java - Hive：将两张地图合并为一列

Question

我有一个蜂巢表

create table mySource(
    col_1   map<string, string>,
    col_2   map<string, string>
)

这是记录的样子

col_1                col_2
{"a":1, "b":"2"}     {"c":3, "d":"4"}

我的目标表看起来像这样

create table myTarget(
        my_col   map<string, string>
    )

现在我想将 mySource 中的两列组合成一个映射并将其提供给我的目标表。基本上我想写类似的东西

insert into myTarget
    select
        some_method(col_1, col_2) as my_col
    from mySource;

蜂巢中是否有内置方法可以做到这一点？我用 collect_set 尝试了一些东西，但出现了很多错误

Answer 1

该解决方案仅使用内置方法。分解两个地图，UNION ALL 结果，收集数组key:value，连接数组','，使用 将字符串转换为地图str_to_map：

with mytable as (--Use your table instead of this
select 
map('a','1', 'b','2') as col_1, map('c','3', 'd','4') as col_2
)

select str_to_map(concat_ws(',',collect_set(concat(key,':',val)))) as mymap
from
(
select m1.key, m1.val 
  from mytable
       lateral view explode(col_1) m1 as key, val
union all
select m2.key, m2.val 
  from mytable
       lateral view explode(col_2) m2 as key, val
)s       
;

结果：

mymap

{"a":"1","b":"2","c":"3","d":"4"}  

使用砖房图书馆会容易得多：

ADD JAR /path/to/jar/brickhouse-0.7.1.jar;
CREATE TEMPORARY FUNCTION COMBINE AS 'brickhouse.udf.collect.CombineUDF';

select combine(col_1, col_2) as mymap from mytable;