相同Spring Boot环境下JPQL查询缺失ID字段的原因排查

在JPA Repository中定义了如下数据查询方法：

@Query(
      "SELECT s FROM Site s "
          + "LEFT JOIN FETCH s.siteType "
          + "LEFT JOIN FETCH s.variables "
          + "LEFT JOIN FETCH s.parameters")
  List<Site> findWithDetails();

本地与同事环境均使用Spring Boot 2.7.0（内置Spring Data 2.7.0）版本，本地环境下该JPQL可正常翻译为完整SQL，所有实体的ID字段、关联外键字段均被正常查询，SQL可正常执行，翻译生成的SQL如下：

select
    site0_.id_site as id_site1_37_0_,
    sitetype1_.id_site_type as id_site_1_39_1_,
    variables2_.id_variable as id_varia1_50_2_,
    parameters3_.id_site_param as id_site_1_38_3_,
    site0_.active as active2_37_0_,
    site0_.coord as coord3_37_0_,
    site0_.name as name4_37_0_,
    site0_.short_name as short_na5_37_0_,
    site0_.id_site_type as id_site_7_37_0_,
    site0_.source as source6_37_0_,
    sitetype1_.display_name as display_2_39_1_,
    sitetype1_.marker_default_color as marker_d3_39_1_,
    sitetype1_.marker_fa_icon as marker_f4_39_1_,
    sitetype1_.marker_fa_icon_color as marker_f5_39_1_,
    sitetype1_.marker_fa_icon_prefix as marker_f6_39_1_,
    sitetype1_.marker_shape as marker_s7_39_1_,
    sitetype1_.site_type as site_typ8_39_1_,
    variables2_.active as active2_50_2_,
    variables2_.description as descript3_50_2_,
    variables2_.id_external_ts_source as id_exter9_50_2_,
    variables2_.frequency as frequenc4_50_2_,
    variables2_.name as name5_50_2_,
    variables2_.popup_display_order as popup_di6_50_2_,
    variables2_.short_name as short_na7_50_2_,
    variables2_.id_site as id_site10_50_2_,
    variables2_.unit_type as unit_typ8_50_2_,
    variables2_.id_variable_type as id_vari11_50_2_,
    variables2_.id_site as id_site10_50_0__,
    variables2_.id_variable as id_varia1_50_0__,
    parameters3_.name as name2_38_3_,
    parameters3_.id_site as id_site4_38_3_,
    parameters3_.value as value3_38_3_,
    parameters3_.id_site as id_site4_38_1__,
    parameters3_.id_site_param as id_site_1_38_1__
from
    site site0_
left outer join site_type sitetype1_ on
    site0_.id_site_type = sitetype1_.id_site_type
left outer join variable variables2_ on
    site0_.id_site = variables2_.id_site
left outer join site_param parameters3_ on
    site0_.id_site = parameters3_.id_site

同事电脑上同一段JPQL翻译生成的SQL缺失了5个关键ID相关字段，导致查询无法正常执行，同事环境生成的SQL如下：

select
    site0_.id_site as id_site1_37_0_,
    sitetype1_.id_site_type as id_site_1_39_1_,
    variables2_.id_variable as id_varia1_50_2_,
    parameters3_.id_site_param as id_site_1_38_3_,
    site0_.active as active2_37_0_,
    site0_.coord as coord3_37_0_,
    site0_.name as name4_37_0_,
    site0_.short_name as short_na5_37_0_,
    site0_.source as source6_37_0_,
    sitetype1_.display_name as display_2_39_1_,
    sitetype1_.marker_default_color as marker_d3_39_1_,
    sitetype1_.marker_fa_icon as marker_f4_39_1_,
    sitetype1_.marker_fa_icon_color as marker_f5_39_1_,
    sitetype1_.marker_fa_icon_prefix as marker_f6_39_1_,
    sitetype1_.marker_shape as marker_s7_39_1_,
    sitetype1_.site_type as site_typ8_39_1_,
    variables2_.active as active2_50_2_,
    variables2_.description as descript3_50_2_,
    variables2_.frequency as frequenc4_50_2_,
    variables2_.name as name5_50_2_,
    variables2_.popup_display_order as popup_di6_50_2_,
    variables2_.short_name as short_na7_50_2_,
    variables2_.unit_type as unit_typ8_50_2_,
    variables2_.id_site as id_site10_50_0__,
    variables2_.id_variable as id_varia1_50_0__,
    parameters3_.name as name2_38_3_,
    parameters3_.value as value3_38_3_,
    parameters3_.id_site as id_site4_38_1__,
    parameters3_.id_site_param as id_site_1_38_1__
from
    site site0_
left outer join site_type sitetype1_ on
    site0_.id_site_type = sitetype1_.id_site_type
left outer join variable variables2_ on
    site0_.id_site = variables2_.id_site
left outer join site_param parameters3_ on
    site0_.id_site = parameters3_.id_site

缺失的字段具体为：

site0_.id_site_type as id_site_7_37_0_,
variables2_.id_external_ts_source as id_exter9_50_2_,
variables2_.id_site as id_site10_50_2_,
variables2_.id_variable_type as id_vari11_50_2_,
parameters3_.id_site as id_site4_38_3_,

目前已确认双方Spring Boot框架版本完全一致，不确定是否存在配置缺失、依赖差异或其他问题导致ID字段被忽略。

• 优先清理编译缓存：这类同代码同版本但SQL生成不一致的问题，90%以上是本地编译缓存脏数据导致。直接删除同事本地项目的target/build编译输出目录，IDEA用户执行File -> Invalidate Caches...勾选清除缓存重启，拉取最新代码后执行全量编译，再验证SQL生成是否正常。同时核对双方实体类代码，确认缺失字段对应的关联注解（@ManyToOne/@OneToMany、@JoinColumn）配置完全一致，没有多余的@Transient注解、异常的insertable/updatable配置。
• 核对全量依赖树：不要只核对Spring Boot版本，执行mvn dependency:tree（Maven）或gradle dependencies（Gradle）导出双方依赖树对比，重点检查hibernate-core、字节码增强相关依赖（byte-buddy、javassist）、Lombok版本是否一致，排查是否有第三方依赖引入了低版本Hibernate、或者自定义JPA增强插件干扰了实体字段解析。
• 对比本地配置差异：逐行核对双方的application.yml/application.properties配置，重点检查Hibernate相关配置：是否有自定义的hibernate.integrator_provider、实体拦截器、EntityTuplizer，是否异常开启了字节码增强相关配置，是否存在本地自定义的SQL格式化/拦截逻辑过滤了查询字段。
• 检查IDE配置：确认同事IDE的注解处理器已正常开启，Lombok、JPA相关注解处理器配置与本地一致，IDEA用户检查JPA Facet是否正确识别了所有实体类和关联字段，没有误标记字段为非持久化字段。
• 临时规避方案：如果短时间找不到根因，可以修改JPQL显式声明需要查询的关联ID字段，强制Hibernate在SQL中查询这些字段，绕过自动字段选择的逻辑异常。

内容的提问来源于stack exchange，提问作者Francois