Subject | Re: Index topic on Firebird book |
---|---|
Author | lysander_fb |
Post date | 2005-06-22T08:11:19Z |
--- In firebird-support@yahoogroups.com, "jameschua1021"
<cai.haibin@g...> wrote:
the plan analyser in IBExpert).
Also it depends on how you are limiting the resultset by using WHERE.
If you do a plain left outer join without a WHERE-clause on the left
side an index should not help you too much.
could understand it.
selectivity of 0.5 means that in average you do have 2 records per
index-key. And that should not be a problem. There are more functions
to find out if this average is outweighted by extreme high
duplications of only a few fields.
Check your documentation and your management-tool for
functions/statistics about "duplications" or "dupes" ... something
like that.
ciao,
André
<cai.haibin@g...> wrote:
> But itIt depends on the join. Look at the generated plan (for example with
> doesn't mention which side of the table's column should be index? Is
> one side enough?
the plan analyser in IBExpert).
Also it depends on how you are limiting the resultset by using WHERE.
If you do a plain left outer join without a WHERE-clause on the left
side an index should not help you too much.
> How is selectivity and statistic related. From the book I learn thatI think that this _is_ explained in the book, and at least so that _I_
> selectivity that closes to 1 is the better(do we have a treshold here
> as to what is acceptable is a selectivity of .5 still good?)
could understand it.
selectivity of 0.5 means that in average you do have 2 records per
index-key. And that should not be a problem. There are more functions
to find out if this average is outweighted by extreme high
duplications of only a few fields.
Check your documentation and your management-tool for
functions/statistics about "duplications" or "dupes" ... something
like that.
ciao,
André