【优化】COUNT(1)、COUNT(*)、COUNT(常量)、COUNT(主键)、COUNT(ROWID)、COUNT(非空列)、COUNT(允许为空列)、COUNT(DISTINCT列名)
1.1BLOG文档结构图
1.2前言部分
1.2.1导读和注意事项
各位技术爱好者,看完本文后,你可以掌握如下的技能,也可以学到一些其它你所不知道的知识,~O(∩_∩)O~:
①COUNT(1)和COUNT(*)的区别(重点)
②10046和10053的使用
③“SELECT COUNT(列)”和“SELECT列”在选择索引方面的区别
④COUNT计数的优化
Tips:
①本文在itpub(http://blog.itpub.net/26736162)、博客园(http://www.cnblogs.com/lhrbest)和微信公众号(xiaomaimiaolhr)上有同步更新。
②文章中用到的所有代码、相关软件、相关资料及本文的pdf版本都请前往小麦苗的云盘下载,小麦苗的云盘地址见:http://blog.itpub.net/26736162/viewspace-1624453/。
③若网页文章代码格式有错乱,请下载pdf格式的文档来阅读。
④在本篇BLOG中,代码输出部分一般放在一行一列的表格中。
⑤ 本文适合于Oracle初中级人员阅读,Oracle大师请略过本文。
本文若有错误或不完善的地方请大家多多指正,您的批评指正是我写作的最大动力。
1.2.2本文简介
看了很多有关COUNT(1)和COUNT(*)的区别和效率,众说纷纭。最终还是决定自己动手实验一番。
-------------------------------------------------------------------------
第二章实验部分
2.1实验环境介绍
项目
source db
db类型
RAC
db version
11.2.0.3.0
db存储
ASM
OS版本及kernel版本
RHEL 6.5
2.2实验目标
弄清楚COUNT(1)、COUNT(*)、COUNT(常量)、COUNT(主键)、COUNT(ROWID)、COUNT(非空列)、COUNT(允许为空列)、COUNT(DISTINCT列名)之间的区别,以及它们之间的效率问题。
2.3实验过程
2.3.1实验脚本
--创建1W行的表
DROP TABLE T_COUNT_LHR;
CREATE TABLE T_COUNT_LHR AS
SELECT OBJECT_ID,
OBJECT_NAME,
OWNER,
DATA_OBJECT_ID,
OBJECT_TYPE,
LAST_DDL_TIME
FROM DBA_OBJECTS D
WHERE D.OBJECT_ID IS NOT NULL
AND D.OBJECT_NAME IS NOT NULL
AND ROWNUM <= 10000;
--更新空值,
UPDATE T_COUNT_LHR t SET t.object_type='' WHERE Rownum<=5;
UPDATE T_COUNT_LHR t SET t.LAST_DDL_TIME=T.LAST_DDL_TIME+ROWNUM;
UPDATE T_COUNT_LHR t SET t.LAST_DDL_TIME='' WHERE Rownum<=1;
COMMIT;
--添加主键、非空约束、唯一索引、普通索引
ALTER TABLE T_COUNT_LHR ADD CONSTRAINT PK_OBJECT_ID PRIMARY KEY(OBJECT_ID);
ALTER TABLE T_COUNT_LHR MODIFY OBJECT_NAME NOT NULL;
CREATE UNIQUE INDEX IDX_LDT ON T_COUNT_LHR(LAST_DDL_TIME);
CREATE INDEX IDX_DATA_OBJECT_ID ON T_COUNT_LHR(DATA_OBJECT_ID);
CREATE INDEX IDX_DATA_OWNERON T_COUNT_LHR(OWNER);
ALTER TABLE T_COUNT_LHR MODIFYOWNERNOT NULL;
--收集统计信息
EXEC dbms_stats.gather_table_stats(USER,'T_COUNT_LHR');
SELECT d.COLUMN_NAME,d.DATA_TYPE,d.NUM_NULLS,d.NUM_DISTINCT,d.LAST_ANALYZED FROM cols d WHERE d.TABLE_NAME='T_COUNT_LHR';
表的信息如下所示:
列名
是否主键
是否允许为空
是否有索引
数据类型
空值的行数
不同值的行数
总行数
OBJECT_ID
Y
N
唯一索引
NUMBER
0
10000
10000
OBJECT_NAME
N
无
VARCHAR2
0
8112
10000
OWNER
N
普通索引(IDX_OWNER)
VARCHAR2
0
5
10000
DATA_OBJECT_ID
Y
普通索引(IDX_DATA_OBJECT_ID)
NUMBER
7645
2318
10000
OBJECT_TYPE
Y
无
VARCHAR2
5
20
10000
LAST_DDL_TIME
Y
唯一索引(IDX_LDT)
DATE
1
9999
10000
需要统计如下几种情况:
SELECT COUNT(1) FROM T_COUNT_LHR;--走索引
SELECT COUNT(*) FROM T_COUNT_LHR;--走索引
SELECT COUNT(ROWID) FROM T_COUNT_LHR; --走索引
SELECT COUNT(OBJECT_ID) FROM T_COUNT_LHR; --走索引
SELECT COUNT(OBJECT_NAME) FROM T_COUNT_LHR;--走索引
SELECT COUNT(OWNER) FROM T_COUNT_LHR D;--走索引
SELECT COUNT(D.DATA_OBJECT_ID) FROM T_COUNT_LHR D; --走索引
SELECT COUNT(D.LAST_DDL_TIME) FROM T_COUNT_LHR D;--走索引
SELECT COUNT(D.LAST_DDL_TIME) FROM T_COUNT_LHR D WHERE D.LAST_DDL_TIME IS NOT NULL;--走索引
SELECT D.LAST_DDL_TIME FROM T_COUNT_LHR D; --不走索引
SELECT D.LAST_DDL_TIME FROM T_COUNT_LHR D WHERE D.LAST_DDL_TIME IS NOT NULL;--走索引
SELECT COUNT(DISTINCT DATA_OBJECT_ID) FROM T_COUNT_LHR D;--不走索引
SELECT COUNT(DISTINCT OWNER) FROM T_COUNT_LHR D;--走索引
SELECT COUNT(DISTINCT DATA_OBJECT_ID) FROM T_COUNT_LHR D WHERE DATA_OBJECT_ID IS NOT NULL ;--走索引
2.3.2执行计划
介绍
SQL命令
执行计划
返回行数
COUNT(1)和COUNT(常量)是一样的,SELECT COUNT(*) "COUNT(1)" FROM "LHR"."T_COUNT_LHR" "T_COUNT_LHR"
SELECT COUNT(1) FROM T_COUNT_LHR;
Plan hash value: 1265209789
|-----------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
|-----------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 7 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | INDEX FAST FULL SCAN| PK_OBJECT_ID | 10000 | 7 (0)| 00:00:01 |
|-----------------------------------------------------------------------------10000
SELECT COUNT(*) "COUNT(*)" FROM "LHR"."T_COUNT_LHR" "T_COUNT_LHR"
SELECT COUNT(*) FROM T_COUNT_LHR;
Plan hash value: 1265209789
|-----------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
|-----------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 7 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | INDEX FAST FULL SCAN| PK_OBJECT_ID | 10000 | 7 (0)| 00:00:01 |
|-----------------------------------------------------------------------------10000
SELECT COUNT("T_COUNT_LHR".ROWID) "COUNT(ROWID)" FROM "LHR"."T_COUNT_LHR" "T_COUNT_LHR"
SELECT COUNT(ROWID) FROM T_COUNT_LHR;
Plan hash value: 1265209789
|-------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
|-------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 12 | 7 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | 12 | | |
| 2 | INDEX FAST FULL SCAN| PK_OBJECT_ID | 10000 | 117K| 7 (0)| 00:00:01 |
|-------------------------------------------------------------------------------------10000
SELECT COUNT(*) "COUNT(OBJECT_ID)" FROM "LHR"."T_COUNT_LHR" "T_COUNT_LHR"
SELECT COUNT(OBJECT_ID) FROM T_COUNT_LHR;
Plan hash value: 1265209789
|-----------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
|-----------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 7 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | INDEX FAST FULL SCAN| PK_OBJECT_ID | 10000 | 7 (0)| 00:00:01 |
|-----------------------------------------------------------------------------10000
SELECT COUNT(*) "COUNT(OBJECT_NAME)" FROM "LHR"."T_COUNT_LHR" "T_COUNT_LHR"
SELECT COUNT(OBJECT_NAME) FROM T_COUNT_LHR;
Plan hash value: 1265209789
|-----------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
|-----------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 7 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | INDEX FAST FULL SCAN| PK_OBJECT_ID | 10000 | 7 (0)| 00:00:01 |
|-----------------------------------------------------------------------------10000
OWNER列含有索引,非空列,选择主键索引
SELECT COUNT(OWNER) FROM T_COUNT_LHR D;
Plan hash value: 1265209789
|-----------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
|-----------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 7 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | INDEX FAST FULL SCAN| PK_OBJECT_ID | 10000 | 7 (0)| 00:00:01 |
|-----------------------------------------------------------------------------10000
DATA_OBJECT_ID列允许为空,含有普通索引
SELECT COUNT(D.DATA_OBJECT_ID) FROM T_COUNT_LHR D;
Plan hash value: 2404962198
|-------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
|-------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 2 | 3 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | 2 | | |
| 2 | INDEX FAST FULL SCAN| IDX_DATA_OBJECT_ID | 10000 | 20000 | 3 (0)| 00:00:01 |
|-------------------------------------------------------------------------------------------2355
注意,COUNT(列)计算的是列值为非空的行数
SELECT COUNT(D.LAST_DDL_TIME) FROM T_COUNT_LHR D;
Plan hash value: 887614938
|--------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
|--------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 8 | 9 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | 8 | | |
| 2 | INDEX FAST FULL SCAN| IDX_LDT | 10000 | 80000 | 9 (0)| 00:00:01 |
|--------------------------------------------------------------------------------9999
SELECT COUNT(D.LAST_DDL_TIME) FROM T_COUNT_LHR D WHERE D.LAST_DDL_TIME IS NOT NULL;
Plan hash value: 887614938
|--------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
|--------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 8 | 9 (0)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | 8 | | |
|* 2 | INDEX FAST FULL SCAN| IDX_LDT | 9999 | 79992 | 9 (0)| 00:00:01 |
|--------------------------------------------------------------------------------9999
由于列中存在空值,所以不会选择索引。加上IS NOT NULL就可以选择索引了。
SELECT D.LAST_DDL_TIME FROM T_COUNT_LHR D;
Plan hash value: 2392171920
|--------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
|--------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 10000 | 80000 | 22 (0)| 00:00:01 |
| 1 | TABLE ACCESS FULL| T_COUNT_LHR | 10000 | 80000 | 22 (0)| 00:00:01 |
|--------------------------------------------------------------------------------10000
SELECT D.LAST_DDL_TIME FROM T_COUNT_LHR D WHERE D.LAST_DDL_TIME IS NOT NULL;
Plan hash value: 2419516343
|-------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
|-------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 9999 | 79992 | 9 (0)| 00:00:01 |
|* 1 | INDEX FAST FULL SCAN| IDX_LDT | 9999 | 79992 | 9 (0)| 00:00:01 |
|-------------------------------------------------------------------------------9999
DISTINCT允许为空列不会选择索引,而DISTINCT非空列会选择索引
SELECT COUNT(DISTINCT DATA_OBJECT_ID) FROM T_COUNT_LHR D;
Plan hash value: 3258478826
|-----------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
|-----------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 13 | 23 (5)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | 13 | | |
| 2 | VIEW | VW_DAG_0 | 2318 | 30134 | 23 (5)| 00:00:01 |
| 3 | HASH GROUP BY | | 2318 | 4636 | 23 (5)| 00:00:01 |
| 4 | TABLE ACCESS FULL| T_COUNT_LHR | 10000 | 20000 | 22 (0)| 00:00:01 |
|-----------------------------------------------------------------------------------2318
SELECT COUNT(DISTINCT OWNER) FROM T_COUNT_LHR D;
Plan hash value: 4008695099
|-----------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
|-----------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 17 | 9 (12)| 00:00:01 |
| 1 | SORT AGGREGATE | | 1 | 17 | | |
| 2 | VIEW | VW_DAG_0 | 5 | 85 | 9 (12)| 00:00:01 |
| 3 | HASH GROUP BY | | 5 | 30 | 9 (12)| 00:00:01 |
| 4 | INDEX FAST FULL SCAN| IDX_DATA_OWNER | 10000 | 60000 | 8 (0)| 00:00:01 |
|-----------------------------------------------------------------------------------------5
2.3.310046事件
alter session set events '10046 trace name context forever, level 12';
SELECT COUNT(*) FROM T_COUNT_LHR;
alter system flush buffer_cache;
alter system flush shared_pool;
SELECT COUNT(1) FROM T_COUNT_LHR;
alter system flush buffer_cache;
alter system flush shared_pool;
SELECT COUNT(ROWID) FROM T_COUNT_LHR;
alter session set events '10046 trace name context OFF';
select value from v$diag_info where name like '%Default%';
exit
tkprof orclasm_ora_13825.trc orclasm_ora_13825.out
alter system flush buffer_cache;
alter system flush shared_pool;
alter session set sql_trace=true;
SELECT COUNT(*) FROM T_COUNT_LHR;
alter system flush buffer_cache;
alter system flush shared_pool;
SELECT COUNT(1) FROM T_COUNT_LHR;
alter system flush buffer_cache;
alter system flush shared_pool;
SELECT COUNT(ROWID) FROM T_COUNT_LHR;
alter session set sql_trace=false;
select value from v$diag_info where name like '%Default%';
********************************************************************************
count = number of times OCI procedure was executed
cpu = cpu time in seconds executing
elapsed = elapsed time in seconds executing
disk = number of physical reads of buffers from disk
query = number of buffers gotten for consistent read
current = number of buffers gotten in current mode (usually for update)
rows = number of rows processed by the fetch or execute call
********************************************************************************
SQL ID: g9rksvy7gkdmj Plan Hash: 1265209789
SELECT COUNT(*)
FROM
T_COUNT_LHR
call count cpu elapsed disk query current rows
------- ------ -------- ---------- ---------- ---------- ---------- ----------
Parse 1 0.00 0.00 0 0 0 0
Execute 1 0.00 0.00 0 0 0 0
Fetch 2 0.00 0.55 22 25 0 1
------- ------ -------- ---------- ---------- ---------- ---------- ----------
total 4 0.01 0.56 22 25 0 1
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 90
Number of plan statistics captured: 1
Rows (1st) Rows (avg) Rows (max) Row Source Operation
---------- ---------- ---------- ---------------------------------------------------
1 1 1 SORT AGGREGATE (cr=25 pr=22 pw=0 time=552894 us)
10000 10000 10000 INDEX FAST FULL SCAN PK_OBJECT_ID (cr=25 pr=22 pw=0 time=2502 us cost=7 size=0 card=10000)(object id 159543)
********************************************************************************
SQL ID: 4abkxq9x7uamj Plan Hash: 1265209789
SELECT COUNT(1)
FROM
T_COUNT_LHR
call count cpu elapsed disk query current rows
------- ------ -------- ---------- ---------- ---------- ---------- ----------
Parse 1 0.00 0.01 0 0 0 0
Execute 1 0.00 0.00 0 0 0 0
Fetch 2 0.00 0.00 22 25 0 1
------- ------ -------- ---------- ---------- ---------- ---------- ----------
total 4 0.00 0.01 22 25 0 1
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 90
Number of plan statistics captured: 1
Rows (1st) Rows (avg) Rows (max) Row Source Operation
---------- ---------- ---------- ---------------------------------------------------
1 1 1 SORT AGGREGATE (cr=25 pr=22 pw=0 time=5934 us)
10000 10000 10000 INDEX FAST FULL SCAN PK_OBJECT_ID (cr=25 pr=22 pw=0 time=2469 us cost=7 size=0 card=10000)(object id 159543)
********************************************************************************
SQL ID: 4tm3jwzff8ub8 Plan Hash: 1265209789
SELECT COUNT(ROWID)
FROM
T_COUNT_LHR
call count cpu elapsed disk query current rows
------- ------ -------- ---------- ---------- ---------- ---------- ----------
Parse 1 0.00 0.00 0 0 0 0
Execute 1 0.00 0.00 0 0 0 0
Fetch 2 0.00 0.01 22 25 0 1
------- ------ -------- ---------- ---------- ---------- ---------- ----------
total 4 0.00 0.01 22 25 0 1
Misses in library cache during parse: 1
Optimizer mode: ALL_ROWS
Parsing user id: 90
Number of plan statistics captured: 1
Rows (1st) Rows (avg) Rows (max) Row Source Operation
---------- ---------- ---------- ---------------------------------------------------
1 1 1 SORT AGGREGATE (cr=25 pr=22 pw=0 time=11264 us)
10000 10000 10000 INDEX FAST FULL SCAN PK_OBJECT_ID (cr=25 pr=22 pw=0 time=762 us cost=7 size=120000 card=10000)(object id 159543)
********************************************************************************
2.3.410053事件
alter session set events '10053 trace name context forever, level 12';
alter session set events '10053 trace name context OFF';
alter system flush buffer_cache;
alter system flush shared_pool;
alter session set events '10053 trace name context forever, level 12';
SELECT COUNT(*) FROM T_COUNT_LHR;
alter system flush buffer_cache;
alter system flush shared_pool;
SELECT COUNT(1) FROM T_COUNT_LHR;
alter system flush buffer_cache;
alter system flush shared_pool;
SELECT COUNT(ROWID) FROM T_COUNT_LHR;
alter session set events '10053 trace name context OFF';
select value from v$diag_info where name like '%Default%';
搜索关键词“Final”、“Starting”:
Finalquery after transformations:******* UNPARSED QUERY IS *******
SELECT COUNT(*) "COUNT(*)" FROM "LHR"."T_COUNT_LHR" "T_COUNT_LHR"
kkoqbc: optimizing query block SEL$1 (#0)
:
call(in-use=1216, alloc=16344), compile(in-use=58000, alloc=58544), execution(in-use=2480, alloc=4032)
kkoqbc-subheap (create addr=0x7f9950a9fb58)
****************
QUERY BLOCK TEXT
****************
SELECT COUNT(*) FROM T_COUNT_LHR
---------------------
QUERY BLOCK SIGNATURE
---------------------
signature (optimizer): qb_name=SEL$1 nbfros=1 flg=0
fro(0): flg=0 objn=159542 hint_alias="T_COUNT_LHR"@"SEL$1"
-----------------------------
SYSTEM STATISTICS INFORMATION
-----------------------------
Using NOWORKLOAD Stats
CPUSPEEDNW: 1752 millions instructions/sec (default is 100)
IOTFRSPEED: 4096 bytes per millisecond (default is 4096)
IOSEEKTIM: 10 milliseconds (default is 10)
MBRC: NO VALUE blocks (default is 8)
***************************************
BASE STATISTICAL INFORMATION
***********************
Table Stats::
Table: T_COUNT_LHR Alias: T_COUNT_LHR
#Rows: 10000 #Blks: 75 AvgRowLen: 44.00 ChainCnt: 0.00
Index Stats::
Index: IDX_DATA_OBJECT_ID Col#: 4
LVLS: 1 #LB: 5 #DK: 2318 LB/K: 1.00 DB/K: 1.00 CLUF: 86.00
Index: IDX_DATA_OWNER Col#: 3
LVLS: 1 #LB: 23 #DK: 5 LB/K: 4.00 DB/K: 28.00 CLUF: 141.00
Index: IDX_LDT Col#: 6
LVLS: 1 #LB: 26 #DK: 9999 LB/K: 1.00 DB/K: 1.00 CLUF: 377.00
Index: PK_OBJECT_ID Col#: 1
LVLS: 1 #LB: 20 #DK: 10000 LB/K: 1.00 DB/K: 1.00 CLUF: 76.00
Access path analysis for T_COUNT_LHR
***************************************
SINGLE TABLE ACCESS PATH
Single Table Cardinality Estimation for T_COUNT_LHR[T_COUNT_LHR]
Table: T_COUNT_LHR Alias: T_COUNT_LHR
Card: Original: 10000.000000 Rounded: 10000 Computed: 10000.00 Non Adjusted: 10000.00
Access Path: TableScan
Cost: 22.10 Resp: 22.10 Degree: 0
Cost_io: 22.00 Cost_cpu: 2034108
Resp_io: 22.00 Resp_cpu: 2034108
Access Path: index (index (FFS))
Index: IDX_DATA_OWNER
resc_io: 8.00 resc_cpu: 1363793
ix_sel: 0.000000 ix_sel_with_filters: 1.000000
Access Path: index (FFS)
Cost: 8.06 Resp: 8.06 Degree: 1
Cost_io: 8.00 Cost_cpu: 1363793
Resp_io: 8.00 Resp_cpu: 1363793
Access Path: index (index (FFS))
Index: PK_OBJECT_ID
resc_io: 7.00 resc_cpu: 1342429
ix_sel: 0.000000 ix_sel_with_filters: 1.000000
Access Path: index (FFS)
Cost: 7.06 Resp: 7.06 Degree: 1
Cost_io: 7.00 Cost_cpu: 1342429
Resp_io: 7.00 Resp_cpu: 1342429
Access Path: index (FullScan)
Index: IDX_DATA_OWNER
resc_io: 24.00 resc_cpu: 2170915
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 24.10 Resp: 24.10 Degree: 1
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 1
****** trying bitmap/domain indexes ******
Access Path: index (FullScan)
Index: IDX_DATA_OWNER
resc_io: 24.00 resc_cpu: 2170915
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 24.10 Resp: 24.10 Degree: 0
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 0
****** finished trying bitmap/domain indexes ******
******** Begin index join costing ********
****** trying bitmap/domain indexes ******
Access Path: index (FullScan)
Index: IDX_DATA_OWNER
resc_io: 24.00 resc_cpu: 2170915
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 24.10 Resp: 24.10 Degree: 0
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 0
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 0
Bitmap nodes:
Used PK_OBJECT_ID
Cost = 26.377821, sel = 1.000000
****** finished trying bitmap/domain indexes ******
******** End index join costing ********
Best:: AccessPath: IndexFFS
Index: PK_OBJECT_ID
Cost: 7.06 Degree: 1 Resp: 7.06 Card: 10000.00 Bytes: 0
***************************************
。。。。。。。。。。。。。。
Starting SQL statement dump
user_id=90 user_name=LHR module=SQL*Plus action=
sql_id=g9rksvy7gkdmj plan_hash_value=1265209789 problem_type=3
----- Current SQL Statement for this session (sql_id=g9rksvy7gkdmj) -----
SELECT COUNT(*) FROM T_COUNT_LHR
sql_text_length=33
sql=SELECT COUNT(*) FROM T_COUNT_LHR
----- Explain Plan Dump -----
----- Plan Table -----
============
Plan Table
============
---------------------------------------------+-----------------------------------+
| Id | Operation | Name | Rows | Bytes | Cost | Time |
---------------------------------------------+-----------------------------------+
| 0 | SELECT STATEMENT | | | | 7 | |
| 1 | SORT AGGREGATE | | 1 | | | |
| 2 | INDEX FAST FULL SCAN | PK_OBJECT_ID| 10K | | 7 | 00:00:01 |
---------------------------------------------+-----------------------------------+
Predicate Information:
----------------------
Content of other_xml column
===========================
db_version : 11.2.0.3
parse_schema : LHR
plan_hash : 1265209789
plan_hash_2 : 3881728982
Outline Data:
/*+
BEGIN_OUTLINE_DATA
IGNORE_OPTIM_EMBEDDED_HINTS
OPTIMIZER_FEATURES_ENABLE('11.2.0.3')
DB_VERSION('11.2.0.3')
OPT_PARAM('star_transformation_enabled' 'true')
ALL_ROWS
OUTLINE_LEAF(@"SEL$1")
INDEX_FFS(@"SEL$1" "T_COUNT_LHR"@"SEL$1" ("T_COUNT_LHR"."OBJECT_ID"))
END_OUTLINE_DATA
*/
。。。。。。。。。。。。。。。。
====================== END SQL Statement Dump ======================
Final query after transformations:******* UNPARSED QUERY IS *******
SELECT COUNT(*) "COUNT(1)" FROM "LHR"."T_COUNT_LHR" "T_COUNT_LHR"
kkoqbc: optimizing query block SEL$1 (#0)
:
call(in-use=1240, alloc=16344), compile(in-use=58312, alloc=58544), execution(in-use=2480, alloc=4032)
kkoqbc-subheap (create addr=0x7f9950a9fb58)
****************
QUERY BLOCK TEXT
****************
SELECT COUNT(1) FROM T_COUNT_LHR
---------------------
QUERY BLOCK SIGNATURE
---------------------
signature (optimizer): qb_name=SEL$1 nbfros=1 flg=0
fro(0): flg=0 objn=159542 hint_alias="T_COUNT_LHR"@"SEL$1"
-----------------------------
SYSTEM STATISTICS INFORMATION
-----------------------------
Using NOWORKLOAD Stats
CPUSPEEDNW: 1752 millions instructions/sec (default is 100)
IOTFRSPEED: 4096 bytes per millisecond (default is 4096)
IOSEEKTIM: 10 milliseconds (default is 10)
MBRC: NO VALUE blocks (default is 8)
***************************************
BASE STATISTICAL INFORMATION
***********************
Table Stats::
Table: T_COUNT_LHR Alias: T_COUNT_LHR
#Rows: 10000 #Blks: 75 AvgRowLen: 44.00 ChainCnt: 0.00
Index Stats::
Index: IDX_DATA_OBJECT_ID Col#: 4
LVLS: 1 #LB: 5 #DK: 2318 LB/K: 1.00 DB/K: 1.00 CLUF: 86.00
Index: IDX_DATA_OWNER Col#: 3
LVLS: 1 #LB: 23 #DK: 5 LB/K: 4.00 DB/K: 28.00 CLUF: 141.00
Index: IDX_LDT Col#: 6
LVLS: 1 #LB: 26 #DK: 9999 LB/K: 1.00 DB/K: 1.00 CLUF: 377.00
Index: PK_OBJECT_ID Col#: 1
LVLS: 1 #LB: 20 #DK: 10000 LB/K: 1.00 DB/K: 1.00 CLUF: 76.00
Access path analysis for T_COUNT_LHR
***************************************
SINGLE TABLE ACCESS PATH
Single Table Cardinality Estimation for T_COUNT_LHR[T_COUNT_LHR]
Table: T_COUNT_LHR Alias: T_COUNT_LHR
Card: Original: 10000.000000 Rounded: 10000 Computed: 10000.00 Non Adjusted: 10000.00
Access Path: TableScan
Cost: 22.10 Resp: 22.10 Degree: 0
Cost_io: 22.00 Cost_cpu: 2034108
Resp_io: 22.00 Resp_cpu: 2034108
Access Path: index (index (FFS))
Index: IDX_DATA_OWNER
resc_io: 8.00 resc_cpu: 1363793
ix_sel: 0.000000 ix_sel_with_filters: 1.000000
Access Path: index (FFS)
Cost: 8.06 Resp: 8.06 Degree: 1
Cost_io: 8.00 Cost_cpu: 1363793
Resp_io: 8.00 Resp_cpu: 1363793
Access Path: index (index (FFS))
Index: PK_OBJECT_ID
resc_io: 7.00 resc_cpu: 1342429
ix_sel: 0.000000 ix_sel_with_filters: 1.000000
Access Path: index (FFS)
Cost: 7.06 Resp: 7.06 Degree: 1
Cost_io: 7.00 Cost_cpu: 1342429
Resp_io: 7.00 Resp_cpu: 1342429
Access Path: index (FullScan)
Index: IDX_DATA_OWNER
resc_io: 24.00 resc_cpu: 2170915
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 24.10 Resp: 24.10 Degree: 1
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 1
****** trying bitmap/domain indexes ******
Access Path: index (FullScan)
Index: IDX_DATA_OWNER
resc_io: 24.00 resc_cpu: 2170915
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 24.10 Resp: 24.10 Degree: 0
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 0
****** finished trying bitmap/domain indexes ******
******** Begin index join costing ********
****** trying bitmap/domain indexes ******
Access Path: index (FullScan)
Index: IDX_DATA_OWNER
resc_io: 24.00 resc_cpu: 2170915
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 24.10 Resp: 24.10 Degree: 0
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 0
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 0
Bitmap nodes:
Used PK_OBJECT_ID
Cost = 26.377821, sel = 1.000000
****** finished trying bitmap/domain indexes ******
******** End index join costing ********
Best:: AccessPath: IndexFFS
Index: PK_OBJECT_ID
Cost: 7.06 Degree: 1 Resp: 7.06 Card: 10000.00 Bytes: 0
***************************************
。。。。。。。。。。。。。。。。。
Starting SQL statement dump
user_id=90 user_name=LHR module=SQL*Plus action=
sql_id=4abkxq9x7uamj plan_hash_value=1265209789 problem_type=3
----- Current SQL Statement for this session (sql_id=4abkxq9x7uamj) -----
SELECT COUNT(1) FROM T_COUNT_LHR
sql_text_length=33
sql=SELECT COUNT(1) FROM T_COUNT_LHR
----- Explain Plan Dump -----
----- Plan Table -----
============
Plan Table
============
---------------------------------------------+-----------------------------------+
| Id | Operation | Name | Rows | Bytes | Cost | Time |
---------------------------------------------+-----------------------------------+
| 0 | SELECT STATEMENT | | | | 7 | |
| 1 | SORT AGGREGATE | | 1 | | | |
| 2 | INDEX FAST FULL SCAN | PK_OBJECT_ID| 10K | | 7 | 00:00:01 |
---------------------------------------------+-----------------------------------+
Predicate Information:
----------------------
Content of other_xml column
===========================
db_version : 11.2.0.3
parse_schema : LHR
plan_hash : 1265209789
plan_hash_2 : 3881728982
Outline Data:
/*+
BEGIN_OUTLINE_DATA
IGNORE_OPTIM_EMBEDDED_HINTS
OPTIMIZER_FEATURES_ENABLE('11.2.0.3')
DB_VERSION('11.2.0.3')
OPT_PARAM('star_transformation_enabled' 'true')
ALL_ROWS
OUTLINE_LEAF(@"SEL$1")
INDEX_FFS(@"SEL$1" "T_COUNT_LHR"@"SEL$1" ("T_COUNT_LHR"."OBJECT_ID"))
END_OUTLINE_DATA
*/
====================== END SQL Statement Dump ======================
******************************************
----- Current SQL Statement for this session (sql_id=4tm3jwzff8ub8) -----
SELECT COUNT(ROWID) FROM T_COUNT_LHR
*******************************************
。。。。。。。。。
Final query after transformations:******* UNPARSED QUERY IS *******
SELECT COUNT("T_COUNT_LHR".ROWID) "COUNT(ROWID)" FROM "LHR"."T_COUNT_LHR" "T_COUNT_LHR"
kkoqbc: optimizing query block SEL$1 (#0)
:
call(in-use=1240, alloc=16344), compile(in-use=58224, alloc=62688), execution(in-use=2480, alloc=4032)
kkoqbc-subheap (create addr=0x7f9950a9fb58)
****************
QUERY BLOCK TEXT
****************
SELECT COUNT(ROWID) FROM T_COUNT_LHR
---------------------
QUERY BLOCK SIGNATURE
---------------------
signature (optimizer): qb_name=SEL$1 nbfros=1 flg=0
fro(0): flg=0 objn=159542 hint_alias="T_COUNT_LHR"@"SEL$1"
-----------------------------
SYSTEM STATISTICS INFORMATION
-----------------------------
Using NOWORKLOAD Stats
CPUSPEEDNW: 1752 millions instructions/sec (default is 100)
IOTFRSPEED: 4096 bytes per millisecond (default is 4096)
IOSEEKTIM: 10 milliseconds (default is 10)
MBRC: NO VALUE blocks (default is 8)
***************************************
BASE STATISTICAL INFORMATION
***********************
Table Stats::
Table: T_COUNT_LHR Alias: T_COUNT_LHR
#Rows: 10000 #Blks: 75 AvgRowLen: 44.00 ChainCnt: 0.00
Index Stats::
Index: IDX_DATA_OBJECT_ID Col#: 4
LVLS: 1 #LB: 5 #DK: 2318 LB/K: 1.00 DB/K: 1.00 CLUF: 86.00
Index: IDX_DATA_OWNER Col#: 3
LVLS: 1 #LB: 23 #DK: 5 LB/K: 4.00 DB/K: 28.00 CLUF: 141.00
Index: IDX_LDT Col#: 6
LVLS: 1 #LB: 26 #DK: 9999 LB/K: 1.00 DB/K: 1.00 CLUF: 377.00
Index: PK_OBJECT_ID Col#: 1
LVLS: 1 #LB: 20 #DK: 10000 LB/K: 1.00 DB/K: 1.00 CLUF: 76.00
Access path analysis for T_COUNT_LHR
***************************************
SINGLE TABLE ACCESS PATH
Single Table Cardinality Estimation for T_COUNT_LHR[T_COUNT_LHR]
Table: T_COUNT_LHR Alias: T_COUNT_LHR
Card: Original: 10000.000000 Rounded: 10000 Computed: 10000.00 Non Adjusted: 10000.00
Access Path: TableScan
Cost: 22.10 Resp: 22.10 Degree: 0
Cost_io: 22.00 Cost_cpu: 2034108
Resp_io: 22.00 Resp_cpu: 2034108
Access Path: index (index (FFS))
Index: IDX_DATA_OWNER
resc_io: 8.00 resc_cpu: 1363793
ix_sel: 0.000000 ix_sel_with_filters: 1.000000
Access Path: index (FFS)
Cost: 8.06 Resp: 8.06 Degree: 1
Cost_io: 8.00 Cost_cpu: 1363793
Resp_io: 8.00 Resp_cpu: 1363793
Access Path: index (index (FFS))
Index: PK_OBJECT_ID
resc_io: 7.00 resc_cpu: 1342429
ix_sel: 0.000000 ix_sel_with_filters: 1.000000
Access Path: index (FFS)
Cost: 7.06 Resp: 7.06 Degree: 1
Cost_io: 7.00 Cost_cpu: 1342429
Resp_io: 7.00 Resp_cpu: 1342429
Access Path: index (FullScan)
Index: IDX_DATA_OWNER
resc_io: 24.00 resc_cpu: 2170915
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 24.10 Resp: 24.10 Degree: 1
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 1
****** trying bitmap/domain indexes ******
Access Path: index (FullScan)
Index: IDX_DATA_OWNER
resc_io: 24.00 resc_cpu: 2170915
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 24.10 Resp: 24.10 Degree: 0
Access Path: index (FullScan)
Index: PK_OBJECT_ID
resc_io: 21.00 resc_cpu: 2149550
ix_sel: 1.000000 ix_sel_with_filters: 1.000000
Cost: 21.10 Resp: 21.10 Degree: 0
****** finished trying bitmap/domain indexes ******
Best:: AccessPath: IndexFFS
Index: PK_OBJECT_ID
Cost: 7.06 Degree: 1 Resp: 7.06 Card: 10000.00 Bytes: 0
***************************************
。。。。。。。。。。。。。。。。。
Starting SQL statement dump
user_id=90 user_name=LHR module=SQL*Plus action=
sql_id=4tm3jwzff8ub8 plan_hash_value=1265209789 problem_type=3
----- Current SQL Statement for this session (sql_id=4tm3jwzff8ub8) -----
SELECT COUNT(ROWID) FROM T_COUNT_LHR
sql_text_length=37
sql=SELECT COUNT(ROWID) FROM T_COUNT_LHR
----- Explain Plan Dump -----
----- Plan Table -----
============
Plan Table
============
---------------------------------------------+-----------------------------------+
| Id | Operation | Name | Rows | Bytes | Cost | Time |
---------------------------------------------+-----------------------------------+
| 0 | SELECT STATEMENT | | | | 7 | |
| 1 | SORT AGGREGATE | | 1 | 12 | | |
| 2 | INDEX FAST FULL SCAN | PK_OBJECT_ID| 10K | 117K | 7 | 00:00:01 |
---------------------------------------------+-----------------------------------+
Predicate Information:
----------------------
Content of other_xml column
===========================
db_version : 11.2.0.3
parse_schema : LHR
plan_hash : 1265209789
plan_hash_2 : 3881728982
Outline Data:
/*+
BEGIN_OUTLINE_DATA
IGNORE_OPTIM_EMBEDDED_HINTS
OPTIMIZER_FEATURES_ENABLE('11.2.0.3')
DB_VERSION('11.2.0.3')
OPT_PARAM('star_transformation_enabled' 'true')
ALL_ROWS
OUTLINE_LEAF(@"SEL$1")
INDEX_FFS(@"SEL$1" "T_COUNT_LHR"@"SEL$1" ("T_COUNT_LHR"."OBJECT_ID"))
END_OUTLINE_DATA
*/
。。。。。。。。。。。。
====================== END SQL Statement Dump ======================
2.3.5COUNT(COL1)和SELECT COL1的区别
其实在2014年的时候,小麦苗发布过一篇博客(http://blog.itpub.net/26736162/viewspace-1329880/),里边对这个问题有详细的实验。今天就把这个实验搬过来吧。
drop table t purge;
Create Table t nologging As select * from dba_objects d ;
create index ind_objectname on t(object_name);
set autotrace traceonly;
select t.object_name from t where t.object_name ='T'; --走索引
select t.object_name from t where UPPER(t.object_name) ='T'; --不走索引
select t.object_name from t where UPPER(t.object_name) ='T' and t.object_name IS NOT NULL ; --走索引(INDEX FAST FULL SCAN)
select t.object_name from t where UPPER(t.object_name) ||'AAA' ='T'||'AAA' and t.object_name IS NOT NULL ; --走索引(INDEX FAST FULL SCAN)
select t.object_name,t.owner from t where UPPER(t.object_name) ||'AAA' ='T'||'AAA' and t.object_name IS NOT NULL ; --不走索引
C:\Users\xiaomaimiao>sqlplus lhr/lhr@orclasm
SQL*Plus: Release 11.2.0.1.0 Production on 星期三 11月 12 10:52:29 2014
Copyright (c) 1982, 2010, Oracle. All rights reserved.
连接到:
Oracle Database 11g Enterprise Edition Release11.2.0.3.0 - 64bitProduction
With the Partitioning, Automatic Storage Management, OLAP, Data Mining
and Real Application Testing options
SQL>
SQL>
SQL> drop table t purge;
表已删除。
SQL> Create Table t nologging As select * from dba_objects d ;
表已创建。
SQL> create index ind_objectname on t(object_name);
索引已创建。
---- t表所有列均可以为空
SQL> desc t
NameNull? Type
----------------------------------------- -------- ----------------------------
OWNERVARCHAR2(30)
OBJECT_NAMEVARCHAR2(128)
SUBOBJECT_NAMEVARCHAR2(30)
OBJECT_IDNUMBER
DATA_OBJECT_IDNUMBER
OBJECT_TYPEVARCHAR2(19)
CREATEDDATE
LAST_DDL_TIMEDATE
TIMESTAMPVARCHAR2(19)
STATUSVARCHAR2(7)
TEMPORARYVARCHAR2(1)
GENERATEDVARCHAR2(1)
SECONDARYVARCHAR2(1)
NAMESPACENUMBER
EDITION_NAMEVARCHAR2(30)
SQL>
SQL> set autotrace traceonly;
SQL>select t.object_name from t where t.object_name ='T';
执行计划
----------------------------------------------------------
Plan hash value: 4280870634
-----------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-----------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 66 | 3 (0)| 00:00:01 |
|* 1 |INDEX RANGE SCAN| IND_OBJECTNAME | 1 | 66 | 3 (0)| 00:00:01 |
-----------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - access("T"."OBJECT_NAME"='T')
Note
-----
- dynamic sampling used for this statement (level=2)
- SQL plan baseline "SQL_PLAN_503ygb00mbj6k165e82cd" used for this statement
统计信息
----------------------------------------------------------
34 recursive calls
43 db block gets
127 consistent gets
398 physical reads
15476 redo size
349 bytes sent via SQL*Net to client
359 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
SQL>select t.object_name from t where UPPER(t.object_name) ='T';
执行计划
----------------------------------------------------------
Plan hash value: 1601196873
--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 12 | 792 | 305 (1)| 00:00:04 |
|* 1 |TABLE ACCESS FULL| T | 12 | 792 | 305 (1)| 00:00:04 |
--------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(UPPER("T"."OBJECT_NAME")='T')
Note
-----
- dynamic sampling used for this statement (level=2)
- SQL plan baseline "SQL_PLAN_9p76pys5gdb2b94ecae5c" used for this statement
统计信息
----------------------------------------------------------
29 recursive calls
43 db block gets
1209 consistent gets
1092 physical reads
15484 redo size
349 bytes sent via SQL*Net to client
359 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
SQL>select t.object_name from t where UPPER(t.object_name) ='T' and t.object_name IS NOT NULL ;
执行计划
----------------------------------------------------------
Plan hash value: 3379870158
---------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 51 | 3366 | 110 (1)| 00:00:02 |
|* 1 |INDEX FAST FULL SCAN| IND_OBJECTNAME | 51 | 3366 | 110 (1)| 00:00:02 |
---------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("T"."OBJECT_NAME" IS NOT NULL AND UPPER("T"."OBJECT_NAME")='T')
Note
-----
- dynamic sampling used for this statement (level=2)
- SQL plan baseline "SQL_PLAN_czkarb71kthws18b0c28f" used for this statement
统计信息
----------------------------------------------------------
29 recursive calls
43 db block gets
505 consistent gets
384 physical reads
15612 redo size
349 bytes sent via SQL*Net to client
359 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
SQL>select t.object_name,t.owner from t where UPPER(t.object_name) ||'AAA' ='T'||'AAA' and t.object_name IS NOT NULL ;
执行计划
----------------------------------------------------------
Plan hash value: 1601196873
--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 51 | 4233 | 304 (1)| 00:00:04 |
|* 1 |TABLE ACCESS FULL| T | 51 | 4233 | 304 (1)| 00:00:04 |
--------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("T"."OBJECT_NAME" IS NOT NULL AND
UPPER("T"."OBJECT_NAME")||'AAA'='TAAA')
Note
-----
- dynamic sampling used for this statement (level=2)
- SQL plan baseline "SQL_PLAN_au9a1c4hwdtb894ecae5c" used for this statement
统计信息
----------------------------------------------------------
30 recursive calls
44 db block gets
1210 consistent gets
1091 physical reads
15748 redo size
408 bytes sent via SQL*Net to client
359 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
SQL>select t.object_name from t where UPPER(t.object_name) ||'AAA' ='T'||'AAA' and t.object_name IS NOT NULL ;
执行计划
----------------------------------------------------------
Plan hash value: 3379870158
---------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
---------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 51 | 3366 | 110 (1)| 00:00:02 |
|* 1 |INDEX FAST FULL SCAN| IND_OBJECTNAME | 51 | 3366 | 110 (1)| 00:00:02 |
---------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("T"."OBJECT_NAME" IS NOT NULL AND
UPPER("T"."OBJECT_NAME")||'AAA'='TAAA')
Note
-----
- dynamic sampling used for this statement (level=2)
- SQL plan baseline "SQL_PLAN_1gu36rnh3s2a318b0c28f" used for this statement
统计信息
----------------------------------------------------------
28 recursive calls
44 db block gets
505 consistent gets
6 physical reads
15544 redo size
349 bytes sent via SQL*Net to client
359 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
SQL>
2.4实验结论
COUNT()函数是Oracle中的聚合函数,用于统计结果集的行数。其语法形式如下所示:
COUNT({ * | [ DISTINCT | ALL ] expr }) [ OVER (analytic_clause) ]
COUNT returns the number of rows returned by the query. You can use it as an aggregate or analytic function.
If you specify DISTINCT, then you can specify only the query_partition_clause of the analytic_clause. The order_by_clause and windowing_clause are not allowed.
If you specify expr, then COUNT returns the number of rows where expr is not null.You can count either all rows, or only distinct values of expr.
If you specify the asterisk (*), then this function returns all rows, including duplicates and nulls.COUNT never returns null.
我们把COUNT的使用情况分为以下3类:
①COUNT(1)、COUNT(*)、COUNT(常量)、COUNT(主键)、COUNT(ROWID)、COUNT(非空列)
②COUNT(允许为空列)
③COUNT(DISTINCT列名)
下面分别从查询结果和效率方面做个比较:
(一)结果区别
①COUNT(1)、COUNT(*)、COUNT(ROWID)、COUNT(常量)、COUNT(主键)、COUNT(非空列)这几种方式统计的行数是表中所有存在的行的总数,包括值为NULL的行和非空行。所以,这几种方式的执行结果相同。这里的常量可以为数字或字符串,例如,COUNT(2)、COUNT(333)、COUNT('x')、COUNT('xiaomaimiao')。需要注意的是:这里的COUNT(1)中的“1”并不表示表中的第一列,它其实是一个表达式,可以换成任意数字或字符或表达式。
②COUNT(允许为空列)这种方式统计的行数不会包括字段值为NULL的行。
③COUNT(DISTINCT列名)得到的结果是除去值为NULL和重复数据后的结果。
④“SELECT COUNT(''),COUNT(NULL) FROM T_COUNT_LHR;”返回0行。
(二)效率、索引
①如果存在主键或非空列上的索引,那么COUNT(1)、COUNT(*)、COUNT(ROWID)、COUNT(常量)、COUNT(主键)、COUNT(非空列)会首先选择主键上的索引快速全扫描(INDEX FAST FULL SCAN)。若主键不存在则会选择非空列上的索引。若非空列上没有索引则肯定走全表扫描(TABLE ACCESS FULL)。其中,COUNT(ROWID)在走索引的时候比其它几种方式要慢。通过10053事件可以看到这几种方式除了COUNT(ROWID)之外,其它最终都会转换成COUNT(*)的方式来执行。
②对于COUNT(COL1)来说,只要列字段上有索引则会选择索引快速全扫描(INDEX FAST FULL SCAN)。而对于“SELECT COL1”来说,除非列上有NOT NULL约束,否则执行计划会选择全表扫描。
③COUNT(DISTINCT列名)若列上有索引,且有非空约束或在WHERE子句中使用IS NOT NULL,则会选择索引快速全扫描。其余情况选择全表扫描。
2.5关于COUNT的优化
统计记录条数,如何才能最快?xb_audit_ddl_lhr表有2303262记录。
(一)全表扫描:
LHR@orclasm > set line 9999
LHR@orclasm > set autot on
LHR@orclasm > set timing on
LHR@orclasm > set time on
12:34:01 LHR@orclasm >
12:35:20 LHR@orclasm > SELECT /*+full(t)*/ COUNT(*) FROM xb_audit_ddl_lhr t;
COUNT(*)
----------
2303262
Elapsed:00:01:16.53
Execution Plan
----------------------------------------------------------
Plan hash value: 3725780224
-------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time | Pstart| Pstop |
-------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 26629 (1)| 00:05:20 | | |
| 1 | SORT AGGREGATE | | 1 | | | | |
| 2 | PARTITION RANGE ALL| | 2247K| 26629 (1)| 00:05:20 | 1 |1048575|
| 3 | PARTITION HASH ALL| | 2247K| 26629 (1)| 00:05:20 | 1 | 5 |
| 4 | TABLE ACCESS FULL| XB_AUDIT_DDL_LHR | 2247K| 26629 (1)| 00:05:20 | 1 |1048575|
-------------------------------------------------------------------------------------------------
Statistics
----------------------------------------------------------
1 recursive calls
57 db block gets
92274 consistent gets
92012 physical reads
0 redo size
529 bytes sent via SQL*Net to client
519 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
12:36:47 LHR@orclasm >
(二)主键索引
12:38:31 LHR@orclasm > SELECT /*+index(SYS_C0011358)*/ COUNT(*) FROM xb_audit_ddl_lhr t;
COUNT(*)
----------
2303262
Elapsed:00:00:00.24
Execution Plan
----------------------------------------------------------
Plan hash value: 2548021478
------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time |
------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 3114 (1)| 00:00:38 |
| 1 | SORT AGGREGATE | | 1 | | |
| 2 | INDEX FAST FULL SCAN| SYS_C0011358 | 2247K| 3114 (1)| 00:00:38 |
------------------------------------------------------------------------------
Statistics
----------------------------------------------------------
0 recursive calls
2 db block gets
12283 consistent gets
12220 physical reads
0 redo size
529 bytes sent via SQL*Net to client
519 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
(三)位图索引
CREATE BITMAP INDEX IND_xb_operation ON xb_audit_ddl_lhr(operation) local;
12:41:44 LHR@orclasm > SELECT /*+index(IND_xb_operation)*/ COUNT(*) FROM xb_audit_ddl_lhr t;
COUNT(*)
----------
2303262
Elapsed: 00:00:00.00
Execution Plan
----------------------------------------------------------
Plan hash value: 1101410508
------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Cost (%CPU)| Time | Pstart| Pstop |
------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 114 (0)| 00:00:02 | | |
| 1 | SORT AGGREGATE | | 1 | | | | |
| 2 | PARTITION RANGE ALL | | 2247K| 114 (0)| 00:00:02 | 1 |1048575|
| 3 | PARTITION HASH ALL | | 2247K| 114 (0)| 00:00:02 | 1 | 5 |
| 4 | BITMAP CONVERSION COUNT | | 2247K| 114 (0)| 00:00:02 | | |
| 5 | BITMAP INDEX FAST FULL SCAN| IND_XB_OPERATION | | | | 1 |1048575|
------------------------------------------------------------------------------------------------------------
Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
312 consistent gets
0 physical reads
0 redo size
529 bytes sent via SQL*Net to client
519 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
1 rows processed
在数据量超大的表上,还可以采用位图索引并行的方式。优化无止境,根据场景选择最适合的才是最好的。
第三章参考文章
3.1MOS
3.1.1How the Oracle CBO Chooses a Path for the SELECT COUNT(*) Command (文档ID 124717.1)
Purpose:========The purpose of this article is to explain how the Oracle Cost Based Optimizer (CBO) evaluates the best path to do a SELECT COUNT(*) or aSELECT COLUMN(column) without using a WHERE clause. Scope & Application:====================This article is meant for all DBAs and anyone involved in SQL Tuning.How the Oracle CBO Chooses a Path for the SELECT COUNT(*) Command:==================================================================CBO is invoked when: o OPTIMIZER_MODE = CHOOSE, or OPTIMIZER_GOAL = CHOOSE--AND-- o At least one of the objects of the select statement has been analyzed.Here some examples of select count() without any where clause:Sel1 : Select count(*) from journal_entries;Sel2 : Select count(1) from journal_entries;Sel3 : Select count(id_je) from journal_entries;Sel4 : Select count(balanced) from journal_entries;* Table journal_entries has been analyzed.* id_je is a column with a NOT NULL constraint.* DATE_JE is a column with a NOT NULL constraint.* balanced is a column without a NOT NULL constraint.* There are four indexes on table journal_entries: . NDX_ECR_ID_JE on column id_je . NDX_ECR_DATE_JE_BALANCED on columns date_je, balanced . NDX_ECR_BALANCED_DATE_JE on columns balanced, date_je . NDX_ECR_BALANCED on column balancedI. Sel1 and Sel2:-----------------For CBO, Sel1 and Sel2 are strictly equivalent: 1. CBO looks at the table, retrieves statistics, and calculates the cost for a full table scan.2. CBO then looks for all indexes on the table. In this example it lists four indexes.3. If CBO finds one or more indices which index a column NOT NULL or index at least one column NOT NULL, it evaluates the cost of the path for each of these indices. In our example CBO evaluates three indices: - NDX_ECR_ID_JE - NDX_ECR_DATE_JE_BALANCED - NDX_ECR_BALANCED_DATE_JE4. CBO chooses the path with the lower cost. If an index has the lowest cost, it is chosen, whatever columns are indexed. II. Sel3:---------For Sel3, CBO does the same as for Sel1 and Sel2 since "id_je" has aNOT NULL constraint.If the column in the count() is NOT NULL, CBO considers the selectstrictly equivalent to a select count(*) or a select count(constant).III. Sel4:----------For Sel4, the column in the count() function can have a NULL value.So, what happens?- CBO will not evaluate index NDX_ECR_BALANCED since "balanced" has no NOT NULL constraint. If you are sure that NDX_ECR_BALANCED is more optimal than performing a full table scan on journal_entries, you must rewrite your query from select count(balanced) from journal_entries to: select count(*) from journal_entries where balanced > " " or 0 (depending on column type) CBO will then use the index. Even if you write your select with a hint: /*+ INDEX (JOURNAL_ENTRIES NDX_ECR_BALANCED) */ CBO will not use it as it is not evaluated.- CBO will not evaluate index NDX_ECR_ID_JE because BALANCED is not present in this index.- CBO evaluates indexes NDX_ECR_DATE_JE_BALANCED and NDX_ECR_BALANCED_DATE_JE since these indexes contain a column (DATE_JE) which has a NOT NULL constraint. If the cost of one of these two indexes is less than the cost of a full table scan, CBO uses this index. CBO does not take care of the position of column balanced in the index.References:===========Note:67522.1Why is my index not used?
3.2博客
3.2.1asmtom
https://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:1156159920245
You Asked
What is the difference between count(1) and count(*) in a sql query
eg.
select count(1) from emp;
and
select count(*) from emp;
and we said...
nothing, they are the same, incur the same amount of work -- do the same thing, take the same amount of resources.
You can see this via:
ops$tkyte@ORA817.US.ORACLE.COM> alter session set sql_trace=true;
Session altered.
ops$tkyte@ORA817.US.ORACLE.COM> select count(*) from all_objects;
COUNT(*)
----------
27044
ops$tkyte@ORA817.US.ORACLE.COM> select count(1) from all_objects
2 /
COUNT(1)
----------
27044
and the tkprof will show:
select count(*)
from
all_objects
call count cpu elapsed disk query current rows
------- ------ -------- ---------- ---------- ---------- ---------- ----------
Parse 1 0.02 0.02 0 0 0 0
Execute 1 0.00 0.00 0 0 0 0
Fetch 2 5.56 5.56 0 234998 4 1
------- ------ -------- ---------- ---------- ---------- ---------- ----------
total 4 5.58 5.58 0 234998 4 1
select count(1)
from
all_objects
call count cpu elapsed disk query current rows
------- ------ -------- ---------- ---------- ---------- ---------- ----------
Parse 1 0.02 0.02 0 0 0 0
Execute 1 0.00 0.00 0 0 0 0
Fetch 2 5.46 5.47 0 234998 4 1
------- ------ -------- ---------- ---------- ---------- ---------- ----------
total 4 5.48 5.49 0 234998 4 1
Same number of blocks read/written/processed, same cpu times (basically) same elapsed times (basically).
they are identical.
Anyone who thinks different (and I know you are out there) will have to post a test case like the above or some scientific proof otherwise to be taken seriously....
-------------------------------------------------------------------------
About Me
...............................................................................................................................
●本文作者:小麦苗,只专注于数据库的技术,更注重技术的运用
●本文在itpub(http://blog.itpub.net/26736162)、博客园(http://www.cnblogs.com/lhrbest)和个人微信公众号(xiaomaimiaolhr)上有同步更新
●本文itpub地址:http://blog.itpub.net/26736162/viewspace-2136339/
●本文博客园地址:http://www.cnblogs.com/lhrbest/p/6645603.html
●本文pdf版及小麦苗云盘地址:http://blog.itpub.net/26736162/viewspace-1624453/
● QQ群:230161599微信群:私聊
●联系我请加QQ好友(646634621),注明添加缘由
●于2017-03-28 09:00 ~ 2017-03-30 22:00在魔都完成
●文章内容来源于小麦苗的学习笔记,部分整理自网络,若有侵权或不当之处还请谅解
●版权所有,欢迎分享本文,转载请保留出处
...............................................................................................................................
拿起手机使用微信客户端扫描下边的左边图片来关注小麦苗的微信公众号:xiaomaimiaolhr,扫描右边的二维码加入小麦苗的QQ群,学习最实用的数据库技术。