绑定变量对于SQL性能的影响

两个小黄鹂2024-09-15 11:12

绑定变量

sql 复制代码 
SQL> set timing on
SQL> alter session set  statistics_level=all;
VAR v4 VARCHAR2(32)
VAR v5 VARCHAR2(128)
VAR v6 VARCHAR2(128)

EXEC :v4:='007680'
EXEC :v5:='P1803585080457506817'
EXEC :v6:='202406200826069960000000000'

SELECT AD.* FROM TEST1.TESTTABLE AD WHERE AD.InstitutionID = :v4 AND AD.PaymentTxSn = :v5 AND AD.SourceSystemNo = :v6 AND Status = 30;
select * from table(dbms_xplan.display_cursor(null,null,'allstats last'));

Elapsed: 00:00:00.87
SQL> SQL_ID     06nzu53vnm5gf, child number 0
-------------------------------------
SELECT AD.* FROM AGGREGATE1.TESTTABLE AD WHERE
AD.InstitutionID = :v4 AND AD.PaymentTxSn = :v5 AND AD.SourceSystemNo =
:v6 AND Status = 30

Plan hash value: 4134968517

------------------------------------------------------------------------------------------------------------------------------
| Id  | Operation                           | Name                         | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
------------------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                    |                              |      1 |        |      0 |00:00:00.86 |     542K|
|*  1 |  TABLE ACCESS BY INDEX ROWID BATCHED| TESTTABLE                    |      1 |      1 |      0 |00:00:00.86 |     542K|
|*  2 |   INDEX RANGE SCAN                  | U_TESTTABLE_INSTID_TXSN      |      1 |      1 |    572K|00:00:00.08 |    6340 |
------------------------------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter(("AD"."PAYMENTTXSN"=:V5 AND "AD"."SOURCESYSTEMNO"=:V6 AND TO_NUMBER("STATUS")=30))
   2 - access("AD"."INSTITUTIONID"=:V4)

Elapsed: 00:00:00.03
SQL> alter session set  statistics_level=all;
VAR v4 VARCHAR2(32)
VAR v5 VARCHAR2(128)
VAR v6 VARCHAR2(128)

EXEC :v4:='007759'
EXEC :v5:='P1803585080457506817'
EXEC :v6:='202406200826069960000000000'

SELECT AD.* FROM AGGREGATE1.TESTTABLE AD WHERE AD.InstitutionID = :v4 AND AD.PaymentTxSn = :v5 AND AD.SourceSystemNo = :v6 AND Status = 30;
select * from table(dbms_xplan.display_cursor(null,null,'allstats last'));

Elapsed: 00:00:12.48
SQL> SQL_ID     06nzu53vnm5gf, child number 0
-------------------------------------
SELECT AD.* FROM AGGREGATE1.TESTTABLE AD WHERE
AD.InstitutionID = :v4 AND AD.PaymentTxSn = :v5 AND AD.SourceSystemNo =
:v6 AND Status = 30

Plan hash value: 4134968517

---------------------------------------------------------------------------------------------------------------------------------------
| Id  | Operation                           | Name                         | Starts | E-Rows | A-Rows |   A-Time   | Buffers | Reads  |
---------------------------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                    |                              |      1 |        |      0 |00:00:12.48 |      11M|      4 |
|*  1 |  TABLE ACCESS BY INDEX ROWID BATCHED| TESTTABLE                    |      1 |      1 |      0 |00:00:12.48 |      11M|      4 |
|*  2 |   INDEX RANGE SCAN                  | U_TESTTABLE_INSTID_TXSN      |      1 |      1 |     12M|00:00:01.54 |     128K|      0 |
---------------------------------------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter(("AD"."PAYMENTTXSN"=:V5 AND "AD"."SOURCESYSTEMNO"=:V6 AND TO_NUMBER("STATUS")=30))
   2 - access("AD"."INSTITUTIONID"=:V4)

Elapsed: 00:00:00.03

全使用绑定变量第一次执行sql之后，紧接着第二次执行不同绑定变量的值情况下第一次执行0.87秒第二次执行12.48秒，是因为oracle数据库为了防止大量硬解析做的优化工作。第二次会沿用第一次的执行计划。

不绑定变量

sql 复制代码 
SQL> alter session set  statistics_level=all;
VAR v5 VARCHAR2(128)
VAR v6 VARCHAR2(128)

EXEC :v5:='P1803585080457506817'
EXEC :v6:='202406200826069960000000000'

SELECT AD.* FROM AGGREGATE1.TESTTABLE AD WHERE AD.InstitutionID = '007759' AND AD.PaymentTxSn = :v5 AND AD.SourceSystemNo = :v6 AND Status = 30;
select * from table(dbms_xplan.display_cursor(null,null,'allstats last'));

Elapsed: 00:00:00.00

SQL_ID  a7x2cx23c1034, child number 0
-------------------------------------
SELECT AD.* FROM AGGREGATE1.TESTTABLE AD WHERE
AD.InstitutionID = '007759' AND AD.PaymentTxSn = :v5 AND
AD.SourceSystemNo = :v6 AND Status = 30

Plan hash value: 1281936248

------------------------------------------------------------------------------------------------------------------------------
| Id  | Operation                           | Name                         | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
------------------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                    |                              |      1 |        |      0 |00:00:00.01 |       4 |
|*  1 |  TABLE ACCESS BY INDEX ROWID BATCHED| TESTTABLE                    |      1 |      1 |      0 |00:00:00.01 |       4 |
|*  2 |   INDEX RANGE SCAN                  | I_TESTTABLE_PAYMENTTXSN      |      1 |      1 |      0 |00:00:00.01 |       4 |
------------------------------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter(("AD"."SOURCESYSTEMNO"=:V6 AND "AD"."INSTITUTIONID"='007759' AND TO_NUMBER("STATUS")=30))
   2 - access("AD"."PAYMENTTXSN"=:V5)

Elapsed: 00:00:00.10

SQL> alter session set  statistics_level=all;
VAR v5 VARCHAR2(128)
VAR v6 VARCHAR2(128)

EXEC :v5:='P1803585080457506817'
EXEC :v6:='202406200826069960000000000'

SELECT AD.* FROM AGGREGATE1.TESTTABLE AD WHERE AD.InstitutionID = '007898' AND AD.PaymentTxSn = :v5 AND AD.SourceSystemNo = :v6 AND Status = 30;
select * from table(dbms_xplan.display_cursor(null,null,'allstats last')); 

Elapsed: 00:00:00.01

SQL_ID  gn57p95cqx1ga, child number 0
-------------------------------------
SELECT AD.* FROM AGGREGATE1.TESTTABLE AD WHERE
AD.InstitutionID = '007898' AND AD.PaymentTxSn = :v5 AND
AD.SourceSystemNo = :v6 AND Status = 30

Plan hash value: 4134968517

------------------------------------------------------------------------------------------------------------------------------
| Id  | Operation                           | Name                         | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
------------------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                    |                              |      1 |        |      0 |00:00:00.01 |       6 |
|*  1 |  TABLE ACCESS BY INDEX ROWID BATCHED| TESTTABLE                    |      1 |      1 |      0 |00:00:00.01 |       6 |
|*  2 |   INDEX RANGE SCAN                  | U_TESTTABLE_INSTID_TXSN      |      1 |      1 |      1 |00:00:00.01 |       5 |
------------------------------------------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter(("AD"."PAYMENTTXSN"=:V5 AND "AD"."SOURCESYSTEMNO"=:V6 AND TO_NUMBER("STATUS")=30))
   2 - access("AD"."INSTITUTIONID"='007898')

Elapsed: 00:00:00.04

把机构id绑定变量取消了，带入不同机构id情况下执行计划改变了，并且每次执行时间都很短，符合开发规范中对于绑定变量不可以使用在分布不均匀的列上的说法

总结：

绑定变量：

SQL在数据库认为就是一条SQL，该SQL在硬解析后会统一使用一个执行计划，在重新硬解析前都会沿用该执行计划，即便有效率高的执行计划，此时数据库也不再考虑使用，因此出现慢SQL。

不绑定变量：

SQL参数不同，在数据库认为是不同的SQL，每条SQL会分别解析出自己合适的执行计划，会减少慢SQL的发生，考虑当前环境因素，建议在报表类或分析类场景不绑定变量。

  1. 对于字段值比较少，个数小于100，且分布很不均匀的情况不要使用绑定变量，比如说状态、性别、类型等，还有绑定变量最好用于等值连接
  2. 对于那种大于小于列上有索引，并且会出现跨度很大、跨度很小两级分化很严重的情况也不要使用绑定变量
相关推荐
爱喝水的鱼丶2 小时前
SAP-ABAP:S/4 HANA ABAP 新特性入门：与 ECC ABAP 的核心差异、架构优势与学习路径规划
运维·数据库·性能优化·sap·abap·hana·开发交流
余槐i8 小时前
PostgreSQL 函数索引报错 must be immutable：3种绕法与代价
数据库·后端·sql·postgresql·性能优化
数据库小学妹9 小时前
PostgreSQL表膨胀怎么办？autovacuum清不动的根因与排查命令
数据库·postgresql·表膨胀·mvcc·数据库运维
mister_guo9 小时前
MySQL-1267错误：排序规则冲突的原因与解决方案
数据库·mysql
骇客野人9 小时前
在麒麟操作系统上 分别安装 jdk1.8， 达梦数据库，redis ，memcache 的操作实施方案和步骤
数据库·redis·memcached
美狐美颜SDK开放平台11 小时前
直播APP如何实现美白滤镜功能？视频美颜SDK技术开发方案详解
前端·数据库·人工智能·音视频·美颜sdk·直播美颜sdk·视频美颜sdk
花间相见12 小时前
【计算基础｜MySQL01】【上手】安装连接、数据类型与建库建表
数据库·后端
花间相见12 小时前
【计算基础｜MySQL02】【上手】SQL 增删改查
数据库·后端
ITxiaobing202313 小时前
IP 库与 AppsFlyer 国家/地区数据不一致，如何系统地对账？
数据库·网络协议·tcp/ip
这个DBA有点耶13 小时前
从时间到语义：时序数据库与向量数据库的核心机制、选型与多模融合趋势
数据库·mysql·架构
热门推荐
01「全语录+省流版」邵艾伦 × 孙宇晨：聊风口、聊个人IP、聊年轻人如何抓住 AI 时代的机会02国内可直接用、免费额度/永久免费的大模型API清单（含 SiliconFlow、火山、阿里、智谱、百度、Kimi、DeepSeek、DMXAPI 等）03历年考研数学一、数学二、数学三真题试卷及答案PDF04GitHub 镜像站点05几个好用的ip纯净度检测网站06蜘蛛磁力 搜索引擎大全，如何使用蜘蛛磁力查找磁力链接07Telegram APK 下载安装完整指南 — 2026年最新08Windows 10/11 如何退出微软账户并改用本地账户登录（含找不到选项的解决方法）09微信历史版本含下载地址（ Windows PC | 安卓 | MAC ）及设置微信不更新10轻量化AI渗透武器库｜ARTEX自主攻防系统、资产梳理、漏洞挖掘、链路分析全流程智能化落地