休眠条件查询需要时间从数据库表中获取记录

Question

在我的项目中，我将 10000 条记录插入到表中。如果任何记录已经可用，我们将通过使用休眠条件获取它并将其与新记录进行比较。我们在表中的数据非常少，大约 10000 条记录。对于休眠标准，我们已经应用了条件。此休眠条件从表中获取 6 条记录。但是，如果我调用 criteria.list() 方法来获取 6 条记录，为什么要花时间

HIbernate 条件查询：

select this_.CLASS_TYPE as CLASS1_15_0_, this_.CLASS_UNIT_TYPE as CLASS2_15_0_, this_.CLASS_UNIT_CODE as CLASS3_15_0_, 
this_.ID as ID15_0_, this_.LANG_CODE_ISO as LANG5_15_0_, this_.CTY_CODE_ISO as CTY6_15_0_, this_.NAME as NAME15_0_, this_.INS_DTIME as INS8_15_0_,this_.UPD_DTIME as UPD9_15_0_, this_.DEL_DTIME as DEL10_15_0_ from FI_SALES_LOCATION_V this_ where (this_.CLASS_TYPE='BU' and this_.CLASS_UNIT_TYPE='STO' and this_.CLASS_UNIT_CODE='A1000') and this_.DEL_DTIME is null and ((this_.LANG_CODE_ISO='en' and this_.CTY_CODE_ISO='GB' and this_.ID='A1000') or (this_.LANG_CODE_ISO='en' and this_.CTY_CODE_ISO='GB' and this_.ID='AP1000') or (this_.LANG_CODE_ISO='pt' and this_.CTY_CODE_ISO='PT' and this_.ID='AP1000') or (this_.LANG_CODE_ISO='pt' and this_.CTY_CODE_ISO='PT' and this_.ID='A1000') or (this_.LANG_CODE_ISO='s1' and this_.CTY_CODE_ISO='SI' and this_.ID='AP1000') or (this_.LANG_CODE_ISO='s1' and this_.CTY_CODE_ISO='SI' and this_.ID='A1000'));

Answer 1

考虑到来自问题和 cmets 的信息，您的代码是这样的：

• 对于要插入的每条记录，您都可以通过条件进行选择。
• 进行检查，然后根据具体情况保留或合并记录。

您必须了解休眠会话在某些情况下的行为方式。当您在休眠中执行操作时，您正在做的（在第一时间）是更新休眠会话。无论是在插入、删除还是更新中，操作都将在会话中执行，Hibernate 甚至会记录一条 SQL 语句，但 它不会立即刷新到您的数据库（是的，它具有误导性)。

通常一切都被刷新并结束事务，但在某些情况下，hibernate 会刷新操作以使会话保持最新。在您的情况下，您有多个要插入/合并的记录，但在每次操作之前执行选择。 Hibernate 检测到这个操作并且会刷新每个之前存储的操作，仅仅是因为如果有挂起的操作要刷新，查询数据库是没有意义的。数据库需要处于可能的最新状态。不要试图与之抗争，它旨在以这种方式工作。

所以我最好的猜测是，当您的 select 执行时间过长时，这是因为 hibernate 正在刷新之前的插入/更新操作。

如何解决：

我想不出任何直接的方法来解决这个问题。您将不得不重新考虑您的方法。

有一件很重要的事情：即使您只需要在没有任何验证的情况下进行纯插入，您也会遇到一些麻烦，因为您的代码试图插入数千条记录。最大的问题是，就像我之前写的那样，hibernate 直到事务结束才会刷新操作。因此，它会将每个要执行的语句保存在内存中。您很有可能会耗尽内存。

我要做的第一件事是检查hibernate docs on batching（从docs开始）。之后，我会想出一个策略来优化插入大量记录的需求和所需的验证。 与其处理单个记录，不如处理数据块。

根据可用的少量信息，我会尝试这个工作流程：

在一切之前，为 hibernate 定义批处理大小属性（再次检查批处理文档）

• 在处理数据时，使用为批量大小定义的相同大小的数据切片
• 在要插入的记录中选择要验证的所有可能记录
• 执行所需的验证以检查切片中的每条记录是合并还是插入
• 组插入和合并并执行操作。刷新和清除休眠会话
• 获取下一个数据片段，然后重复该过程

基本上这是您正在执行的“相同”过程，但它将针对记录块执行，而不是在每条记录中执行。

Answer 2

尝试这样写查询：

select this_.CLASS_TYPE as CLASS1_15_0_, this_.CLASS_UNIT_TYPE as CLASS2_15_0_, this_.CLASS_UNIT_CODE as CLASS3_15_0_, 
this_.ID as ID15_0_, this_.LANG_CODE_ISO as LANG5_15_0_, this_.CTY_CODE_ISO as CTY6_15_0_, this_.NAME as NAME15_0_, this_.INS_DTIME as INS8_15_0_,this_.UPD_DTIME as UPD9_15_0_, this_.DEL_DTIME as DEL10_15_0_ 
from FI_SALES_LOCATION_V this_ 
where 
(this_.CLASS_TYPE='BU' and this_.CLASS_UNIT_TYPE='STO' and this_.CLASS_UNIT_CODE='A1000') 
and this_.DEL_DTIME is null 
and (this_.LANG_CODE_ISO='en' and this_.CTY_CODE_ISO='GB' and this_.ID='A1000') 
UNION ALL
select this_.CLASS_TYPE as CLASS1_15_0_, this_.CLASS_UNIT_TYPE as CLASS2_15_0_, this_.CLASS_UNIT_CODE as CLASS3_15_0_, 
this_.ID as ID15_0_, this_.LANG_CODE_ISO as LANG5_15_0_, this_.CTY_CODE_ISO as CTY6_15_0_, this_.NAME as NAME15_0_, this_.INS_DTIME as INS8_15_0_,this_.UPD_DTIME as UPD9_15_0_, this_.DEL_DTIME as DEL10_15_0_ 
from FI_SALES_LOCATION_V this_ 
where 
(this_.CLASS_TYPE='BU' and this_.CLASS_UNIT_TYPE='STO' and this_.CLASS_UNIT_CODE='A1000') 
and this_.DEL_DTIME is null 
and (this_.LANG_CODE_ISO='en' and this_.CTY_CODE_ISO='GB' and this_.ID='AP1000') 
UNION ALL
select this_.CLASS_TYPE as CLASS1_15_0_, this_.CLASS_UNIT_TYPE as CLASS2_15_0_, this_.CLASS_UNIT_CODE as CLASS3_15_0_, 
this_.ID as ID15_0_, this_.LANG_CODE_ISO as LANG5_15_0_, this_.CTY_CODE_ISO as CTY6_15_0_, this_.NAME as NAME15_0_, this_.INS_DTIME as INS8_15_0_,this_.UPD_DTIME as UPD9_15_0_, this_.DEL_DTIME as DEL10_15_0_ 
from FI_SALES_LOCATION_V this_ 
where 
(this_.CLASS_TYPE='BU' and this_.CLASS_UNIT_TYPE='STO' and this_.CLASS_UNIT_CODE='A1000') 
and this_.DEL_DTIME is null 
and (this_.LANG_CODE_ISO='pt' and this_.CTY_CODE_ISO='PT' and this_.ID='AP1000') 
UNION ALL
select this_.CLASS_TYPE as CLASS1_15_0_, this_.CLASS_UNIT_TYPE as CLASS2_15_0_, this_.CLASS_UNIT_CODE as CLASS3_15_0_, 
this_.ID as ID15_0_, this_.LANG_CODE_ISO as LANG5_15_0_, this_.CTY_CODE_ISO as CTY6_15_0_, this_.NAME as NAME15_0_, this_.INS_DTIME as INS8_15_0_,this_.UPD_DTIME as UPD9_15_0_, this_.DEL_DTIME as DEL10_15_0_ 
from FI_SALES_LOCATION_V this_ 
where 
(this_.CLASS_TYPE='BU' and this_.CLASS_UNIT_TYPE='STO' and this_.CLASS_UNIT_CODE='A1000') 
and this_.DEL_DTIME is null 
and (this_.LANG_CODE_ISO='pt' and this_.CTY_CODE_ISO='PT' and this_.ID='A1000') 
UNION  ALL
select this_.CLASS_TYPE as CLASS1_15_0_, this_.CLASS_UNIT_TYPE as CLASS2_15_0_, this_.CLASS_UNIT_CODE as CLASS3_15_0_, 
this_.ID as ID15_0_, this_.LANG_CODE_ISO as LANG5_15_0_, this_.CTY_CODE_ISO as CTY6_15_0_, this_.NAME as NAME15_0_, this_.INS_DTIME as INS8_15_0_,this_.UPD_DTIME as UPD9_15_0_, this_.DEL_DTIME as DEL10_15_0_ 
from FI_SALES_LOCATION_V this_ 
where 
(this_.CLASS_TYPE='BU' and this_.CLASS_UNIT_TYPE='STO' and this_.CLASS_UNIT_CODE='A1000') 
and this_.DEL_DTIME is null 
and (this_.LANG_CODE_ISO='s1' and this_.CTY_CODE_ISO='SI' and this_.ID='AP1000') 
UNION ALL
select this_.CLASS_TYPE as CLASS1_15_0_, this_.CLASS_UNIT_TYPE as CLASS2_15_0_, this_.CLASS_UNIT_CODE as CLASS3_15_0_, 
this_.ID as ID15_0_, this_.LANG_CODE_ISO as LANG5_15_0_, this_.CTY_CODE_ISO as CTY6_15_0_, this_.NAME as NAME15_0_, this_.INS_DTIME as INS8_15_0_,this_.UPD_DTIME as UPD9_15_0_, this_.DEL_DTIME as DEL10_15_0_ 
from FI_SALES_LOCATION_V this_ 
where 
(this_.CLASS_TYPE='BU' and this_.CLASS_UNIT_TYPE='STO' and this_.CLASS_UNIT_CODE='A1000') 
and this_.DEL_DTIME is null 
and (this_.LANG_CODE_ISO='s1' and this_.CTY_CODE_ISO='SI' and this_.ID='A1000');

即使写得更多，执行 UNION ALL 而不是 OR 时性能会有所提高。

或者你可以索引表，但是你会在 INSERT 上失去性能。