本文只涉及使用dbm_stats來收集或刪除列的統計信息的一些命令，以備查詢。
測試表如下(實驗環境爲10.2.0.4):
SQL> create table test(i int,a varchar2(30)); Table created. SQL> insert into test select rownum,object_name from all_objects; 9907 rows created.
簡單的說，列的統計信息，主要包括兩種類型：
- 只有基本信息：收集的統計信息只有1個桶(bucket)
- 包含柱狀圖信息：收集的統計信息包含2到254個桶
也就是說，如果想收集列的基本信息，同時不希望收集柱狀圖，則需要指定bucket的size爲1：
SQL> exec dbms_stats.gather_table_stats(user, 'TEST', cascade=>false, method_opt=>'for columns i size 1'); PL/SQL procedure successfully completed. SQL> select column_name,num_distinct,num_buckets,low_value,high_value,density, avg_col_len,histogram 2 from user_tab_columns where table_name='TEST'; COLUMN_NAM DISTINCT BUCKETS LOW_VALUE HIGH_VALUE DENSITY AVG_COL_LEN HISTOGRAM ---------- -------- ------- ---------- ---------- ---------- ----------- ---------- I 9907 1 C102 C26408 .000100939 4 NONE A NONE
如果要收集列的柱狀圖信息，則bucket的個數必須大於等於2(最多不超過254)
SQL> exec dbms_stats.gather_table_stats(user, 'TEST', cascade=>false, method_opt=>'for columns i size 2'); PL/SQL procedure successfully completed. SQL> select column_name,num_distinct,num_buckets,low_value,high_value,density, avg_col_len,histogram 2 from user_tab_columns where table_name='TEST'; COLUMN DISTINCT BUCKETS LOW_VALUE HIGH_VALUE DENSITY AVG_COL_LEN HISTOGRAM ------ -------- ------- --------- ---------- ---------- ----------- --------------- I 9907 2 C102 C26408 .000100939 4 HEIGHT BALANCED A NONE
如果要刪除列已有的柱狀圖信息而保留列的基本統計信息，則需要重新收集bucket爲1的統計信息
SQL> exec dbms_stats.gather_table_stats(user, 'TEST', cascade=>false, method_opt=>'for columns i size 1'); PL/SQL procedure successfully completed. SQL> select column_name,num_distinct,num_buckets,low_value,high_value,density, avg_col_len,histogram 2 from user_tab_columns where table_name='TEST'; COLUMN DISTINCT BUCKETS LOW_VALUE HIGH_VALUE DENSITY AVG_COL_LEN HISTOGRAM ------ -------- ------- ---------- ---------- ---------- ----------- ---------- I 9907 1 C102 C26408 .000100939 4 NONE A NONE
這個操作明顯不太合理，重新收集統計信息的代價有時候是很大的，所以Oracle11g對此做出了改進，允許只刪除柱狀圖而保留基本統計信息，命令語法如下：
exec dbms_stats.delete_column_stats(user, 'TEST','I',col_stat_type=>'HISTOGRAM');
而要徹底刪除整個列的統計信息，則需要調用delete_column_stats過程
SQL> exec dbms_stats.delete_column_stats(user, 'TEST', 'I'); PL/SQL procedure successfully completed. SQL> select column_name,num_distinct,num_buckets,low_value,high_value,density, avg_col_len,histogram 2 from user_tab_columns where table_name='TEST'; COLUMN DISTINCT BUCKETS LOW_VALUE HIGH_VALUE DENSITY AVG_COL_LEN HISTOGRAM ------ -------- ------- ---------- ---------- ------- ----------- ---------- I NONE A NONE
可以在同一個過程中收集多個列的統計信息，並且可以爲不同的列指定不同的bucket個數：
SQL> exec dbms_stats.gather_table_stats(user, 'TEST',
cascade=>false, method_opt => 'for columns size 1 T for columns size 2 A');
PL/SQL procedure successfully completed.
SQL> select column_name,num_distinct,num_buckets,low_value,high_value,density,
avg_col_len,histogram
2 from user_tab_columns where table_name='TEST';
COLUMN DISTINCT BUCKETS LOW_VALUE HIGH_VALUE DENSITY AVG_COL_LEN HISTOGRAM
------ -------- ------- --------- -------------------- ---------- ----------- ---------------
I 9907 1 C102 C26408 .000100939 4 NONE
A 7376 2 41 5F75746C245F6C6E635F .000185239 18 HEIGHT BALANCED
696E645F7061727473
值得注意的是，9i的dbms_stats中，method_opt的默認值是FOR ALL COLUMNS SIZE 1，也就是收集列的基本統計信息而不收集柱狀圖信息，而10g的默認值則變成了FOR ALL COLUMNS SIZE AUTO，則Oracle在收集列的基本信息之外，還會根據情況收集某些列的柱狀圖。