>
>> based on the index values because 2004-12-02 is between the from and end
>> date. However, b and d would be excluded immediately because the the
>> from
>> date is greater than 2004-12-02 and would save the optimizer from even
>> reading the table for these index entries because the fromDate is in the
>> index.
>
> That is not correct. Postgres currently doesn't have a way to skip ahead
> on an index scan. So what will happen is that a, b, c, d and will all be
> considered and b and d removed by a filter rule.
>
OK. I got it now. Thanks :)