About estimate size of index extent after rebulid:

In first: u have to know type of indexes
-- compressed or not compressed
-- b+ or bit map
In second: (and this is more interesting thing) u have to know
HOW ROWS WITH SOME INDEX KEY VALUES DISTRIBUTED IN REAL TABLE.
Why:
index segment (not compressed) can keep 2 kind of entry:
1) key value + rowid
2) key value + (rowid_start, rowid_end)

In this case we have (NOTE with the same # of rows in 1 table block)

1 variant (in 1 db block):
1 row -- key1
2 row -- key1
3 row -- key2
4 row -- key2
5 row -- key1
6 row -- key2
7 row -- key1
8 row -- key2
9 row -- key2

in index we will have --> 6 entries

2 variant (in 1 db block):
1 row -- key1
2 row -- key1
3 row -- key1
4 row -- key2
5 row -- key2
6 row -- key2
7 row -- key2
8 row -- key2
9 row -- key1

in index we will have --> 3 entries

So field cluster_factor from all_indexxes in this case doesn't work.
This is one of the main problem with estimation of size of index extents.



[Edited by Shestakov on 09-04-2002 at 11:38 AM]