well -- you need to be a little more specific. We can find the allocated space (how much space has been carved out to be used only by this index segment) and we can find the amount of space within that space that is/has been actually used (the 'high water mark').

Consider -- I have an index T_IDX:

ops$tkyte@8i> select bytes, blocks, extents from user_segments where segment_name = 'T_IDX';

BYTES BLOCKS EXTENTS
---------- ---------- ----------
1048576 128 2

There is the amount of space dedicated to this segment. 128 blocks.


If I analyze the index (estimate on a big one)

ops$tkyte@8i> analyze index t_idx compute statistics;
Index analyzed.

ops$tkyte@8i> select blevel, leaf_blocks, num_rows from user_indexes where index_name = 'T_IDX';

BLEVEL LEAF_BLOCKS NUM_ROWS
---------- ----------- ----------
1 89 20252


I find that there really are about 89 blocks in use. These stats will remain in the data dictonary until i re-analyze or delete the stats.

Further, if I "validate" the index, I can get a little more data:

ops$tkyte@8i> analyze index t_idx validate structure;
Index analyzed.

ops$tkyte@8i> select height, blocks, lf_blks, br_blks, btree_space, used_space from index_stats where name = 'T_IDX';

HEIGHT BLOCKS LF_BLKS BR_BLKS BTREE_SPACE USED_SPACE
---------- ---------- ---------- ---------- ----------- ----------
2 128 89 1 719672 634258

That shows that of the allocated and used space -- I have about 700k, The space ACTUALLY in use in that allocated and used space is about 630k. This is exactly what I expected as this is a brand new index I just created and the pctfree was set to 10% so 10% of each block has been reserved for furture inserts into these blocks --700k - 10% is 630k. Note that the index stats view "empties" itself after you disconnect -- you would have to save this data into another table if you wanted to analyze it further.