Oracle里countcount(*)和count(主键)哪个更快

Posted 2020-08-23

这两天听了将近20场演讲，感觉收获很多，最深的感觉就是自己还有很长的路要走。有几个点记录一下：

昨天听老猫讲，提到一个普遍的问题就是Oracle里count(*)、count(1)和count(主键)到底哪个快的问题。这个问题看起来很简单，每个人都会有自己的答案，去百度上搜会出来一大堆帖子来讲哪个更快。但是老猫说了它们三个其实是一样的，我听到之后也觉得挺诧异的，因为我记得别人跟我说过count(主键)会快，然后自己简单想了一下，觉得好像是那么回事的就没有深入去追究。接着老猫说官方有这样的说法这三个其实是等价的。晚上回来之后到MOS上查了一下，居然被我找到了How the Oracle CBO Chooses a Path for the SELECT COUNT(*) Command (文档 ID 124717.1)。这篇文档讲的就是在CBO优化器模式下，Oracle怎样去评估没有where条件select count(*)和select count(colum)语句的最优路径。

1、创建测试表并设计测试场景：

--创建测试表
[email protected]>create table journal_entries 
  2  (id_je number(8) ,
  3  date_je date not null,
  4  balanced number ,
  5  constraint indx_ecr_id_je primary key(id_je)
  6  );

Table created.
--创建索引
[email protected]>create index indx_ecr_date_je_balanced on journal_entries(date_je,balanced);

Index created.

[email protected]>create index indx_ecr_balanced_date_je on journal_entries(balanced,date_je);

Index created.

[email protected]>create index indx_ecr_balanced on journal_entries(balanced);

Index created.
--插入测试数据
[email protected]>insert into journal_entries values(1,sysdate,11);

1 row created.

[email protected]>insert into journal_entries values(2,sysdate,21);

1 row created.

[email protected]>insert into journal_entries values(3,sysdate,31);

1 row created.

[email protected]>insert into journal_entries values(4,sysdate,41);

1 row created.

[email protected]>insert into journal_entries values(5,sysdate,51);

1 row created.

[email protected]>insert into journal_entries values(6,sysdate,61);

1 row created.

[email protected]>insert into journal_entries values(7,sysdate,71);

1 row created.

[email protected]>insert into journal_entries values(8,sysdate,81);

1 row created.

[email protected]>insert into journal_entries values(9,sysdate,91);

1 row created.

[email protected]>commit;

Commit complete.
--收集统计信息
[email protected]>exec dbms_stats.gather_table_stats(ownname=>USER,tabname=>‘JOURNAL_ENTRIES‘,cascade=>true);

PL/SQL procedure successfully completed.

设计四个场景进行对比：

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;

1、场景1和场景2等价

For CBO, Sel1 and Sel2 are strictly equivalent

[email protected]>alter session set statistics_level=all;

Session altered.

[email protected]>select count(*) from journal_entries;

  COUNT(*)
----------
         9

[email protected]>select * from table(dbms_xplan.display_cursor(null,null,‘runstats_last‘));

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  5ja3ukp4wd73p, child number 0
-------------------------------------
select count(*) from journal_entries

Plan hash value: 42135099

---------------------------------------------------------------------------------------------
| Id  | Operation        | Name           | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
---------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT |                |      1 |        |      1 |00:00:00.01 |       1 |
|   1 |  SORT AGGREGATE  |                |      1 |      1 |      1 |00:00:00.01 |       1 |
|   2 |   INDEX FULL SCAN| INDX_ECR_ID_JE |      1 |      9 |      9 |00:00:00.01 |       1 |
---------------------------------------------------------------------------------------------


14 rows selected.

[email protected]>select count(1) from journal_entries;

  COUNT(1)
----------
         9

[email protected]>select * from table(dbms_xplan.display_cursor(null,null,‘runstats_last‘));

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  gbxjjuqj9j7ww, child number 0
-------------------------------------
select count(1) from journal_entries

Plan hash value: 42135099

---------------------------------------------------------------------------------------------
| Id  | Operation        | Name           | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
---------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT |                |      1 |        |      1 |00:00:00.01 |       1 |
|   1 |  SORT AGGREGATE  |                |      1 |      1 |      1 |00:00:00.01 |       1 |
|   2 |   INDEX FULL SCAN| INDX_ECR_ID_JE |      1 |      9 |      9 |00:00:00.01 |       1 |
---------------------------------------------------------------------------------------------


14 rows selected.

可以看到两个语句的执行计划是完全相同的。

2、场景3也与前两个场景等价，因为id_je有NOT NULL约束

For Sel3, CBO does the same as for Sel1 and Sel2 since "id_je" has a NOT NULL constraint.

[email protected]>select count(id_je) from journal_entries;

COUNT(ID_JE)
------------
           9

[email protected]>select * from table(dbms_xplan.display_cursor(null,null,‘runstats_last‘));

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  b1p4v15dwx7hs, child number 0
-------------------------------------
select count(id_je) from journal_entries

Plan hash value: 42135099

---------------------------------------------------------------------------------------------
| Id  | Operation        | Name           | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
---------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT |                |      1 |        |      1 |00:00:00.01 |       1 |
|   1 |  SORT AGGREGATE  |                |      1 |      1 |      1 |00:00:00.01 |       1 |
|   2 |   INDEX FULL SCAN| INDX_ECR_ID_JE |      1 |      9 |      9 |00:00:00.01 |       1 |
---------------------------------------------------------------------------------------------


14 rows selected.

可以看到执行计划与前两个也是完全相同的。

4、场景4跟前边3个不同，因为balanced列上没有NOT NULL约束，但是balanced列上有索引，那会走这个列上的索引么？我们来看一下执行计划：

[email protected]>select count(balanced) from journal_entries;

COUNT(BALANCED)
---------------
              9

[email protected]>select * from table(dbms_xplan.display_cursor(null,null,‘runstats_last‘));

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  bc3bc8c0fg14z, child number 0
-------------------------------------
select count(balanced) from journal_entries

Plan hash value: 3638043346

--------------------------------------------------------------------------------------------------------
| Id  | Operation        | Name                      | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
--------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT |                           |      1 |        |      1 |00:00:00.01 |       1 |
|   1 |  SORT AGGREGATE  |                           |      1 |      1 |      1 |00:00:00.01 |       1 |
|   2 |   INDEX FULL SCAN| INDX_ECR_DATE_JE_BALANCED |      1 |      9 |      9 |00:00:00.01 |       1 |
--------------------------------------------------------------------------------------------------------


14 rows selected.

我们看到这个执行计划没有走balanced列上的索引，而是走了和date_je的联合索引。这个可以查看另一篇文档：Note:67522.1 Why is my index not used?

小结一下：

我这里只是简单的从执行计划上看count(*)、count(1)和count(主键)其实是一致，MOS的文档中详细的讲解了Oracle是如何评估执行计划的，也可以使用10053 event查看CBO优化器是如何做出选择的。由于我的功力还不够，对于10053事件还不是很明白，暂时就先不做演示了，要不哪说错了就不好了，这也可以做为以后博客分享的内容。

从这个事情上来看，我们对于一件事情应该做一个深入的研究，有充足的证据来证明，尤其是想要在某一方面有深入发展的时候。

本文出自 “DBA Fighting!” 博客，请务必保留此出处http://hbxztc.blog.51cto.com/1587495/1883642