HIVE SQL 根据主键去重并实现其余字段分组聚合

永远在减肥永远110的的小潘

已于 2023-11-23 10:49:57 修改

阅读量2.2k

点赞数

文章标签： hive sql hadoop

于 2023-07-18 11:36:06 首次发布

本文链接：https://blog.csdn.net/p1306252/article/details/131782412

版权

相同个人id下所有字段按时间顺序补位，取首个不为空值
在这里插入图片描述

--数据建表
drop table if exists db.tb_name;
create table if not exists db.tb_name
( id string
	,name string
	,tele string
	,email string
	,`date` string
)
;
insert overwrite table db.tb_name
values 
("32001","张三","23456789",null,"2023-07-18")
,("32001",null,null,"23456789@163.com","2023-07-19")
,("32002","李四",null,"23456876@qq.com","2023-07-18")
,("32003","王二",null,null,"2023-07-18")
,("32003",null,"9876789",null,"2023-06-18")
,("32003",null,null,"9876789@gmail.com","2023-07-18")
,("32004","刘五","987456798",null,"2023-07-18")
,("32004","刘七","1987456798",null,"2023-07-20")
;

--distribute by 分区排序：类似MR中partition，进行分区，结合sort by使用

drop table if exists db.tb_name_new;
create table if not exists db.tb_name_new as
select id 
	,collect_list(`name`)[0] as `name`
	,collect_list(`tele`)[0] as `tele`
	,collect_list(`email`)[0] as `email`
	,collect_list(`date`)[0] as `date`
from 
(
select id
	,name
	,tele
	,email
	,`date`
from 
db.tb_name
distribute by id 
sort by id,`date` desc
) t 
group by id
;

*注意：此处是取的首个不为空（即不为null）的字段，所以在实际使用过程中应提前将空字符串转为null值。

--剔除字符串中的不可见字符，若该字段中均为不可见字符或该字段为空字符串，则转为空
case when length(regexp_replace(col_name,'[\\x00-\\x08\\x0B-\\x0C\\x0E-\\x1F]+|\\s+',''))>0 
	then regexp_replace(col_name,'[\\x00-\\x08\\x0B-\\x0C\\x0E-\\x1F]+|\\s+','')
	else null end as new_col_name