On 18.11.10 20:27, Bergquist, Brett wrote:
Thanks for the explanation! So if I understand this correctly, if the indexes that I am querying on are
always unique (the computed value that is index is always unique), then as long as the optimizer knows
about these statistics (ie I have run "sysc
trigali [mailto:mikem_...@sbcglobal.net]
Sent: Thursday, November 18, 2010 1:51 PM
To: derby-dev@db.apache.org
Subject: Re: Question on cardinality, statistics, and when things go stale
derby uses 2 types of statistics, for this discussion I will call them
distribution and cardinality.
For distrib
derby uses 2 types of statistics, for this discussion I will call them
distribution and cardinality.
For distribution derby uses the index themselves at query compile time,
so this info never goes stale. For distribution the optimizer might
need to know what percentage of the keys in an index ar
I have read as much as I can about this subject and am a little confused. I
understand the cardinality statistics are not computed if the tables are empty
when indexes are created and I can work around that.
Say I have a table with 15 million records in it and there is a main index that
I need