count(*)、count(val)和count(1)的解釋

原創 十二试炼 2020-02-21 08:42

一、關於count的一些謠言：

    1、count(*)比count(val)更慢！項目組必須用count(val)，不準用count(*)，誰用扣誰錢！

    2、count(*)用不到索引，count(val)才能用到。

    3、count(*)是統計出全表的記錄，是吞吐量的操作，肯定用不到索引。

    4、count(1)比count(*)的速度快。

二、驗證count(*)和count(val)

    1、首先創建一個表，使用count(*)和count(val)查詢比較：

<span style="font-size:14px;">----刪除echo表----
SQL> drop table echo purge;
drop table echo purge
           *
第 1 行出現錯誤:
ORA-00942: 表或視圖不存在

----創建一張echo的測試表----
SQL> create table echo as select * from dba_objects;

表已創建。

SQL> update echo set object_id =  rownum;

已更新72509行。

SQL> commit;

提交完成。

SQL> set timing on
SQL> set linesize 100
SQL> set autotrace on
SQL> select count(*) from echo;

  COUNT(*)
----------

已用時間:  00: 00: 00.01

執行計劃
----------------------------------------------------------
Plan hash value: 99109176

-------------------------------------------------------------------
| Id  | Operation       | Name | Rows  | Cost (%CPU)| Time      |
-------------------------------------------------------------------
|   0 | SELECT STATEMENT   |      |    1 |   290   (1)| 00:00:04 |
|   1 |  SORT AGGREGATE    |      |    1 |           |      |
|   2 |   TABLE ACCESS FULL| ECHO | 80064 |   290   (1)| 00:00:04 |
-------------------------------------------------------------------

Note
-----
   - dynamic sampling used for this statement (level=2)


統計信息
----------------------------------------------------------
 recursive calls
 db block gets
 consistent gets
 physical reads
 redo size
 bytes sent via SQL*Net to client
 bytes received via SQL*Net from client
 SQL*Net roundtrips to/from client
 sorts (memory)
 sorts (disk)
 rows processed

SQL> select count(*) from echo;

  COUNT(*)
----------

已用時間:  00: 00: 00.01

執行計劃
----------------------------------------------------------
Plan hash value: 99109176

-------------------------------------------------------------------
| Id  | Operation       | Name | Rows  | Cost (%CPU)| Time      |
-------------------------------------------------------------------
|   0 | SELECT STATEMENT   |      |    1 |   290   (1)| 00:00:04 |
|   1 |  SORT AGGREGATE    |      |    1 |           |      |
|   2 |   TABLE ACCESS FULL| ECHO | 80064 |   290   (1)| 00:00:04 |
-------------------------------------------------------------------

Note
-----
   - dynamic sampling used for this statement (level=2)


統計信息
----------------------------------------------------------
 recursive calls
 db block gets
 consistent gets
 physical reads
 redo size
 bytes sent via SQL*Net to client
 bytes received via SQL*Net from client
 SQL*Net roundtrips to/from client
 sorts (memory)
 sorts (disk)
 rows processed

SQL> select count(object_id) from echo;

COUNT(OBJECT_ID)
----------------

已用時間:  00: 00: 00.01

執行計劃
----------------------------------------------------------
Plan hash value: 99109176

---------------------------------------------------------------------------
| Id  | Operation       | Name | Rows  | Bytes | Cost (%CPU)| Time      |
---------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |      |    1 |    13 |   290   (1)| 00:00:04 |
|   1 |  SORT AGGREGATE    |      |    1 |    13 |           |      |
|   2 |   TABLE ACCESS FULL| ECHO | 80064 |  1016K|   290   (1)| 00:00:04 |
---------------------------------------------------------------------------

Note
-----
   - dynamic sampling used for this statement (level=2)


統計信息
----------------------------------------------------------
 recursive calls
 db block gets
 consistent gets
 physical reads
 redo size
 bytes sent via SQL*Net to client
 bytes received via SQL*Net from client
 SQL*Net roundtrips to/from client
 sorts (memory)
 sorts (disk)
 rows processed

SQL> select count(object_id) from echo;

COUNT(OBJECT_ID)
----------------

已用時間:  00: 00: 00.01

執行計劃
----------------------------------------------------------
Plan hash value: 99109176

---------------------------------------------------------------------------
| Id  | Operation       | Name | Rows  | Bytes | Cost (%CPU)| Time      |
---------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |      |    1 |    13 |   290   (1)| 00:00:04 |
|   1 |  SORT AGGREGATE    |      |    1 |    13 |           |      |
|   2 |   TABLE ACCESS FULL| ECHO | 80064 |  1016K|   290   (1)| 00:00:04 |
---------------------------------------------------------------------------

Note
-----
   - dynamic sampling used for this statement (level=2)


統計信息
----------------------------------------------------------
 recursive calls
 db block gets
 consistent gets
 physical reads
 redo size
 bytes sent via SQL*Net to client
 bytes received via SQL*Net from client
 SQL*Net roundtrips to/from client
 sorts (memory)
 sorts (disk)
 rows processed</span>

從上面的執行計劃可以看出count(*)和count(val)是一樣快的。

    2、建立索引做比較

<span style="font-size:14px;">SQL> create index idx_object_id on echo(object_id);

索引已創建。

已用時間:  00: 00: 05.69
SQL> select count(*) from echo;

  COUNT(*)
----------

已用時間:  00: 00: 00.05

執行計劃
----------------------------------------------------------
Plan hash value: 99109176

-------------------------------------------------------------------
| Id  | Operation       | Name | Rows  | Cost (%CPU)| Time      |
-------------------------------------------------------------------
|   0 | SELECT STATEMENT   |      |    1 |   290   (1)| 00:00:04 |
|   1 |  SORT AGGREGATE    |      |    1 |           |      |
|   2 |   TABLE ACCESS FULL| ECHO | 80064 |   290   (1)| 00:00:04 |
-------------------------------------------------------------------

Note
-----
   - dynamic sampling used for this statement (level=2)


統計信息
----------------------------------------------------------
 recursive calls
 db block gets
 consistent gets
 physical reads
 redo size
 bytes sent via SQL*Net to client
 bytes received via SQL*Net from client
 SQL*Net roundtrips to/from client
 sorts (memory)
 sorts (disk)
 rows processed

SQL> select count(object_id) from echo;

COUNT(OBJECT_ID)
----------------

已用時間:  00: 00: 00.08

執行計劃
----------------------------------------------------------
Plan hash value: 1131838604

---------------------------------------------------------------------------------------
| Id  | Operation          | Name          | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT      |           |     1 |    13 |    49    (0)| 00:00:01 |
|   1 |  SORT AGGREGATE       |           |     1 |    13 |        |          |
|   2 |   INDEX FAST FULL SCAN| IDX_OBJECT_ID | 80064 |  1016K|    49    (0)| 00:00:01 |
---------------------------------------------------------------------------------------

Note
-----
   - dynamic sampling used for this statement (level=2)


統計信息
----------------------------------------------------------
 recursive calls
 db block gets
 consistent gets
 physical reads
 redo size
 bytes sent via SQL*Net to client
 bytes received via SQL*Net from client
 SQL*Net roundtrips to/from client
 sorts (memory)
 sorts (disk)
 rows processed</span>

    哇，原來真的是用count(val)比count(*)要快啊，因爲count(*)不能用到索引，而count(val)可以，真相真是如此嗎？
    3、將object_id設置爲非空SQL> alter table echo modify object_id not null;
 

<span style="font-size:14px;">SQL> alter table echo modify object_id not null;

表已更改。

已用時間:  00: 00: 01.41

SQL> select count(*) from echo;

  COUNT(*)
----------

已用時間:  00: 00: 00.00

執行計劃
----------------------------------------------------------
Plan hash value: 1131838604

-------------------------------------------------------------------------------
| Id  | Operation          | Name          | Rows  | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------
|   0 | SELECT STATEMENT      |           |     1 |    49    (0)| 00:00:01 |
|   1 |  SORT AGGREGATE       |           |     1 |        |          |
|   2 |   INDEX FAST FULL SCAN| IDX_OBJECT_ID | 80064 |    49    (0)| 00:00:01 |
-------------------------------------------------------------------------------

Note
-----
   - dynamic sampling used for this statement (level=2)


統計信息
----------------------------------------------------------
 recursive calls
 db block gets
 consistent gets
 physical reads
 redo size
 bytes sent via SQL*Net to client
 bytes received via SQL*Net from client
 SQL*Net roundtrips to/from client
 sorts (memory)
 sorts (disk)
 rows processed

SQL> select count(object_id) from echo;

COUNT(OBJECT_ID)
----------------

已用時間:  00: 00: 00.01

執行計劃
----------------------------------------------------------
Plan hash value: 1131838604

-------------------------------------------------------------------------------
| Id  | Operation          | Name          | Rows  | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------
|   0 | SELECT STATEMENT      |           |     1 |    49    (0)| 00:00:01 |
|   1 |  SORT AGGREGATE       |           |     1 |        |          |
|   2 |   INDEX FAST FULL SCAN| IDX_OBJECT_ID | 80064 |    49    (0)| 00:00:01 |
-------------------------------------------------------------------------------

Note
-----
   - dynamic sampling used for this statement (level=2)


統計信息
----------------------------------------------------------
 recursive calls
 db block gets
 consistent gets
 physical reads
 redo size
 bytes sent via SQL*Net to client
 bytes received via SQL*Net from client
 SQL*Net roundtrips to/from client
 sorts (memory)
 sorts (disk)
 rows processed</span>

--看來count(val)和count(*)其實一樣快，如果索引列是非空的，count(*)可用到索引，此時一樣快！真相真是如此嗎？
其實兩者根本沒有可比性，性能比較首先考慮寫法等價，這兩個語句根本就不等價。

結論：

    其實優化器裏的算法是這麼玩的，列的偏移量決定性能，列越靠後，訪問的開銷越大。由於count(*)的算法與列偏移量無關，所以count(*)最快，count(最後列val)最慢。

    3、用實驗驗證上面的結論 

<span style="font-size:14px;">SQL> set serveroutput on
SQL> set echo on
SQL> drop table t purge;

表已刪除。
----構造出有25個字段的表T----
DECLARE
  l_sql VARCHAR2(32767);
BEGIN
  l_sql := 'CREATE TABLE t (';
  FOR i IN 1..25 
  LOOP
    l_sql := l_sql || 'n' || i || ' NUMBER,';
  END LOOP;
  l_sql := l_sql || 'pad VARCHAR2(1000)) PCTFREE 10';
  EXECUTE IMMEDIATE l_sql;
END;
 /

PL/SQL 過程已成功完成。
----將記錄還有這個表T中填充----
DECLARE
  l_sql VARCHAR2(32767);
BEGIN
  l_sql := 'INSERT INTO t SELECT ';
  FOR i IN 1..25
  LOOP
    l_sql := l_sql || '0,';
  END LOOP;
  l_sql := l_sql || 'NULL FROM dual CONNECT BY level <= 10000';
  EXECUTE IMMEDIATE l_sql;
  COMMIT;
END;
 /

PL/SQL 過程已成功完成。

execute dbms_stats.gather_table_stats(ownname=>user, tabname=>'t')
SELECT num_rows, blocks FROM user_tables WHERE table_name = 'T';

PL/SQL 過程已成功完成。

SQL> 
  NUM_ROWS     BLOCKS
---------- ----------
      80
----以下動作觀察執行速度，比較發現count(*)最快，count(最大列)最慢----
SQL> DECLARE
  l_dummy PLS_INTEGER;
  l_start PLS_INTEGER;
  l_stop PLS_INTEGER;
  l_sql VARCHAR2(100);
BEGIN
  l_start := dbms_utility.get_time;
  FOR j IN 1..1000
  LOOP
    EXECUTE IMMEDIATE 'SELECT count(*) FROM t' INTO l_dummy;
  END LOOP;
  l_stop := dbms_utility.get_time;
  dbms_output.put_line((l_stop-l_start)/100);
 
  FOR i IN 1..25
  LOOP
    l_sql := 'SELECT count(n' || i || ') FROM t';
    l_start := dbms_utility.get_time;
    FOR j IN 1..1000
    LOOP
      EXECUTE IMMEDIATE l_sql INTO l_dummy;
    END LOOP;
    l_stop := dbms_utility.get_time;
    dbms_output.put_line((l_stop-l_start)/100);
  END LOOP;
END;
 /
.18
.33
.39
.38
.42
.4
.45
.49
.48
.46
.48
.48
.55
.51
.56
.57
.61
.62
.75
.67
.68
.7
.73
.78
.77
.81
PL/SQL 過程已成功完成。</span>

三、驗證count(*)和count(1)
    沿用TOM大師的解釋：

 
十二試煉
發佈了17 篇原創文章 · 獲贊 9 · 訪問量 14萬+
私信 關注
發表評論
所有評論
還沒有人評論，想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章

優酷鴻蒙開發實踐 | 鴻蒙卡片開發

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image"

阿里巴巴移动技术 2021-10-22 15:33:55

數據庫測試的基礎要素

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"typ

Jonathan Allen 2021-10-22 15:33:55

一年40萬次實驗，字節跳動A/B測試平臺是怎麼煉出來的？

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"typ

Tina 2021-09-26 15:43:52

走過二十年的軟件測試業

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockq

Tim Bray 2021-08-06 17:03:57

爲了生存重寫軟件！面對“缺芯”，特斯拉很慌

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragr

Tina 2021-08-02 14:58:59

谷歌開發團隊犯低級錯誤？因一個字符拼寫Bug，Chromebook用戶被鎖在系統之外

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragr

Kent Duke 2021-07-25 16:38:56

10年磨一劍，雲測試頭部玩家如何爲數萬家客戶落地AI自動化測試

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null

施尧 2021-07-25 10:04:06

怎樣在企業裏推廣TDD等技術實踐？

{"type":"doc","content":[{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"爲什麼TDD很重要"

Emily Bache 2021-07-20 10:03:55

從Kratos設計看Go微服務工程實踐

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null

腾讯云中间件 2021-07-16 15:33:59

從10次宕機事件中，我學到5個重要的經驗

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"typ

Tom Kleinpeter 2021-07-13 11:13:50

複雜SQL治理實踐

一、前言 軟件在持續的開發和維護過程中，會不斷添加新功能和修復舊的缺陷，這往往伴隨着代碼的快速增長和複雜性的提升。若代碼庫沒有得到良好的管理和重構，就可能積累大量的技術債務，包括不一致的設計、冗餘代碼、過時的庫和框架以及不再使用的功能。

京東雲開發者 2024-05-24 11:56:56

Linux mpstat 命令

mpstat的語法如下 mpstat [-P {cpu|ALL}] [internal [count]] 其中，各參數含義如下： 參數    含義 -P {cpu l ALL}    表示監控哪個CPU， cpu在[0,cpu個數-1]中取

原創 2024-04-07 13:07:06

盤點一個Python字符串格式化處理的問題（AI+Python）

大家好，我是Python進階者。 一、前言 前幾天在Python白銀交流羣【空翼】問了一個字符串格式化處理的問題，問題如下： 二、實現過程 這裏【東哥】給了一個指導，如下所示： {word:<10}：這是一個格式化字段，word是變量名，

原創 2024-03-29 10:02:11

無人不識又無人不迷糊的this

本文分享自華爲雲社區《3月閱讀周·你不知道的JavaScript | 無人不識又無人不迷糊的this》，作者： 葉一一。 關於this this關鍵字是JavaScript中最複雜的機制之一。它是一個很特別的關鍵字，被自動定義在所有函數的

原創 2024-03-26 11:34:42

R語言讀取大型NetCDF文件

失蹤人口迴歸，本篇來介紹下R語言讀取大型NetCDF文件的一些實踐。 1 NetCDF數據簡介 先給一段Wiki上關於NetCDF的定義。 NetCDF (Network Common Data Form) is a set of soft

原創 2024-03-11 02:07:23