演示

create table src (myjson string);

insert into src values
    ('{"Rtype":{"ver":"1","os":"ms","type":"ns","vehicle":"Mh-3412","MOD":{"Version":[{"ABC":{"XYZ":"123.dfer","founder":"3.0","GHT":"Florida","fashion":"fg45","cdc":"new","dof":"yes","ts":"2000-04-01T00:00:00.171Z"}}]}}}')
   ,('{"Rtype":{"ver":"1","os":"ms","type":"ns","vehicle":"Mh-3412","MOD":{"Version":[{"GAP":{"XVY":"123.dfer","FAH":"3.0","GHT":"Florida","fashion":"fg45","cdc":"new","dof":"yes","ts":"2000-04-01T00:00:00.171Z"}}]}}}')
   ,('{"Rtype":{"ver":"1","os":"ms","type":"ns","vehicle":"Mh-3412","MOD":{"Version":[{"BOX":{"VOG":"123.dfer","FAH":"3.0","FAX":"Florida","fashion":"fg45","cdc":"new","dof":"yes","ts":"2000-04-01T00:00:00.171Z"}}]}}}')
;

create table trg_abc (myjson string);
create table trg_gap (myjson string);
create table trg_box (myjson string);

from src
insert into trg_abc select myjson where get_json_object(myjson,'$.Rtype.MOD.Version[0].ABC') is not null
insert into trg_gap select myjson where get_json_object(myjson,'$.Rtype.MOD.Version[0].GAP') is not null
insert into trg_box select myjson where get_json_object(myjson,'$.Rtype.MOD.Version[0].BOX') is not null
;

首先，您需要将数据作为json存储在配置单元表中：
我假设您的配置单元表是外部的（通常是-请检查它） SHOW CREATE TABLE your_table ).
例如，如果是这样，整个数据集就位于某个hdfs/s3路径中 s3a://your_bucket/your_jsons_location/ 下载json-udf-1.3.7-jar-with-dependencies.jar并运行 ADD JARS s3a://your_bucket/lib/json-udf-1.3.7-jar-with-dependencies.jar; 然后必须为每个json模式创建一个专用的json表：

CREATE EXTERNAL TABLE boxes
(Rtype struct<ver:string,os:string,type:string,vehicle:string,MOD:struct<Version:array<struct<BOX:struct<VOG:string,FAH:string,FAX:string,fashion:string,cdc:string,dof:string,ts:string>>>>>)
ROW FORMAT SERDE 'org.openx.data.jsonserde.JsonSerDe' location 's3a://your_bucket/your_jsons_location/';

CREATE EXTERNAL TABLE gaps
(Rtype struct<ver:string,os:string,type:string,vehicle:string,MOD:struct<Version:array<struct<GAP:struct<XVY:string,FAH:string,GHT:string,fashion:string,cdc:string,dof:string,ts:string>>>>>)
ROW FORMAT SERDE 'org.openx.data.jsonserde.JsonSerDe' location 's3a://your_bucket/your_jsons_location/';

CREATE EXTERNAL TABLE abcs
(Rtype struct<ver:string,os:string,type:string,vehicle:string,MOD:struct<Version:array<struct<ABC:struct<XYZ:string,founder:string,GHT:string,fashion:string,cdc:string,dof:string,ts:string>>>>>)
ROW FORMAT SERDE 'org.openx.data.jsonserde.JsonSerDe' location 's3a://your_jsons_location/';

现在如果你要跑：

SELECT * FROM boxes;
SELECT * FROM gaps;
SELECT * FROM abcs;

您将看到每个表只正确地解析了匹配的json（根据create语句中指定的模式）。每个表中不匹配的为空。
要筛选出不相关的记录，请运行： SELECT * FROM s WHERE Rtype.mod.version[0]. IS NOT NULL; 注意：这个完整的解释假设json存储在hive表的外部（我特别使用了s3，但也可以是hdfs）