram of value distribution.

see you !

--
Alexandre Benson Smith
Development
THOR Software e Comercial Ltda
Santo Andre - Sao Paulo - Brazil
www.thorsoftware.com.br