如何在Spring for GraphQL中结合JPA+QueryDSL实现按需查询字段

问题场景 #

你使用Spring for GraphQL结合Querydsl查询员工数据，当执行仅请求firstName的GraphQL查询时：

query {
  employees {
    firstName
  }
}

实际SQL却查询了id、first_name、last_name所有字段，需要实现SQL仅返回GraphQL选择集对应的字段。

核心原因 #

QuerydslPredicateExecutor默认会查询实体的全部字段，要实现字段裁剪，需要结合Spring Data投影和Spring for GraphQL的选择集解析能力，让底层框架根据选择集生成针对性的SQL。

具体解决方案 #

1. 定义投影接口 #

创建与GraphQL字段匹配的投影接口，仅包含需要查询的字段：

// 基础投影接口，仅包含firstName
public interface EmployeeProjection {
    String getFirstName();
}

// 扩展投影接口，支持全字段查询（按需定义）
interface EmployeeFullProjection extends EmployeeProjection {
    Integer getId();
    String getLastName();
}

投影接口的方法名必须和实体类的getter方法完全对应（比如getFirstName对应EmployeeEntity的firstName属性）。

2. 调整仓库接口 #

在EmployeeRepository中添加支持投影的查询方法，让Querydsl可以返回投影实例：

@GraphQlRepository
public interface EmployeeRepository extends JpaRepository<EmployeeEntity, Integer>,
        QuerydslPredicateExecutor<EmployeeEntity> {
    // 支持动态投影的通用查询方法
    <T> Iterable<T> findAll(Predicate predicate, Class<T> projectionType);
}

3. 改造数据获取器，利用投影实现字段裁剪 #

直接借助QuerydslDataFetcher的内置投影支持，它会自动解析GraphQL选择集并生成对应SQL：

@Controller
public class EmployeeController {

    private final EmployeeRepository employeeRepository;
    private final EmployeeMapper employeeMapper;

    // 构造注入替代@Autowired（更符合Spring规范）
    public EmployeeController(EmployeeRepository employeeRepository, EmployeeMapper employeeMapper) {
        this.employeeRepository = employeeRepository;
        this.employeeMapper = employeeMapper;
    }

    @QueryMapping
    List<Employee> employees(DataFetchingEnvironment environment) {
        // 指定投影接口，QuerydslDataFetcher会自动根据选择集裁剪SQL字段
        Iterable<EmployeeProjection> projections = QuerydslDataFetcher.builder(employeeRepository)
                .projection(EmployeeProjection.class)
                .many()
                .get(environment);

        return StreamSupport.stream(projections.spliterator(), false)
                .map(proj -> new Employee(null, proj.getFirstName(), null))
                .toList();
    }

    record Employee(Integer id, String firstName, String lastName) {}
}

如果需要支持多种字段组合的动态场景，可以根据选择集动态切换投影类型：

@QueryMapping
List<Employee> employees(DataFetchingEnvironment environment) {
    // 提取GraphQL请求中选择的字段
    List<String> selectedFields = environment.getSelectionSet().getFields().stream()
            .map(Field::getName)
            .toList();

    // 根据选择的字段组合确定使用的投影类型
    Class<? extends EmployeeProjection> projectionType = determineProjectionType(selectedFields);

    Iterable<? extends EmployeeProjection> projections = QuerydslDataFetcher.builder(employeeRepository)
            .projection(projectionType)
            .many()
            .get(environment);

    // 转换为DTO，按需填充字段
    return StreamSupport.stream(projections.spliterator(), false)
            .map(this::convertToEmployeeDto)
            .toList();
}

private Class<? extends EmployeeProjection> determineProjectionType(List<String> selectedFields) {
    if (selectedFields.equals(List.of("firstName"))) {
        return EmployeeProjection.class;
    } else if (selectedFields.containsAll(List.of("id", "firstName", "lastName"))) {
        return EmployeeFullProjection.class;
    }
    // 可扩展更多字段组合的投影类型
    return EmployeeProjection.class;
}

private Employee convertToEmployeeDto(EmployeeProjection projection) {
    Integer id = (projection instanceof EmployeeFullProjection) ? ((EmployeeFullProjection) projection).getId() : null;
    String lastName = (projection instanceof EmployeeFullProjection) ? ((EmployeeFullProjection) projection).getLastName() : null;
    return new Employee(id, projection.getFirstName(), lastName);
}

效果验证 #

修改后执行目标GraphQL查询，SQL日志会显示仅查询first_name字段：

SELECT first_name FROM employee

内容的提问来源于stack exchange，提问作者Arun Chandrasekaran