Oinone远程调用链路源码分析

前提

源码分析版本是 5.1.x版本

概要

在服务启动时,获取注解REMOTE的函数,通过dubbo的泛化调用发布。在调用函数时,通过dubbo泛化调用获取结果。

注册服务者

  1. 在spring 启动方法installOrLoad中初始化
  2. 寻找定义REMOTE的方法
  3. 组装dubbo的服务配置
  4. 组装服务对象实现引用,内容如下,用于注册
    • 调用前置处理
      • 放信息到SessionApi
      • 函数调用链追踪,放到本地TransmittableThreadLocal
      • 从redis中获取到的数据进行反序列化并存在到本地的线程里
      • Trace信息,放一份在sessionApi中 和ThreadLocal
    • 调用函数执行
    • 返回数据转成特定格式
  5. 通过线程组调用dubbo的ServiceConfig.export 服务发布

时序图

Oinone远程调用链路源码分析
注册

源码分析

根据条件判断,确定向dubbo进行服务发布
RemoteServiceLoader

public void publishService(List<FunctionDefinition> functionList,Map<String,Runnable> isPublished) {
        // 因为泛化接口只能控制到namespace,控制粒度不能到fun级别,这里进行去重处理
        Map<String, Function> genericNamespaceMap = new HashMap<>();
        for (FunctionDefinition functionDefinition : functionList) {
            Function function = new Function(functionDefinition)

            try {
               //定义REMOTE, 才给予远程调用
                if (FunctionOpenEnum.REMOTE.in(function.getOpen()) && !ClassUtils.isInterface(function.getClazz())) {
                    genericNamespaceMap.putIfAbsent(RegistryUtils.getRegistryInterface(function), function);
                }
            } catch (PamirsException e) {
            }
        }
        // 发布远程服务
        for (String namespace : genericNamespaceMap.keySet()) {
            Function function = genericNamespaceMap.get(namespace);
            if(isPublished.get(RegistryUtils.getRegistryInterface(function)) == null){
                // 发布,注册远程函数服务,底层使用dubbo的泛化调用
                Runnable registryTask = () -> remoteRegistry.registryService(function);
                isPublished.put(RegistryUtils.getRegistryInterface(function),registryTask);
            }else{

            }
        }
    }

构造ServiceConfig方法,设置成泛化调用,进行发布export()
DefaultRemoteRegistryComponent

     public void registryGenericService(String interfaceName, List<MethodConfig> methods,
                                       String group, String version, Integer timeout, Integer retries) {
        ....
        try {
            ServiceConfig<GenericService> service = new ServiceConfig<>();
            // 服务接口名
            service.setInterface(interfaceName);
            // 服务对象实现引用
            service.setRef(genericService(interfaceName));
            if (null != methods) {
                service.setMethods(methods);
            }
            // 声明为泛化接口
            service.setGeneric(Boolean.TRUE.toString());
            // 基础元数据
            constructService(group, version, timeout, retries, service);
            service.export();
        } catch (Exception e) {
           .....
        }
    }

// 服务对象实现引用
private GenericService genericService(String interfaceName) {
        return (method, parameterTypes, args) -> {
            PamirsSession.clear();
            Function function = Objects.requireNonNull(PamirsSession.getContext()).getFunction(RegistryUtils.getFunctionNamespace(method), RegistryUtils.getFunctionFun(method));
            if (log.isDebugEnabled()) {
                log.debug("interfaceName: " + interfaceName + ", isDataManage: " + function.isDataManager());
            }
            try {
                //前置处理:服务提供者,对请求参数进行对象化拆解,并对请求携带的上下文进行处理
                // 放信息到SessionApi
                // 函数调用链追踪,放到本地TransmittableThreadLocal
                // CommonMetaDataCacheApi.computeMetaData() 从redis中获取到的数据进行反序列化并存在到本地的TL中
                // DataAuditApi.computeDataAuditSession() Trace信息,放一份在sessionApi中 和ThreadLocal中
                Object[] args1 = Spider.getDefaultExtension(RemoteRequestArgApi.class).providerHandle(function.getNamespace(), function.getFun(), args, function.getArguments());

                Object result = FunEngine.get().exclude(ScriptType.REMOTE).run(function, args1);

                //后置处理:服务提供者,对结果进行对象化封装、携带请求上下文进行处理
                return Spider.getDefaultExtension(RemoteResponseApi.class).providerHandle(function, method, result);
            } catch (Throwable e) {
                return Spider.getDefaultExtension(RemoteResponseApi.class).providerExceptionHandle(function, method, e);
            } finally {
                PamirsSession.clear();
            }
        };
    }

注册消费者

  1. 函数处理调用
  2. 注册服务消费者
    • 从ReferenceConfigCache获取泛化
  3. 调用dubbo泛化调用接口
  4. 获取返回信息
    • 获取用户id,放入PamirsSession
    • 如果开启debug模式
      • 存入DEBUG_THREAD_LOCAL本地线程
    • 返回格式
      • IWrapper
      • Pagination
      • Result

时序图

Oinone远程调用链路源码分析

源码分析

泛化调用dobbo接口,并解析返回对象
RemoteComputer

public Object compute(Function function, Object... args) {

        .....
        List<Arg> functionArguments = function.getArguments();
        String methodName = RegistryUtils.getGenericServiceMethodName(function);
        String[] argTypes = FunctionUtils.fetchArgTypes(functionArguments);
        //前置处理:服务消费者,对请求参数进行对象化封装、携带请求上下文进行处理
        Object[] arguments = getRemoteRequestApi().consumerHandle(function.getNamespace(),function.getFun(), args, functionArguments);
        // 泛化调用
        Object result = invoke(function, methodName, argTypes, arguments);
        .....
        //后置处理:服务消费者,对返回结果进行对象化拆解,并对结果携带的上下文进行处理
        // 数据转成IWrapper/Pagination/Result
        return getRemoteResponseApi().consumerHandle(function, result);
}

// 配置请求信息,通过$invoke 实际调用
private Object invoke(Function function, String methodName, String[] argTypes, Object[] arguments) {
        Object result;
        String configCdOwnSign = getSessionFillOwnSignApi().getConfigCdOwnSign();
        if (StringUtils.isBlank(configCdOwnSign)) {
.....
        } else {
            try {
            // 获取服务,由于是泛化调用,所以获取的一定是GenericService类型
                GenericService remoteClient = CommonApiFactory.getApi(RemoteRegistry.class).registryOriginConsumer(function);
         // 第一个参数是需要调用的方法名
         // 第二个参数是需要调用的方法的参数类型数组,为String数组,里面存入参数的全类名。
         // 第三个参数是需要调用的方法的参数数组,为Object数组,里面存入需要的参数
                result = remoteClient.$invoke(methodName, argTypes, arguments);
            } catch (RpcException e) {
                ....            }
        }
        return result;
    }

从缓存中获取泛化
DefaultRemoteRegistryComponent

    public GenericService registryGenericConsumer(String interfaceName, List<MethodConfig> methods,
                                                  String group, String version, Integer timeout, Integer retries) {
        ....
        // 创建服务引用配置
        ReferenceConfig<GenericService> reference = new ReferenceConfig<>();
        reference.setInterface(interfaceName);
        // 设置为泛化调用
        reference.setGeneric(Boolean.TRUE.toString());
        if (null != methods) {
            reference.setMethods(methods);
        }
        constructReference(group, version, timeout, retries, reference);
        return ReferenceConfigCache.getCache().get(reference);
    }

名词解释

泛化调用是指在调用方没有服务方提供的API(SDK)的情况下,对服务方进行调用,并且可以正常拿到调用结果
泛化调用(客户端泛化)
实现泛化实现(服务端泛化)

Oinone社区 作者:oinone原创文章,如若转载,请注明出处:https://doc.oinone.top/backend/17027.html

访问Oinone官网:https://www.oinone.top获取数式Oinone低代码应用平台体验

(0)
oinone的头像oinone
上一篇 2024年9月3日 pm12:53
下一篇 2024年9月5日 pm8:13

相关推荐

  • 【Excel导入/导出】多Sheet导入导出示例(4.7.x及以上版本)

    代码示例 示例仅供参考 点击下载代码示例 业务场景 准备工作:两个模型,物料(Material)和物料类别(MaterialCategory)。目标:在一个Excel模板中同时导入和导出两个模型的数据。 Material模型 @Model.model(Material.MODEL_MODEL) @Model.Advanced(unique = {"code"}) @Model(displayName = "物料", labelFields = {"name"}) public class Material extends IdModel { private static final long serialVersionUID = -2594216864389636135L; public static final String MODEL_MODEL = "maas.Material"; @Field.String @Field(displayName = "物料编码", required = true) private String code; @Field.String @Field(displayName = "物料名称", required = true) private String name; } MaterialCategory模型 @Model.model(MaterialCategory.MODEL_MODEL) @Model.Advanced(unique = {"code"}) @Model(displayName = "物料类别", labelFields = {"name"}) public class MaterialCategory extends IdModel { private static final long serialVersionUID = 6300896634558908349L; public static final String MODEL_MODEL = "maas.MaterialCategory"; @Field.String @Field(displayName = "类别编码", required = true) private String code; @Field.String @Field(displayName = "类别名称", required = true) private String name; } 模板定义 MaterialTemplate @Component public class MaterialTemplate implements ExcelTemplateInit { public static final String TEMPLATE_NAME = "materialTemplate"; @Override public List<ExcelWorkbookDefinition> generator() { WorkbookDefinitionBuilder builder = WorkbookDefinitionBuilder.newInstance(Material.MODEL_MODEL, TEMPLATE_NAME) .setDisplayName("物料和物料类别") .setEachImport(Boolean.FALSE);//设置importData的入参为 (ExcelImportContext importContext, List<MaterialCategory> data),如入参是单个对象,请删除setEachImport(Boolean.FALSE) createMaterialSheet(builder); createMaterialCategorySheet(builder); return Collections.singletonList(builder.build()); } private static void createMaterialSheet(WorkbookDefinitionBuilder builder) { builder.createSheet().setName("物料") .createBlock(Material.MODEL_MODEL, ExcelAnalysisTypeEnum.FIXED_HEADER, ExcelDirectionEnum.HORIZONTAL, "A1:B2") .createHeader().setStyleBuilder(ExcelHelper.createDefaultStyle()).setIsConfig(Boolean.TRUE) .createCell().setField("code").setAutoSizeColumn(Boolean.TRUE).and() .createCell().setField("name").setAutoSizeColumn(Boolean.TRUE).and() .and() .createHeader().setStyleBuilder(ExcelHelper.createDefaultStyle(v -> v.setBold(Boolean.TRUE)).setHorizontalAlignment(ExcelHorizontalAlignmentEnum.CENTER)) .createCell().setValue("物料编码").and() .createCell().setValue("物料名称"); } private static void createMaterialCategorySheet(WorkbookDefinitionBuilder builder) { builder.createSheet().setName("物料类别") .createBlock(MaterialCategory.MODEL_MODEL, ExcelAnalysisTypeEnum.FIXED_HEADER, ExcelDirectionEnum.HORIZONTAL, "A1:B2") .createHeader().setStyleBuilder(ExcelHelper.createDefaultStyle()).setIsConfig(Boolean.TRUE) .createCell().setField("code").setAutoSizeColumn(Boolean.TRUE).and()…

    2024年4月23日
    1.3K00
  • 字段类型之关系描述的特殊场景(常量关联)

    场景概述 【字段类型之关系与引用】一文中已经描述了各种关系字段的常规写法,还有一些特殊场景如:关系映射中存在常量,或者M2M中间表是大于两个字段构成。 场景描述 1、PetTalent模型增加talentType字段2、PetItem与PetTalent的多对多关系增加talentType(达人类型),3、PetItemRelPetTalent中间表维护petItemId、petTalentId以及talentType,PetDogItem和PetCatItem分别重写petTalents字段,关系中增加常量描述。示意图如下: 实际操作步骤 Step1 新增 TalentTypeEnum package pro.shushi.pamirs.demo.api.enumeration; import pro.shushi.pamirs.meta.annotation.Dict; import pro.shushi.pamirs.meta.common.enmu.BaseEnum; @Dict(dictionary = TalentTypeEnum.DICTIONARY,displayName = "达人类型") public class TalentTypeEnum extends BaseEnum<TalentTypeEnum,Integer> { public static final String DICTIONARY ="demo.TalentTypeEnum"; public final static TalentTypeEnum DOG =create("DOG",1,"狗达人","狗达人"); public final static TalentTypeEnum CAT =create("CAT",2,"猫达人","猫达人"); } Step2 PetTalent模型增加talentType字段 package pro.shushi.pamirs.demo.api.model; import pro.shushi.pamirs.demo.api.enumeration.TalentTypeEnum; import pro.shushi.pamirs.meta.annotation.Field; import pro.shushi.pamirs.meta.annotation.Model; @Model.model(PetTalent.MODEL_MODEL) @Model(displayName = "宠物达人",summary="宠物达人",labelFields ={"name"}) public class PetTalent extends AbstractDemoIdModel{ public static final String MODEL_MODEL="demo.PetTalent"; @Field(displayName = "达人") private String name; @Field(displayName = "达人类型") private TalentTypeEnum talentType; } Step3 修改PetItem的petTalents字段,在关系描述中增加talentType(达人类型) @Field.many2many(relationFields = {"petItemId"},referenceFields = {"petTalentId","talentType"},through = PetItemRelPetTalent.MODEL_MODEL ) @Field.Relation(relationFields = {"id"}, referenceFields = {"id","talentType"}) @Field(displayName = "推荐达人",summary = "推荐该商品的达人们") private List<PetTalent> petTalents; Step4 PetDogItem增加petTalents字段,重写父类PetItem的关系描述 talentType配置为常量,填入枚举的值 增加domain描述用户页面选择的时候自动过滤出特定类型的达人,RSQL用枚举的name @Field(displayName = "推荐达人") @Field.many2many( through = "PetItemRelPetTalent", relationFields = {"petItemId"}, referenceFields = {"petTalentId","talentType"} ) @Field.Relation(relationFields = {"id"}, referenceFields = {"id", "#1#"}, domain = " talentType == DOG") private List<PetTalent> petTalents; Step5 PetCatItem增加petTalents字段,重写父类PetItem的关系描述 talentType配置为常量,填入枚举的值 增加domain描述用户页面选择的时候自动过滤出特定类型的达人,RSQL用枚举的name @Field(displayName = "推荐达人") @Field.many2many( through = "PetItemRelPetTalent", relationFields = {"petItemId"}, referenceFields = {"petTalentId","talentType"} ) @Field.Relation(relationFields = {"id"}, referenceFields = {"id", "#2#"}, domain = " talentType == CAT") private List<PetTalent> petTalents; Step6 PetCatItem增加petTalents字段,many2one关系示例 talentType配置为常量,填入枚举的值 增加domain描述用户页面选择的时候自动过滤出特定类型的达人,RSQL用枚举的name @Model.model(PetPet.MODEL_MODEL) @Model(displayName…

    2024年5月25日
    1.9K00
  • 【PostgreSQL】后端部署使用PostgreSQL数据库

    PostgreSQL数据库配置 驱动配置 Maven配置(14.3版本可用) <postgresql.version>42.6.0</postgresql.version> <dependency> <groupId>org.postgresql</groupId> <artifactId>postgresql</artifactId> <version>${postgresql.version}</version> </dependency> 离线驱动下载 postgresql-42.2.18.jarpostgresql-42.6.0.jarpostgresql-42.7.3.jar JDBC连接配置 pamirs: datasource: base: type: com.alibaba.druid.pool.DruidDataSource driverClassName: org.postgresql.Driver url: jdbc:postgresql://127.0.0.1:5432/pamirs?currentSchema=base username: xxxxxx password: xxxxxx 连接url配置 暂无官方资料 url格式 jdbc:postgresql://${host}:${port}/${database}?currentSchema=${schema} 在jdbc连接配置时,${database}和${schema}必须完整配置,不可缺省。 其他连接参数如需配置,可自行查阅相关资料进行调优。 方言配置 pamirs方言配置 pamirs: dialect: ds: base: type: PostgreSQL version: 14 major-version: 14.3 pamirs: type: PostgreSQL version: 14 major-version: 14.3 数据库版本 type version majorVersion 14.x PostgreSQL 14 14.3 PS:由于方言开发环境为14.3版本,其他类似版本(14.x)原则上不会出现太大差异,如出现其他版本无法正常支持的,可在文档下方留言。 schedule方言配置 pamirs: event: enabled: true schedule: enabled: true dialect: type: PostgreSQL version: 14 major-version: 14.3 type version majorVersion PostgreSQL 14 14.3 PS:由于schedule的方言在多个版本中并无明显差异,目前仅提供一种方言配置。 其他配置 逻辑删除的值配置 pamirs: mapper: global: table-info: logic-delete-value: (EXTRACT(epoch FROM CURRENT_TIMESTAMP) * 1000000 + EXTRACT(MICROSECONDS FROM CURRENT_TIMESTAMP))::bigint PostgreSQL数据库用户初始化及授权 — init root user (user name can be modified by oneself) CREATE USER root WITH PASSWORD 'password'; — if using automatic database and schema creation, this is very important. ALTER USER root CREATEDB; SELECT * FROM pg_roles; — if using postgres database, this authorization is required. GRANT CREATE ON DATABASE postgres TO root;

    2023年11月1日
    1.4K00
  • 如何自定义覆盖内置模块的页面

    1.首先通过sql查询找到我们需要的页面,从其中的template字段复制出原视图的配置 通过模型编码model在base_view查找需要修改的视图 select * from base_view where model='workflow.WorkflowUserTask' and is_deleted = 0; 2.将base_view的template内容复制到java的core工程的resources目录下新建一个xml文件,修改里面的动作名称 <view widget="WorkFlowImplement"> <template slot="actions"> <action name="$$internal_GotoListTableRouter" priority="1" model="workflow.WorkflowUserTask" tag="contextFreeAction"/> <action name="approveStaging" widget="FlowTaskCommonAction" invisible="!(activeRecord.allowStaging)" priority="2" label="新暂存" model="workflow.WorkflowUserTask" displayName="新暂存"/> <action name="workflow_agree" invisible="!(activeRecord.allowAgree && activeRecord.status == &apos;ACTIVE&apos;)" priority="3" label="新同意" model="workflow.WorkflowUserTask" load="fetchDetailReadOnly" displayName="新同意" goBack="true" validateForm="true" loadRootData="true"/> <action name="workflow_rejust" invisible="!(activeRecord.allowReject && activeRecord.status == &apos;ACTIVE&apos;)" priority="4" label="新拒绝" model="workflow.WorkflowUserTask" displayName="新拒绝" goBack="true" loadRootData="true" tag="contextFreeAction"/> <action name="workflow_turnon" invisible="!(activeRecord.taskType == &apos;APPROVE&apos; && activeRecord.allowTransfer && activeRecord.status == &apos;ACTIVE&apos;)" priority="5" label="新转交" model="workflow.WorkflowUserTask" load="fetchDetailReadOnly" displayName="新转交" goBack="true" loadRootData="true" tag="contextFreeAction"/> <action name="workflow_addsign" invisible="!(activeRecord.taskType == &apos;APPROVE&apos; && activeRecord.allowAddSign && activeRecord.status == &apos;ACTIVE&apos;)" priority="6" label="新加签" model="workflow.WorkflowUserTask" load="fetchDetailReadOnly" displayName="新加签" goBack="true" loadRootData="true" tag="contextFreeAction"/> <action name="workflow_write_fallback" invisible="!(activeRecord.taskType == &apos;WRITE&apos; && activeRecord.allowFallback && activeRecord.status == &apos;ACTIVE&apos;)" priority="7" label="新回退" model="workflow.WorkflowUserTask" displayName="新回退" goBack="true" loadRootData="true" tag="contextFreeAction"/> <action name="workflow_sharing" priority="8" label="新分享" model="workflow.WorkflowUserTask" displayName="新分享" goBack="true" loadRootData="true" tag="contextFreeAction"/> </template> <template slot="fields"> <field name="remark" widget="TextArea" invisible="activeRecord.taskType != &apos;APPROVE&apos;" priority="8" model="workflow.WorkflowUserTask" data="remark" displayName="意见备注"/> </template> </view> 3.在生命周期的元数据编辑方法内覆盖视图 package pro.shushi.pamirs.demo.core.init; import org.springframework.core.Ordered; import org.springframework.core.annotation.Order; import org.springframework.stereotype.Component; import pro.shushi.pamirs.boot.common.api.command.AppLifecycleCommand; import pro.shushi.pamirs.boot.common.extend.MetaDataEditor; import pro.shushi.pamirs.core.common.InitializationUtil; import pro.shushi.pamirs.demo.api.DemoModule; import pro.shushi.pamirs.meta.annotation.fun.extern.Slf4j; import pro.shushi.pamirs.meta.api.dto.meta.Meta; import pro.shushi.pamirs.meta.enmu.ViewTypeEnum; import pro.shushi.pamirs.workflow.app.api.model.WorkflowUserTask; import java.util.Map; @Slf4j @Component @Order(Ordered.LOWEST_PRECEDENCE) public class DemoModuleMetaInstall implements MetaDataEditor { @Override public void edit(AppLifecycleCommand command,…

    2024年7月2日
    1.6K00
  • 查询时自定义排序字段和排序规则

    指定字段排序 平台默认排序字段,参考IdModel,按创建时间和ID倒序(ordering = "createDate DESC, id DESC") 方法1:模型指定排序 模型定义增加排序字段。@Model.Advanced(ordering = "xxxxx DESC, yyyy DESC") @Model.model(PetShop.MODEL_MODEL) @Model(displayName = "宠物店铺",summary="宠物店铺",labelFields ={"shopName"}) @Model.Code(sequence = "DATE_ORDERLY_SEQ",prefix = "P",size=6,step=1,initial = 10000,format = "yyyyMMdd") @Model.Advanced(ordering = "createDate DESC") public class PetShop extends AbstractDemoIdModel { public static final String MODEL_MODEL="demo.PetShop"; // ………… } 方法2:Page查询中可以自定排序规则 API参考 pro.shushi.pamirs.meta.api.dto.condition.Pagination#orderBy public <G, R> Pagination<T> orderBy(SortDirectionEnum direction, Getter<G, R> getter) { if (null == getSort()) { setSort(new Sort()); } getSort().addOrder(direction, getter); return this; } 具体示例 @Function.Advanced(type= FunctionTypeEnum.QUERY) @Function.fun(FunctionConstants.queryPage) @Function(openLevel = {FunctionOpenEnum.API}) public Pagination<PetShop> queryPage(Pagination<PetShop> page, IWrapper<PetShop> queryWrapper){ page.orderBy(SortDirectionEnum.DESC, PetShop::getCreateDate); page = new PetShop().queryPage(page, queryWrapper); return page; } 方法3:查询的wapper中指定 API参考:pro.shushi.pamirs.framework.connectors.data.sql.AbstractWrapper#orderBy @Override public Children orderBy(boolean condition, boolean isAsc, R… columns) { if (ArrayUtils.isEmpty(columns)) { return typedThis; } SqlKeyword mode = isAsc ? ASC : DESC; for (R column : columns) { doIt(condition, ORDER_BY, columnToString(column), mode); } return typedThis; } 具体示例 public List<PetShop> queryList(String name) { List<PetShop> petShops = Models.origin().queryListByWrapper( Pops.<PetShop>lambdaQuery().from(PetShop.MODEL_MODEL) .orderBy(true, true, PetShop::getCreateDate) .orderBy(true, true, PetShop::getId) .like(PetShop::getShopName, name)); return petShops; } 设置查询不排序 方法1:关闭平台默认排序字段,设置模型的ordering,改成:ordering = "1=1" 模型定义增加排序字段。@Model.Advanced(ordering = "1=1") @Model.model(PetShop.MODEL_MODEL) @Model(displayName = "宠物店铺",summary="宠物店铺",labelFields ={"shopName"}) @Model.Code(sequence = "DATE_ORDERLY_SEQ",prefix = "P",size=6,step=1,initial = 10000,format = "yyyyMMdd") @Model.Advanced(ordering =…

    2024年5月25日
    2.0K00

Leave a Reply

登录后才能评论