Thread: how the planner decides between bitmap/index scan

Hi,

does the planner know how "scattered" rows are? If yes: where is this info stored?

I'm asking because it looks to me that not only the number of rows, but also how data is on disk (more or less
sequential)is used to determine what type of index scan (bitmap/regular) has to be used... 

but in the docs:
http://www.postgresql.org/docs/8.4/static/using-explain.html
I've found only "there are so few that the extra cost of sorting the row locations is not worth it", which would mean
thatonly the number of rows is actually used to choose between the two index access methods. 

Can someone explain?

Thank you

Ops, found it:

http://www.postgresql.org/docs/8.4/static/view-pg-stats.html

("correlation" column)

I guess it would be worth pointing it out in the planner docs...



Scara Maccai wrote:
> Hi,
>
> does the planner know how "scattered" rows are? If yes: where is this
> info stored?