Re: RFC: planner statistics in 7.2

Philip Warner <pjw@rhyme.com.au> writes:
> At 21:14 19/04/01 -0400, Tom Lane wrote:
>> But you don't really need to look at the index (if it even exists
>> at the time you do the ANALYZE).  The extent to which the data is
>> ordered in the table is a property of the table, not the index.

> But the value (and cost) of using a specific index in an indexscan depends
> on that index (or am I missing something?). 

All that we're discussing here is one specific parameter in the cost
estimation for an indexscan, viz, the extent to which the table ordering
agrees with the index ordering.  As long as they both agree about the
ordering operator, this number doesn't depend on the index --- the index
is by definition in perfect agreement with the ordering operator.  There
are other parameters in the total cost estimate that will depend on the
index, but this one doesn't AFAICS.
        regards, tom lane