On Thu, 18 Apr 2002, Tom Lane wrote:
> Curt Sampson's nearby remarks about partial indexes are not a bad
> suggestion.
I just tried this out, and the disk space savings alone were pretty
stunning. On a 300,000 row table with about 1750 TRUE values and
the rest FALSE, the full index was over 5 MB and the partial was
less than 50K.
But it turns out that the analyzer's stats were good enough that
it made little difference to performance. Once I analyzed the table,
even with the full index postgres figured out that the index scan
(estimating 1300 values, in this case) would be faster.
So I guess it's key correlation thing that did it, or perhaps he
just had not analzyed the table.
cjs
--
Curt Sampson <cjs@cynic.net> +81 90 7737 2974 http://www.netbsd.org
Don't you know, in this new Dark Age, we're all light. --XTC
