【转载】hive 中 统计某字段json数组中每个value出现的次数

有这样一个json串

{"viewdata":[{"city_id":"59","position_id":0,"qd_title":"网红打卡地","list_id":35},{"city_id":"59","position_id":1,"qd_title":"看青山游绿水","list_id":37}]}

需要将json数组里的qd_title都提取出来转换成hive中的array数组。下面介绍两种方法

法一get_json_object+正则

1.首先可以使用get_json_object函数，提取出数组，但是这个返回的是一个字符串

select get_json_object('{"viewdata":[{"city_id":"39","position_id":0,"qd_title":"网红打卡地","list_id":135},
{"city_id":"39","position_id":1,"qd_title":"看青山游绿水","list_id":327}]}',
'$.viewdata[*].qd_title')

-- 返回,注意这不是一个array数组，只是一个字符串
["网红打卡地","看青山游绿水"]

2.将字符串中的[ ] "都去掉，形成一个,分割的字符串

regexp_replace('${刚刚得到的字符串}','(\\[|\\]|")','')

3.使用字符串分割函数split

select 
split(
   regexp_replace(
      get_json_object('{"viewdata":[{"city_id":"39","position_id":0,"qd_title":"网红打      卡地","list_id":135},
      {"city_id":"39","position_id":1,"qd_title":"看青山游绿水","list_id":327}]}',
      '$.viewdata[*].qd_title'),
   '(\\[|\\]|")',''),
",")

4.整体使用LATERAL VIEW 打平数组进行统计

SELECT qdtitle,COUNT(DISTINCT uuid) uv
  FROM ba_travel.bas_log_sdk_mt_mv a LATERAL VIEW explode(split(regexp_replace(get_json_object(a.event_attribute['custom'],'$.viewdata[*].qd_title'),'(\\[|\\]|")',''),",")) b AS qdtitle
 GROUP BY qdtitle

法二 正则匹配
1.观察json数组中每一个元素都是由{}保卫，由,分割，所以可以使用``}，```对字符串进行拆分

-- event_attribute['custom'] 对应的就是上面的json字符串
split(event_attribute['custom'],'"}')

2.对分割出来的每一个元素进行正则匹配，提取出qd_title对应的value

-- qd_titles 为上面分割出数组的一个元素
regexp_extract(qd_titles,'qd_title...([^"]+)',1)

3.整体使用later view 将数组打平

SELECT regexp_extract(qd_titles,'qd_title...([^"]+)',1) title,
       COUNT(DISTINCT uuid) uv
  FROM ba_travel.bas_log_sdk_mt_mv a LATERAL VIEW  explode(split(event_attribute['custom'],'"}')) b as qd_titles 
 GROUP BY regexp_extract(qd_titles,'qd_title...([^"]+)',1)

原文链接：https://cloud.tencent.com/developer/article/1336159