ORACLE表统计信息与列统计信息

我们在分析某些语句的性能时，会分析一些信息。像表、列、索引、直方图等等，本篇主要讲表与列、索引的统计信息收集与分析。

一、表统计信息

1. 表已创建，需要收集统计信息
BEGIN
DBMS_STATS.GATHER_TABLE_STATS(OWNNAME => '',
TABNAME => '',
ESTIMATE_PERCENT => 100,
METHOD_OPT => 'for all columns size skewonly',
CASCADE => TRUE);
END;

2. 查看表的统计信息用dba_tab_statistics。

col OWNER for a16
col table_name for a20
col PARTITION_NAME for a18
col OBJECT_TYPE for a12
col LAST_ANALYZED for a20
SELECT OWNER,TABLE_NAME,PARTITION_NAME,OBJECT_TYPE,NUM_ROWS, BLOCKS, EMPTY_BLOCKS, AVG_SPACE, CHAIN_CNT, AVG_ROW_LEN,to_char(LAST_ANALYZED,'yyyy-mm-dd hh24:mi:ss') LAST_ANALYZED,USER_STATS
FROM DBA_TAB_STATISTICS WHERE TABLE_NAME = 'MSTB_ORDER_HEADER' ;
OWNER TABLE_NAME PARTITION_NAME OBJECT_TYPE NUM_ROWS BLOCKS EMPTY_BLOCKS AVG_SPACE CHAIN_CNT AVG_ROW_LEN LAST_ANALYZED USER_S
---------------- ----------- -------------- ------------ ---------- ---------- ------------ ---------- ---------- ----------- ------------- ------
POS_ORDER MSTB_ORDER_ TABLE 40445748 1928239 0 0 0 317 2016-01-07 17 NO
POS_ORDER MSTB_ORDER_ P_MAXVALUE PARTITION 1 19 0 0 0 197 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2009_10 PARTITION 0 0 0 0 0 0 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2009_11 PARTITION 0 0 0 0 0 0 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2009_12 PARTITION 0 0 0 0 0 0 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2012_1 PARTITION 0 0 0 0 0 0 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2012_10 PARTITION 73496 3299 0 0 0 260 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2012_11 PARTITION 69908 3138 0 0 0 259 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2012_12 PARTITION 66918 3018 0 0 0 260 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2012_6 PARTITION 45502 1940 0 0 0 256 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2012_7 PARTITION 53236 2279 0 0 0 256 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2012_8 PARTITION 52568 2279 0 0 0 254 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2012_9 PARTITION 71614 3219 0 0 0 260 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2013_1 PARTITION 72958 3293 0 0 0 260 2016-01-07 16 NO
POS_ORDER MSTB_ORDER_ P_MOH_2013_10 PARTITION 75330 3449 0 0 0 261 2016-01-07 16 NO
关于每一列的解释联机文档上都有(http://docs.oracle.com/cd/E11882_01/server.112/e40402/statviews_2114.htm#REFRN20376)
这里blocks是高水位以下的数据块数，empty_blocks是高水位以上的数据块数。
Dbms_stats不计算EMPTY_BLOCKS、AVG_SPACE、CHAIN_CNT。可以使用ANALYZE收取(ANALYZE TABLE <tabname> COMPUTE STATISTICS)</tabname<>
chain_cnt字段表示行迁移和行链接的数量信息.

二、列统计信息

查看列的统计信息用DBA_TAB_COL_STATISTICS
col COLUMN_NAME for a30
col LOW_VALUE for a18
col HIGH_VALUE for a24
col HISTOGRAM for a20
SELECT COLUMN_NAME,
NUM_DISTINCT,
LOW_VALUE,
HIGH_VALUE,
DENSITY ,
NUM_NULLS ,
AVG_COL_LEN ,
HISTOGRAM,
NUM_BUCKETS
FROM DBA_TAB_COL_STATISTICS
WHERE TABLE_NAME = '&TABLE_NAME';

COLUMN_NAME NUM_DISTINCT LOW_VALUE HIGH_VALUE DENSITY NUM_NULLS AVG_COL_LEN HISTOGRAM NUM_BUCKETS
------------------- ------------ --------------- ---------------- ---------- ---------- ----------- --------------- -----------
ORIGINAL_ORDER_PV 3707 80 C24C5F19 .00026976 40419026 2 NONE 1
RSVST3 4 4230363235 4230363238 .25 40194414 2 NONE 1
INVOICE_FLAG 2 80 C102 4.7902E-08 30012940 2 FREQUENCY 2
INVOICE_REPRINT 2 80 C102 1.7177E-07 37531990 2 FREQUENCY 2
POSWTO_FLAG 1 80 80 1 0 2 NONE 1
INVOICE_SALES_DATE 705 78720106010101 78730A1A010101 .002169197 32955696 3 HEIGHT BALANCED 254
ORDER_SUM_QUANTITY 1400 C102 C4025A645B .000714286 9823640 3 NONE 1
RECIEVABLE_AMOUNT 81810 3C1E2E5B66 C50A1C143415 .000844595 0 4 HEIGHT BALANCED 254
ORDER_BV 158632 3C640B66 C508590C3B35 6.3039E-06 0 4 NONE 1
ORDER_PV 167806 3D582C5166 C4400707450B 5.9593E-06 0 4 NONE 1
SALE_DATE 1293 78700206010101 78C70C04010101 .001996008 0 8 HEIGHT BALANCED 254
REGION_CODE 2 303530 333630 1.2352E-08 0 4 FREQUENCY 2
SHOP_CODE 318 3031 6E756C6C .00304878 0 5 HEIGHT BALANCED 254
SHIPPING_WAREHOUSE 389 3031 5443305A .002808989 0 5 HEIGHT BALANCED 254

字段解释，可参考联机文档（http://docs.oracle.com/cd/E11882_01/server.112/e40402/statviews_2101.htm#REFRN20275）

统计信息解释一下：
NUM_DISTINCT：该列中唯一值的数量。
LOW_VALUE：该列的最小值。显示为内部存储格式。对于字符串列，只存储前32字节。
HIGH_VALUE：该列的最大值。显示为内部存储格式。对于字符串列，只存储前32字节。
DENSITY：0到1之间的一个小数。接近0表示对于该列的过滤操作能去掉大多数的行。接近1表示对于该列的过滤操作起不到什么作用。如果没有直方图，DENSITY=1/NUM_DISTINCT。
HISTOGRAM：表明是否有直方图信息，如果有，是什么类型？FREQUENCY表示频率类型，HEIGHT BALANCED表示平均分布类型；如果没有，则为NONE。
NUM_BUCKETS：直方图里的桶数。它表示一组同类的数值放在一起。直方图最少由一个桶组成。如果没有直方图，则为1，最大桶数为254。

这里的LOW_VALUE和HIGH_VALUE都是内部格式，所以必须转换为可读懂的格式，有两种方法。

1）使用工具包utl_raw提供的函数cast_to_binary_double、cast_to_binary_float、cast_to_binary_integer、cast_to_number、cast_to_nvarchar2、cast_to_raw和cast_to_varchar2。这些函数就是把内部存储格式转换为实际值。

SELECT UTL_RAW.CAST_TO_NUMBER(LOW_VALUE), UTL_RAW.CAST_TO_NUMBER(HIGH_VALUE)
FROM DBA_TAB_COL_STATISTICS
WHERE TABLE_NAME = 'MSTB_ORDER_HEADER'
AND COLUMN_NAME = 'GROUP_ORDER_ID';

UTL_RAW.CAST_TO_NUMBER(LOW_VALUE) UTL_RAW.CAST_TO_NUMBER(HIGH_VALUE)
--------------------------------- ----------------------------------
0 7821636

2) 使用dbms_stats提供的过程convert_raw_value、convert_raw_value_nvarchar和convert_raw_value_rowid。该过程不能直接在SQL语句中使用，通常只用于PL/SQL程序中。

DECLARE
L_LOW_VALUE DBA_TAB_COL_STATISTICS.LOW_VALUE%TYPE;
L_HIGH_VALUE DBA_TAB_COL_STATISTICS.HIGH_VALUE%TYPE;
L_VAL1 T.VAL1%TYPE;
BEGIN
SELECT LOW_VALUE, HIGH_VALUE
INTO L_LOW_VALUE, L_HIGH_VALUE
FROM DBA_TAB_COL_STATISTICS
WHERE TABLE_NAME = 'MSTB_ORDER_HEADER'
AND COLUMN_NAME = 'GROUP_ORDER_ID';

DBMS_STATS.CONVERT_RAW_VALUE(L_LOW_VALUE, L_VAL1);
DBMS_OUTPUT.PUT_LINE('low_value: ' || L_VAL1);
DBMS_STATS.CONVERT_RAW_VALUE(L_HIGH_VALUE, L_VAL1);
DBMS_OUTPUT.PUT_LINE('high_value: ' || L_VAL1);
END;
/

UTL_RAW.CAST_TO_NUMBER(LOW_VALUE) UTL_RAW.CAST_TO_NUMBER(HIGH_VALUE)
--------------------------------- ----------------------------------
0 7821636

详细使用方法可参考 “ORACLE utl_raw函数与dbms_stats.convert_raw_value函数使用”
http://blog.itpub.net/17086096/viewspace-1983619/

ITPub博客

ORACLE表统计信息与列统计信息、索引统计信息

最新文章