Tom Lane wrote:
> The thing that jumps out at me from this example is the much larger
> estimate of returned rows in the second case. The planner is clearly
Good catch! There were 296 possible issues the table. One had 86,544
articles associated with it. The next highest was 5,949. Then the
numbers drop to 630, 506, 412, 184 and then the rest are all under 62.
Out of curiosity, how does vacuum decide on the large estimate?
The maximum is 86,544.
The average row return for ixissue = x is 3412.
The median is 25.
The mode is 25.
ixissue is the result of a sequence.
Thanks for the heads up on this...
- K
Kristofer Munn * KMI * 973-509-9414 * AIM KrMunn * ICQ 352499 * www.munn.com