从graphql-java-kickstart迁移至Spring GraphQL时，如何重构DataLoader依赖配置？

问题场景 #

从graphql-java-kickstart迁移至Spring GraphQL，原项目中存在一个DataLoader作为另一个DataLoader方法参数的场景（核心逻辑需保留，后续再解耦重构）。原kickstart配置可直接传递已定义的DataLoader实例，但Spring GraphQL使用BatchLoaderRegistry配置时，无法直接从Lambda中获取其他已注册的DataLoader。

原kickstart核心代码 #

@Component
public class RequestContextBuilder extends DefaultGraphQLServletContextBuilder {

    @Override
    public GraphQLKickstartContext build(HttpServletRequest request, HttpServletResponse response) {
        
        Map<Object, Object> map = new HashMap<>();
        // 其他上下文初始化逻辑
        return GraphQLKickstartContext.of(buildDataLoaderRegistry(), map);
    }
    
    private DataLoaderRegistry buildDataLoaderRegistry() {

        var noBatching = new DataLoaderOptions().setBatchingEnabled(false);
        var dataLoaderRegistry = new DataLoaderRegistry();

        // 注册第一个DataLoader
        var firstLoader = DataLoaderFactory.newDataLoader(
                ids1 -> supplyAsync(() -> firstService.getItems(ids1), traceableExecutorService),
                noBatching
        );
        dataLoaderRegistry.register("first", firstLoader);

        // 注册第二个DataLoader，并直接传入firstLoader实例
        var secondLoader = DataLoaderFactory.newDataLoader(
                ids2 -> supplyAsync(() -> secondService.loadUsingOtherLoader(ids2, firstLoader), traceableExecutorService),
                noBatching
        );
        dataLoaderRegistry.register("second", secondLoader);

        return dataLoaderRegistry;
    }
}

@Service
public class SecondService {
  
  public List<Item> loadUsingOtherLoader(List<ID> ids, DataLoader<ID, Item> firstLoader) {
    
    return ids.stream()
              .map(firstLoader::load)
              // 业务处理逻辑
              .toList();
  }
}

Spring GraphQL当前配置问题 #

@Configuration
public class RequestContextBuilder {

    public RequestContextBuilder(BatchLoaderRegistry registry, FirstService firstService, SecondService secondService) {
    
        registry.forName("first")
                .withOptions(BATCHING_DISABLED)
                .registerBatchLoader((ids, env) -> Flux.fromStream(ids.stream()).map(firstService::getItems));
        
        // 无法直接获取firstLoader传递给secondService
        registry.forName("second")
                .withOptions(BATCHING_DISABLED)
                .registerBatchLoader((ids, env) -> Flux.fromStream(ids.stream()).map(secondService::loadUsingOtherLoader));
        }
    }

解决方案 #

利用Spring GraphQL提供的BatchLoaderEnvironment中的DataLoaderAccessor，可以获取已注册的DataLoader实例，无需大幅修改原有业务代码。

方案1：修改SecondService参数为DataLoaderAccessor #

@Service
public class SecondService {
  
  public List<Item> loadUsingOtherLoader(List<ID> ids, DataLoaderAccessor dataLoaderAccessor) {
    // 通过名称获取已注册的firstLoader
    DataLoader<ID, Item> firstLoader = dataLoaderAccessor.getDataLoader("first");
    return ids.stream()
              .map(firstLoader::load)
              // 业务处理逻辑
              .toList();
  }
}

然后在配置类中从环境取出DataLoaderAccessor传递给方法：

@Configuration
public class DataLoaderConfig {

    public DataLoaderConfig(BatchLoaderRegistry registry, FirstService firstService, SecondService secondService) {
    
        registry.forName("first")
                .withOptions(BATCHING_DISABLED)
                .registerBatchLoader((ids, env) -> Flux.fromIterable(firstService.getItems(ids)));
        
        registry.forName("second")
                .withOptions(BATCHING_DISABLED)
                .registerBatchLoader((ids, env) -> {
                    // 从环境中获取DataLoaderAccessor
                    DataLoaderAccessor accessor = env.getDataLoaderAccessor();
                    return Flux.fromIterable(secondService.loadUsingOtherLoader(ids, accessor));
                });
        }
    }

方案2：不修改SecondService，直接在配置中传递DataLoader #

如果不想改动SecondService的方法签名，可以在BatchLoader的Lambda中先获取firstLoader，再传入方法：

@Configuration
public class DataLoaderConfig {

    public DataLoaderConfig(BatchLoaderRegistry registry, FirstService firstService, SecondService secondService) {
    
        registry.forName("first")
                .withOptions(BATCHING_DISABLED)
                .registerBatchLoader((ids, env) -> Flux.fromIterable(firstService.getItems(ids)));
        
        registry.forName("second")
                .withOptions(BATCHING_DISABLED)
                .registerBatchLoader((ids, env) -> {
                    // 直接获取firstLoader实例
                    DataLoader<ID, Item> firstLoader = env.getDataLoaderAccessor().getDataLoader("first");
                    return Flux.fromIterable(secondService.loadUsingOtherLoader(ids, firstLoader));
                });
        }
    }

两种方案都能在保留原有业务逻辑的前提下，实现Spring GraphQL中DataLoader的跨依赖调用，满足迁移初期的需求。

内容的提问来源于stack exchange，提问作者Michiel