Blink 会话窗口解决双流关联ID关联不上问题

一、实现方案

表1 是用户进店记录

表2 是用户留资信息 

业务：用户留资在前，然后才会有跟进记录，数据上留资时间必然在进店跟进之前，最小相差时间毫秒级别。

通过将留资信息实时同步holo，利用holo做维表的方式，进店记录表通过会话窗口延迟5分钟关联确保数据完全能关联到不会丢失。

二、代码实现

--SQL
--********************************************************************--
--Author: mdp_prd
--CreateTime: 2023-05-05 11:51:38
--Comment: 事件引擎 展厅接待
--********************************************************************--
--声明组织区域维度表 dealer_shop_a_d
CREATE TABLE dealer_shop_a_d(
  dealer_shop_no      VARCHAR COMMENT '专营店编码,长度默认为50',
  province_name       VARCHAR COMMENT '省份名称',
  city_name           VARCHAR COMMENT '城市名称',
  warzone_name        VARCHAR COMMENT '战区名称',
  latest_shop_no      VARCHAR COMMENT '最新的专营店编码,长度默认为50',
  short_name          VARCHAR COMMENT '专营店简称',
  group_name_full   VARCHAR COMMENT '集团店全称',
  group_name        VARCHAR COMMENT '集团简称',
  oz_name_full      VARCHAR COMMENT '区域店全称',
  oz_name           VARCHAR  COMMENT '区域简称',
  category_name   VARCHAR COMMENT '店面类别',
  investor_name   VARCHAR COMMENT '投资人',
  name            VARCHAR Comment '专营店名称',
  PRIMARY KEY (dealer_shop_no),
  PERIOD FOR SYSTEM_TIME
) with (
  type = 'odps',
  endPoint = 'http://阿里云odpsendpoint.cn/api',
  project = 'dwd',
  accessId = '*******************',
  accessKey = '&&&&&&&&&&&&&&&&&&&&&&&&&',
  tableName = 'dealer_shop_a_d',
  maxRowCount = '99999999',
  `partition` = 'data_date=max_pt()',
  cache = 'ALL',
  cacheTTLMs = '3600000'
);

--声明dict码表 follow_sys_dict_a_d
CREATE TABLE follow_sys_dict_a_d(
    id          varchar COMMENT 'ID',
    dict_type   BIGINT COMMENT '字典分组',
    dict_code   BIGINT COMMENT '字典明细项Code',
    dict_name   varchar COMMENT '字典明细项名称',
    dict_sort   BIGINT COMMENT '排序',
    p_code      BIGINT COMMENT '父级字典 关联主表中的dict_code字段 默认0表示无父级节点',
    remark      varchar COMMENT '备注',
    application varchar COMMENT '字典所属系统,默认为空,所有系统都可使用',
    version     BIGINT COMMENT '版本',
    create_time varchar COMMENT '创建时间',
    create_by   varchar COMMENT '创建人',
    update_time varchar COMMENT '更新时间',
    update_by   varchar COMMENT '更新人',
  PRIMARY KEY (dict_code),
  PERIOD FOR SYSTEM_TIME
) with (
  type = 'odps',
  endPoint = 'http://阿里云odpsendpoint.cn/api',
  project = 'ads_rpt',
  accessId = '***************',
  accessKey = '&&&&&&&&&&&&&&&&&&&&&',
  tableName = 'follow_sys_dict_a_d',
  maxRowCount = '99999999',
  `partition` = 'data_date=max_pt()',
  cache = 'ALL',
  cacheTTLMs = '3600000'
);

--声明SCRM商机表 tb_scrm_opportunity_d
CREATE TABLE tb_scrm_opportunity_d(
  id VARCHAR        --潜客表ID
  ,shop_id VARCHAR  --店铺ID
  ,opportunity_type BIGINT  --线索类型（线上/线下）
  ,brand_name VARCHAR
  ,model_name VARCHAR
  ,model_code VARCHAR
  ,customer_name VARCHAR
  ,customer_mobile_phone VARCHAR
  ,PRIMARY KEY (id),
  PERIOD FOR SYSTEM_TIME
) with (
  type = 'hologres',
  dbname = 'ods',
  tablename = 'scrm.tb_scrm_opportunity_d',
  username = '********************',
  password = '&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&',
  endpoint = '阿里云hologres的url:80',
  cache = 'None',
  cacheSize = '100000',
  cacheTTLMs = '60000',
  async = 'false'  
);

--声明车型维表
CREATE TABLE brand_model_a_d (
  model_code      VARCHAR COMMENT '车型编码',
  model_name       VARCHAR COMMENT '车型',
  PRIMARY KEY (model_code),
  PERIOD FOR SYSTEM_TIME
) with (
  type = 'odps',
  endPoint = 'http://阿里云odps.cn/api',
  project = 'dwd',
  accessId = '***************',
  accessKey = '&&&&&&&&&&&&&&&&&&&&&&&&&&&',
  tableName = 'brand_model_a_d ',
  maxRowCount = '99999999',
  `partition` = 'data_date=max_pt()',
  cache = 'ALL',
  cacheTTLMs = '3600000'
);

--声明SCRM FoLLOW
CREATE TABLE follow_p(
  id VARCHAR  Comment '跟进单号'      --ID
  ,opportunity_id VARCHAR comment '商机ID'
  ,shop_id VARCHAR Comment '虚拟店ID'
  ,follow_time TIMESTAMP    Comment '实际跟进时间'
  ,follow_mode BIGINT comment '跟进方式编码'
  ,follow_result_code BIGINT comment '跟进结果编码'
  ,sa_name VARCHAR comment '销售顾问名称'
  ,event_type BIGINT comment '事件类型'
  ,new_dts_sync_dts_after_flag  VARCHAR      --所有列的值是否更新后的值取值：Y或N。
  ,new_dts_sync_dts_operation_flag  VARCHAR  --操作类型取值：I：INSERT操作。D：DELETE操作。U：UPDATE操作。
--这里通过process time 时间窗口延迟流数据源5分钟的作用，在5分02秒是触发计算 
,ts as proctime()
（,ts as TO_TIMESTAMP (NOW () * 1000),
  WATERMARK wk FOR ts as withOffset (ts, 2000) -- 为rowtime定义watermark
这样定义有bug，窗口不会到时关闭，需要通过阿里自己的关键词）
) with (
  type = 'datahub',
  endPoint = 'http://datahub的endPoint ',
  accessId = '**************************',
  accessKey = '&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&',
  project = 'scrm',
  topic = 'follow_p',
  maxRetryTimes = '20',
  retryIntervalMs = '1000',
  batchReadSize = '500',
  lengthCheck = 'NONE',
  columnErrorDebug = 'true'
);



create table print_table0(
    event_name VARCHAR        --事件名称
    ,json_message     VARCHAR      --事件内容
    ,unique_id  VARCHAR  --unique_id
    ,etl_time TIMESTAMP --数据时间
) with (
    type='print'
    -- type='datahub',
    -- endPoint = 'https://datahub的endPoint ',
    -- project = 'it',
    -- topic='event_info',
    -- accessId = '*****************',
    -- accessKey = '&&&&&&&&&&&&&&&&&&&&&&&&',
    -- batchSize='300',
    -- batchWriteTimeoutMs='1000'
);
create view view0 as (
  select
    follow.id,
    follow.opportunity_id,
    follow.shop_id,
    follow.follow_time,
    follow.follow_mode,
    follow.follow_result_code,
    follow.sa_name,
    follow.event_type,
    follow.new_dts_sync_dts_after_flag,
    follow.new_dts_sync_dts_operation_flag
  FROM (
      select
        *
      FROM
        follow_p
      where
        new_dts_sync_dts_after_flag = 'Y'
        AND new_dts_sync_dts_operation_flag <> 'D'
        AND follow_mode = '10420004'
        AND follow_result_code is not null
        AND event_type is not null
        AND DATE_FORMAT (
          TO_TIMESTAMP (FROM_UNIXTIME (cast (follow_time as bigint) / 1000)),
          'yyyy-MM-dd HH:mm:ss'
        ) > Date_format (
          substr (cast (current_timestamp as varchar), 1, 10),
          'yyyy-MM-dd',
          'yyyy-MM-dd HH:mm:ss'
        )
    ) as follow
  GROUP
    BY SESSION (ts, INTERVAL '5' minute),
    follow.id,
    follow.opportunity_id,
    follow.shop_id,
    follow.follow_time,
    follow.follow_mode,
    follow.follow_result_code,
    follow.sa_name,
    follow.event_type,
    follow.new_dts_sync_dts_after_flag,
    follow.new_dts_sync_dts_operation_flag
);

create view view1 as (
  select   
   follow.id 
  ,follow.opportunity_id 
  ,follow.shop_id  
  ,follow.follow_time  
  ,follow.follow_mode  
  ,follow.follow_result_code  
  ,follow.sa_name  
  ,follow.event_type  
  ,follow.new_dts_sync_dts_after_flag   
  ,follow.new_dts_sync_dts_operation_flag
  ,opportunity.id    opportunityid      --潜客表ID
  ,opportunity.shop_id    --店铺ID
  ,opportunity.opportunity_type    --线索类型（线上/线下）
  ,opportunity.brand_name  
  ,opportunity.model_name  
  ,opportunity.model_code  
  ,opportunity.customer_name  
  ,opportunity.customer_mobile_phone  
  FROM
view0  as follow 
LEFT JOIN
tb_scrm_opportunity_d FOR SYSTEM_TIME AS OF PROCTIME()   AS  opportunity
ON follow.opportunity_id=opportunity.id
);


insert into print_table0
select 
'exhibithall_reception'  event_name
,concat('{',CONCAT_WS(',',brand,car_model,`name` ,phone,id,dealer_code,dealer_shop_name,follow_time,action_time,follow_mode,follow_result,sa_name,event_type
),'}') json_message
,concat('event_exhibithall_reception','-',followid) unique_id
,cast(cast(CURRENT_TIMESTAMP as bigint)*1000 as timestamp)  etl_time
FROM
(select 
CONCAT_WS(':','\"brand\"',concat('\"',case when follow.brand_name='A' then 'A'
                                        WHEN follow.brand_name='B'  then 'B'
                                        WHEN follow.brand_name='C' or  follow.brand_name='c'  then 'C'
                                        WHEN follow.brand_name='D' then 'D'
                                        when follow.brand_name='E' then 'E'
                                       ELSE '' END ,'\"')) brand  
,CONCAT_WS(':','\"car_model\"',concat('\"',coalesce(model.model_name,''),'\"'))  car_model
,CONCAT_WS(':','\"name\"',concat('\"',coalesce(follow.customer_name,''),'\"')) `name` 
,CONCAT_WS(':','\"phone\"',concat('\"',coalesce(follow.customer_mobile_phone,''),'\"')) phone 
,CONCAT_WS(':','\"id\"',concat('\"',coalesce(follow.id,''),'\"')) id 
,CONCAT_WS(':','\"dealer_code\"',concat('\"',coalesce(dealer_shop.latest_shop_no,''),'\"')) dealer_code 
,CONCAT_WS(':','\"dealer_shop_name\"',concat('\"',coalesce(dealer_shop.name,''),'\"'))  dealer_shop_name --专营店名称
,CONCAT_WS(':','\"follow_time\"',concat('\"',coalesce(DATE_FORMAT(TO_TIMESTAMP(FROM_UNIXTIME(cast(follow.follow_time as bigint)/1000)) ,'yyyy-MM-dd HH:mm:ss'),''),'\"')) follow_time
,CONCAT_WS(':','\"action_time\"',concat('\"',coalesce(DATE_FORMAT(TO_TIMESTAMP(FROM_UNIXTIME(cast(follow.follow_time as bigint)/1000)) ,'yyyy-MM-dd HH:mm:ss'),''),'\"')) action_time
,CONCAT_WS(':','\"follow_mode\"',concat('\"',coalesce(dict0.dict_name,''),'\"')) as follow_mode --跟进方式
,CONCAT_WS(':','\"follow_result\"',concat('\"',coalesce(dict1.dict_name,''),'\"')) as follow_result --跟进结果
,CONCAT_WS(':','\"sa_name\"',concat('\"',coalesce(follow.sa_name,''),'\"')) sa_name
,CONCAT_WS(':','\"event_type\"',concat('\"',coalesce(dict2.dict_name,''),'\"')) as event_type --跟进事件类型
,follow.id as followid
FROM
view1  as follow 
LEFT JOIN
brand_model_a_d FOR SYSTEM_TIME AS OF PROCTIME()   AS  model
ON follow.model_code=model.model_code
left join 
dealer_shop_a_d FOR SYSTEM_TIME AS OF PROCTIME()   AS dealer_shop 
ON follow.shop_id = dealer_shop.dealer_shop_no
left JOIN 
follow_sys_dict_a_d  FOR SYSTEM_TIME AS OF PROCTIME()   AS   dict0
ON follow.follow_mode=dict0.dict_code
left JOIN 
follow_sys_dict_a_d  FOR SYSTEM_TIME AS OF PROCTIME()   AS   dict1
ON follow.follow_result_code=dict1.dict_code
left JOIN 
follow_sys_dict_a_d  FOR SYSTEM_TIME AS OF PROCTIME()   AS   dict2
ON follow.event_type=dict2.dict_code
WHERE DATE_FORMAT(TO_TIMESTAMP(FROM_UNIXTIME(cast(follow_time as bigint)/1000)),'yyyy-MM-dd HH:mm:ss')
>Date_format(substr(cast(current_timestamp as varchar),1,10),'yyyy-MM-dd','yyyy-MM-dd HH:mm:ss')
) T 
;