表的連接原理
表的連接方式：
1、循環嵌套連接（nested loop join）
2、哈西連接（hash join）
3、排序合併連接（merge join）
一、循環嵌套連接
A B 循環嵌套連接
方法：
1、掃描其中的一個小表，每讀到一條記錄，就根據這條記錄（連接字段）的值到另外一個表中去查找。
2、另外一個表一般是大表，大表的連接字段上有索引，而且索引用的比較好。
3、掃描小表每讀到一個連接字段的指，就到大表上走索引去掃描，以此循環小表，最後的結果合集就是要求查詢的結果。
小表——又叫做驅動表，或者外表
大表——又叫做被驅動表，或者內表
CBO法則下，oracle會自動去選擇驅動表，被驅動表的連接字段要求有索引而且條件要好，如果驅動表過大，也不適合這種方法。
SQL> conn plsql/plsql
Connected.
SQL> set autotrace trace exp
SQL> set linesize 1000
SQL> select h.hrc_descr,o.org_short_name from org_tab o,hrc_tab h where o.hrc_code=h.hrc_code order by 2;
Execution Plan
----------------------------------------------------------
Plan hash value: 566430324
-------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9 | 306 | 8 (25)| 00:00:01 |
| 1 | SORT ORDER BY | | 9 | 306 | 8 (25)| 00:00:01 |
|* 2 | HASH JOIN | | 9 | 306 | 7 (15)| 00:00:01 |
| 3 | TABLE ACCESS FULL| ORG_TAB | 9 | 198 | 3 (0)| 00:00:01 |
| 4 | TABLE ACCESS FULL| HRC_TAB | 9 | 108 | 3 (0)| 00:00:01 |
-------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("O"."HRC_CODE"="H"."HRC_CODE")
強制循環嵌套連接
SQL> select /*+ use_nl(o h) */ h.hrc_descr,o.org_short_name from org_tab o,hrc_tab h where o.hrc_code=h.hrc_code order by 2;
Execution Plan
----------------------------------------------------------
Plan hash value: 3315346234
--------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9 | 306 | 13 (8)| 00:00:01 |
| 1 | SORT ORDER BY | | 9 | 306 | 13 (8)| 00:00:01 |
| 2 | NESTED LOOPS | | 9 | 306 | 12 (0)| 00:00:01 |
| 3 | TABLE ACCESS FULL | ORG_TAB | 9 | 198 | 3 (0)| 00:00:01 |
| 4 | TABLE ACCESS BY INDEX ROWID| HRC_TAB | 1 | 12 | 1 (0)| 00:00:01 |
|* 5 | INDEX UNIQUE SCAN | PK_HRC_TAB | 1 | | 0 (0)| 00:00:01 |
--------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
5 - access("O"."HRC_CODE"="H"."HRC_CODE")
將HRC_TAB的數據加大
SQL> conn plsql/plsql
Connected.
SQL> desc hrc_tab
Name Null? Type
----------------------------------------- -------- ----------------------------
HRC_CODE NOT NULL NUMBER(4)
HRC_DESCR NOT NULL VARCHAR2(20)
SQL> alter table hrc_tab modify hrc_code number(10);
Table altered.
SQL> begin
2 for i in 1..100000 loop
3 insert into hrc_tab values(hrc_org_seq.nextval,'111');
4 end loop;
5 end;
6 /
PL/SQL procedure successfully completed.
SQL> commit;
Commit complete.
SQL> select h.hrc_descr,o.org_short_name from org_tab o,hrc_tab h where o.hrc_code=h.hrc_code order by 2;
Execution Plan
----------------------------------------------------------
Plan hash value: 566430324
-------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9 | 306 | 8 (25)| 00:00:01 |
| 1 | SORT ORDER BY | | 9 | 306 | 8 (25)| 00:00:01 |
|* 2 | HASH JOIN | | 9 | 306 | 7 (15)| 00:00:01 | －－爲什麼還是hash join
| 3 | TABLE ACCESS FULL| ORG_TAB | 9 | 198 | 3 (0)| 00:00:01 |
| 4 | TABLE ACCESS FULL| HRC_TAB | 9 | 108 | 3 (0)| 00:00:01 |
-------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("O"."HRC_CODE"="H"."HRC_CODE")
因爲性能數據還是舊的。
分析：
SQL> exec dbms_stats.gather_table_stats(user,'hrc_tab',cascade=>true,estimate_percent=>100);
PL/SQL procedure successfully completed.
索引的結構沒有問題，因爲插入數據是一條條插入的。
SQL> analyze index PK_HRC_TAB validate structure;
Index analyzed.
SQL> set autotrace off
SQL> select btree_space,height,pct_used,(del_lf_rows/decode(lf_rows,0,1,lf_rows))*100||'%' as delete_pct from index_stats;
BTREE_SPACE HEIGHT PCT_USED DELETE_PCT
----------- ---------- ---------- -----------------------------------------
1503280 2 100 0%
SQL> set autotrace trace exp
SQL> exec dbms_stats.gather_table_stats(user,'org_tab',cascade=>true,estimate_percent=>100);
PL/SQL procedure successfully completed.
SQL> select h.hrc_descr,o.org_short_name from org_tab o,hrc_tab h where o.hrc_code=h.hrc_code order by 2;
Execution Plan
----------------------------------------------------------
Plan hash value: 3315346234
--------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9 | 270 | 13 (8)| 00:00:01 |
| 1 | SORT ORDER BY | | 9 | 270 | 13 (8)| 00:00:01 |
| 2 | NESTED LOOPS | | 9 | 270 | 12 (0)| 00:00:01 |
| 3 | TABLE ACCESS FULL | ORG_TAB | 9 | 198 | 3 (0)| 00:00:01 |
| 4 | TABLE ACCESS BY INDEX ROWID| HRC_TAB | 1 | 8 | 1 (0)| 00:00:01 |
|* 5 | INDEX UNIQUE SCAN | PK_HRC_TAB | 1 | | 0 (0)| 00:00:01 |
--------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
5 - access("O"."HRC_CODE"="H"."HRC_CODE")
SQL> select /*+ use_hash(o h) */ h.hrc_descr,o.org_short_name from org_tab o,hrc_tab h where o.hrc_code=h.hrc_code order by 2;
Execution Plan
----------------------------------------------------------
Plan hash value: 566430324
-------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9 | 270 | 64 (10)| 00:00:01 |
| 1 | SORT ORDER BY | | 9 | 270 | 64 (10)| 00:00:01 |
|* 2 | HASH JOIN | | 9 | 270 | 63 (8)| 00:00:01 |
| 3 | TABLE ACCESS FULL| ORG_TAB | 9 | 198 | 3 (0)| 00:00:01 |
| 4 | TABLE ACCESS FULL| HRC_TAB | 100K| 781K| 58 (6)| 00:00:01 |
-------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("O"."HRC_CODE"="H"."HRC_CODE")
此時，hash join代價比nested loop高。
二、哈西連接方式
hash算法——數據結構中講述的一種算法，需要有一個哈西函數，用哈西函數將集合的值哈西化以後，集合的重複值將會變多，叫做簇值，根據簇值來分簇。
x和y關聯
mod(x,10) --> 10個簇
mod(y,10) --> 10個簇
簇間匹配，再簇內匹配。
假設兩個數據集合：
S＝｛1，1，1，3，3，4，4，4，4，5，8，8，8，8，10｝
B＝｛0，0，1，1，2，2，3，8，9，9，9，10，11｝
hash函數還是mod(x,10),將每個子分區用hash算法分出來，匹配後再進行原值的匹配：
分區 B0 B1 B2 B3 B4 B5 B6 B7 B8 B9
值 ｛0，0，10｝ ｛1，1，11｝ ｛2，2｝ ｛3｝ null null null null ｛8｝ ｛9，9，9｝
－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－
S0｛10｝ Y
S1｛1，1，1｝ Y
S2 NULL N
S3｛3，3｝ Y
S4｛4，4，4，4｝ N
S5｛5｝ N
S6 NULL N
S7 NULL N
S8｛8，8，8，8｝ Y
S9 NULL N
-------------------------------------------------------------------------------------------------
只有4個簇可以匹配
hash join的過程：
1、存放數據到hash區域——PGA區域
將小表（B）的數據，經過hash算法得到哈西簇（bucket），將原值放到相應的哈西簇中。
2、創建位圖向量
將B表的連接字段集合讀取到hash簇，oracle記錄連接鍵的唯一值（原值），構建位圖向量，這裏的位圖向量｛0，1，2，3，8，9，10，11｝
3、超出內存大小的部分（私有PGA區域耗完）會被移動到磁盤的臨時表空間，任何需要寫入到磁盤的操作都會生成IO的損耗，會影響性能， 所以最理想的情況是在PGA中完成HASH的操作。
4、對散列簇進行排序
爲了能充分的利用內存，儘量存儲更多的散列簇，oracle按照各個散列簇的大小在內存和磁盤中排序。
5、讀取大表數據進行hash運算簇間匹配。如果一致，再進行原值的匹配返回結果。如果沒有匹配上，就將S中沒有匹配到的寫到新的散列簇中。算法是一樣的，也就是說在S中，S2，6，7，9被過濾了，同樣把B中的位圖向量進行過濾，這叫做位圖向量的過濾。
hash區域：
SQL> show parameter hash
NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
hash_area_size integer 131072
因爲PGA是自動管理方式，所以這個大小不需要手工調整的。
使用hash join的場合：
1、確認小表存在，驅動表，因爲小表是優先在hash區域構建hash簇的表
2、確認涉及到的表和表的連接鍵都被分析過了。
3、確定用hash join方式了，不需要在連接鍵上創建B樹索引。
4、如果SQL語句中，有where的過濾條件，此時這個字段不是連接字段，還是需要考慮創建B樹索引的。如果過濾條件是連接字段，也要考慮創建B樹索引，可以通過索引的搜索將參與運算結果集變小。
5、如果連接鍵值數值分佈不均勻還是要做直方圖的。
6、PGA不足的話，要調整PGA
7、hash join適合大表和超大表的連接，返回大型的結果集合。
SQL> conn scott/scott
Connected.
SQL> create table tt as select * from all_objects;
Table created.
SQL> create table tt1 as select * from all_objects;
Table created.
SQL> exec dbms_stats.gather_table_stats(user,'tt',estimate_percent=>100);
PL/SQL procedure successfully completed.
SQL> exec dbms_stats.gather_table_stats(user,'tt1',estimate_percent=>100);
PL/SQL procedure successfully completed.
SQL> set autotrace trace exp
SQL> set linesize 1000
SQL> select * from tt,tt1 where tt.object_id=tt1.object_id;
Execution Plan
----------------------------------------------------------
Plan hash value: 2918007165
-----------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time |
-----------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 40909 | 7590K| | 686 (2)| 00:00:09 |
|* 1 | HASH JOIN | | 40909 | 7590K| 4280K| 686 (2)| 00:00:09 |
| 2 | TABLE ACCESS FULL| TT1 | 40910 | 3795K| | 134 (3)| 00:00:02 |
| 3 | TABLE ACCESS FULL| TT | 40909 | 3795K| | 134 (3)| 00:00:02 |
-----------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("TT"."OBJECT_ID"="TT1"."OBJECT_ID")
SQL> create index ind_tt on tt(object_id);
Index created.
SQL> create index ind_tt1 on tt1(object_id);
Index created.
SQL> exec dbms_stats.gather_table_stats(user,'tt1',estimate_percent=>100,cascade=>true);
PL/SQL procedure successfully completed.
SQL> exec dbms_stats.gather_table_stats(user,'tt',estimate_percent=>100,cascade=>true);
PL/SQL procedure successfully completed.
SQL> select * from tt,tt1 where tt.object_id=tt1.object_id;
Execution Plan
----------------------------------------------------------
Plan hash value: 2918007165
-----------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time |
-----------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 40909 | 7590K| | 686 (2)| 00:00:09 |
|* 1 | HASH JOIN | | 40909 | 7590K| 4280K| 686 (2)| 00:00:09 |
| 2 | TABLE ACCESS FULL| TT1 | 40910 | 3795K| | 134 (3)| 00:00:02 |
| 3 | TABLE ACCESS FULL| TT | 40909 | 3795K| | 134 (3)| 00:00:02 |
-----------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("TT"."OBJECT_ID"="TT1"."OBJECT_ID")
此時加一個非連接字段的過濾條件：
SQL> select * from tt,tt1 where tt.object_id=tt1.object_id and tt.object_type='TYPE';
Execution Plan
----------------------------------------------------------
Plan hash value: 3719458023
---------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1705 | 316K| 268 (3)| 00:00:04 |
|* 1 | HASH JOIN | | 1705 | 316K| 268 (3)| 00:00:04 |
|* 2 | TABLE ACCESS FULL| TT | 1705 | 158K| 133 (2)| 00:00:02 |
| 3 | TABLE ACCESS FULL| TT1 | 40910 | 3795K| 134 (3)| 00:00:02 |
---------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("TT"."OBJECT_ID"="TT1"."OBJECT_ID")
2 - filter("TT"."OBJECT_TYPE"='TYPE')
SQL> create index ind_tt2 on tt(object_type);
Index created.
SQL> exec dbms_stats.gather_table_stats(user,'tt',estimate_percent=>100,cascade=>true);
PL/SQL procedure successfully completed.
SQL> select * from tt,tt1 where tt.object_id=tt1.object_id and tt.object_type='TYPE';
Execution Plan
----------------------------------------------------------
Plan hash value: 712899355
----------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
----------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1249 | 231K| 179 (3)| 00:00:03 |
|* 1 | HASH JOIN | | 1249 | 231K| 179 (3)| 00:00:03 |
| 2 | TABLE ACCESS BY INDEX ROWID| TT | 1249 | 115K| 44 (0)| 00:00:01 |
|* 3 | INDEX RANGE SCAN | IND_TT2 | 1249 | | 4 (0)| 00:00:01 |
| 4 | TABLE ACCESS FULL | TT1 | 40910 | 3795K| 134 (3)| 00:00:02 |
----------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("TT"."OBJECT_ID"="TT1"."OBJECT_ID")
3 - access("TT"."OBJECT_TYPE"='TYPE')
此時一個過濾條件落在了連接鍵上。
SQL> select * from tt,tt1 where tt.object_id=tt1.object_id and tt.object_id=259;
Execution Plan
----------------------------------------------------------
Plan hash value: 2300561305
-----------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-----------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 190 | 4 (0)| 00:00:01 |
| 1 | MERGE JOIN CARTESIAN | | 1 | 190 | 4 (0)| 00:00:01 |
| 2 | TABLE ACCESS BY INDEX ROWID | TT | 1 | 95 | 2 (0)| 00:00:01 |
|* 3 | INDEX RANGE SCAN | IND_TT | 1 | | 1 (0)| 00:00:01 |
| 4 | BUFFER SORT | | 1 | 95 | 2 (0)| 00:00:01 |
| 5 | TABLE ACCESS BY INDEX ROWID| TT1 | 1 | 95 | 2 (0)| 00:00:01 |
|* 6 | INDEX RANGE SCAN | IND_TT1 | 1 | | 1 (0)| 00:00:01 |
-----------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
3 - access("TT"."OBJECT_ID"=259)
6 - access("TT1"."OBJECT_ID"=259)
SQL> drop index ind_tt1;
Index dropped.
SQL> drop index ind_tt;
Index dropped.
SQL> exec dbms_stats.gather_table_stats(user,'tt',estimate_percent=>100,cascade=>true);
PL/SQL procedure successfully completed.
SQL> exec dbms_stats.gather_table_stats(user,'tt1',estimate_percent=>100,cascade=>true);
PL/SQL procedure successfully completed.
SQL> select * from tt,tt1 where tt.object_id=tt1.object_id and tt.object_id=259;
Execution Plan
----------------------------------------------------------
Plan hash value: 3719458023
---------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 190 | 267 (2)| 00:00:04 |
|* 1 | HASH JOIN | | 1 | 190 | 267 (2)| 00:00:04 |
|* 2 | TABLE ACCESS FULL| TT | 1 | 95 | 133 (2)| 00:00:02 |
|* 3 | TABLE ACCESS FULL| TT1 | 1 | 95 | 133 (2)| 00:00:02 |
---------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("TT"."OBJECT_ID"="TT1"."OBJECT_ID")
2 - filter("TT"."OBJECT_ID"=259)
3 - filter("TT1"."OBJECT_ID"=259)
三、sort merge join排序合併連接
連接過程：
1、對連接的每個表作全表掃描或者索引全掃————代價很大
2、對其中一個表或者兩個全表掃的結果排序（如果其中的一個表是全索引掃描的方式是不需要排序的，因爲索引本身是排序的）——代價最大
3、連接的時候將兩個結果集合合併 ——代價不大
SQL> select /*+ use_merge(o h) */ h.hrc_descr,o.org_short_name from org_tab o,hrc_tab h where o.hrc_code=h.hrc_code order by 2;
Execution Plan
----------------------------------------------------------
Plan hash value: 3620910538
--------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9 | 270 | 394 (3)| 00:00:05 |
| 1 | SORT ORDER BY | | 9 | 270 | 394 (3)| 00:00:05 |
| 2 | MERGE JOIN | | 9 | 270 | 393 (2)| 00:00:05 |
| 3 | TABLE ACCESS BY INDEX ROWID| HRC_TAB | 100K| 781K| 389 (2)| 00:00:05 |
| 4 | INDEX FULL SCAN | PK_HRC_TAB | 100K| | 191 (2)| 00:00:03 |
|* 5 | SORT JOIN | | 9 | 198 | 4 (25)| 00:00:01 |
| 6 | TABLE ACCESS FULL | ORG_TAB | 9 | 198 | 3 (0)| 00:00:01 |
--------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
5 - access("O"."HRC_CODE"="H"."HRC_CODE")
filter("O"."HRC_CODE"="H"."HRC_CODE")
SQL> select h.hrc_descr,o.org_short_name from org_tab o,hrc_tab h where o.hrc_code=h.hrc_code order by 2;
Execution Plan
----------------------------------------------------------
Plan hash value: 3315346234
--------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9 | 270 | 13 (8)| 00:00:01 |
| 1 | SORT ORDER BY | | 9 | 270 | 13 (8)| 00:00:01 |
| 2 | NESTED LOOPS | | 9 | 270 | 12 (0)| 00:00:01 |
| 3 | TABLE ACCESS FULL | ORG_TAB | 9 | 198 | 3 (0)| 00:00:01 |
| 4 | TABLE ACCESS BY INDEX ROWID| HRC_TAB | 1 | 8 | 1 (0)| 00:00:01 |
|* 5 | INDEX UNIQUE SCAN | PK_HRC_TAB | 1 | | 0 (0)| 00:00:01 |
--------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
5 - access("O"."HRC_CODE"="H"."HRC_CODE")
merge join使用場合：
1、9i開始，因爲其排序和掃描的成本太高，大多被hash join取代。
2、如果源表（或者索引）在存儲的時候已經排過序了，或者兩個表的連接列都有索引且都是索引全掃的過程，此時性能會高於hash join、循環嵌套，出現的機率不大。
－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－
Hint——hint是oracle中的SQL語法，允許用戶在SQL語句中插入提示（hint），影響SQL的執行方式。
誤區：
1、/*+ index(table_name index_name) */認爲這種做法一定能加快執行。
2、hint只要語法沒錯，一定會生效。
SQL> create table t as select rownum id,object_name from all_objects where rownum<100;
Table created.
SQL> create index t_ind on t(id);
Index created.
SQL> exec dbms_stats.gather_table_stats(user,'t',cascade=>true,estimate_percent=>100);
PL/SQL procedure successfully completed.
SQL> set autotrace trace exp
SQL> set linesize 1000
SQL> select * from t where id=100;
Execution Plan
----------------------------------------------------------
Plan hash value: 1376202287
-------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 18 | 2 (0)| 00:00:01 |
| 1 | TABLE ACCESS BY INDEX ROWID| T | 1 | 18 | 2 (0)| 00:00:01 |
|* 2 | INDEX RANGE SCAN | T_IND | 1 | | 1 (0)| 00:00:01 |
-------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("ID"=100)
SQL> select count(*) from t;
Execution Plan
----------------------------------------------------------
Plan hash value: 2966233522
-------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
-------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 3 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | TABLE ACCESS FULL| T | 99 | 3 (0)| 00:00:01 |
-------------------------------------------------------------------
SQL> select /*+ index(t t_ind)*/ count(*) from t;
Execution Plan
----------------------------------------------------------
Plan hash value: 2966233522
-------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
-------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 3 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | TABLE ACCESS FULL| T | 99 | 3 (0)| 00:00:01 | －－爲什麼hint沒有生效？
-------------------------------------------------------------------
被忽略的原因：count是聚合函數，聚合函數對於null值是不起作用的。如果強制走索引的話，索引中是不索引NULL值的，NULL肯定不會被計算進去的，需求是計算所有不爲null的行，算出的結果中，如果ID爲null，name也不爲空，此時走索引，忽略掉這種情況。強制走索引回導致錯誤的結果。所以被忽略。
SQL> select count(comm),count(1) from emp;
COUNT(COMM) COUNT(1)
----------- ----------
4 14
SQL> select /*+ index(t t_ind)*/ count(id) from t; －－需求就是不計算id爲null的值
Execution Plan
----------------------------------------------------------
Plan hash value: 646498162
--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 3 | 1 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | 3 | | |
| 2 | INDEX FULL SCAN| T_IND | 99 | 297 | 1 (0)| 00:00:01 |
--------------------------------------------------------------------------
如果此時索引是unique索引，會不會被忽略？
unique索引含義——不索引null值，要求不爲null的鍵值是不重複的。
SQL> drop index t_ind;
Index dropped.
SQL> create unique index t_ind on t(id);
Index created.
SQL> exec dbms_stats.gather_table_stats(user,'t',cascade=>true,estimate_percent=>100);
PL/SQL procedure successfully completed.
SQL> select /*+ index(t t_ind)*/ count(*) from t; －－走索引計算的是不爲null的值，但是需求是所有行只要有一個字段不爲null就要計算的。
Execution Plan
----------------------------------------------------------
Plan hash value: 2966233522
-------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
-------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 3 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | TABLE ACCESS FULL| T | 99 | 3 (0)| 00:00:01 | －－忽略
-------------------------------------------------------------------
如果此時的索引是主鍵的話，會不會被忽略？
SQL> drop index t_ind;
Index dropped.
SQL> alter table t add constraint t_pk primary key(id);
Table altered.
SQL> exec dbms_stats.gather_table_stats(user,'t',cascade=>true,estimate_percent=>100);
PL/SQL procedure successfully completed.
SQL> select /*+ index(t t_pk)*/ count(*) from t;
Execution Plan
----------------------------------------------------------
Plan hash value: 56794325
-----------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
-----------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 1 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | INDEX FULL SCAN| T_PK | 99 | 1 (0)| 00:00:01 |
-----------------------------------------------------------------
－－沒有被忽略，主鍵不能爲NULL的，需求所有行只要有一個字段不爲null就要計算的，主鍵不爲NULL，此時計算主鍵的個數和計算行數是一樣的。不會被忽略。
SQL> select count(*) from t;
Execution Plan
----------------------------------------------------------
Plan hash value: 56794325
-----------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
-----------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 1 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | INDEX FULL SCAN| T_PK | 99 | 1 (0)| 00:00:01 |
-----------------------------------------------------------------
SQL> select /*+ full(t) */ count(*) from t; －－全表掃的代價高
Execution Plan
----------------------------------------------------------
Plan hash value: 2966233522
-------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
-------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 3 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | TABLE ACCESS FULL| T | 99 | 3 (0)| 00:00:01 |
-------------------------------------------------------------------
第二種情況：如果表使用了別名，提示中也要用別名，否則會被忽略。
SQL> select /*+ full(t) */ count(*) from t a;
Execution Plan
----------------------------------------------------------
Plan hash value: 56794325
-----------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
-----------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 1 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | INDEX FULL SCAN| T_PK | 99 | 1 (0)| 00:00:01 |
-----------------------------------------------------------------
SQL> select /*+ full(a) */ count(*) from t a;
Execution Plan
----------------------------------------------------------
Plan hash value: 2966233522
-------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
-------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 3 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | TABLE ACCESS FULL| T | 99 | 3 (0)| 00:00:01 |
-------------------------------------------------------------------
總結：
1、如果所強制走索引可能導致錯誤的查詢結果的時候，hint會被忽略。
2、如果使用了別名，hint中沒有用，也會被忽略。