Oracle_scripte/Tanel/ast/03_index_density.sql
2014-09-17 13:23:48 +02:00

41 lines
1.4 KiB
SQL

-- analyze index density (how many rows per leaf block)
-- using SYS_OP_LBID function
-- replace the first argument of SYS_OP_LBID number with the object_id of the index you are scanning
-- also you need to modify the table name and index name hint to query the table/index of your interest
-- make sure that the index is actually accessed in the execution plan (both fast full scan and range/full
-- scans do work, but fast full scan is the fastest if you want to scan through the entire index segment)
--
-- additionally, you can use SAMPLE BLOCK syntax (below) to sample only some index blocks (when using
-- fast full scan)
COL blocks_histogram HEAD "Log(2,blocks) Histogram" FOR A30
SELECT
LPAD(NVL(TO_CHAR(rows_per_block), 'Total:'), 15, ' ') num_rows_in_blk
, blocks
, NVL2(rows_per_block, LPAD('#', LOG(2,blocks), '#'), null) blocks_histogram
FROM (
SELECT
CEIL(num_rows/10) * 10 rows_per_block
, COUNT(*) blocks
FROM (
SELECT /*+ INDEX_FFS(o i99) */
count(*) num_rows
FROM
indexed_objects o -- SAMPLE BLOCK (1000) o
WHERE
owner IS NOT NULL
GROUP BY
SYS_OP_LBID( 82824, 'L', o.ROWID)
)
GROUP BY ROLLUP
( CEIL(num_rows/10) * 10 )
ORDER BY
CEIL(num_rows/10) * 10
)
/