CODEWITHSUNDEEP
sqlarrayssplithivehiveql
Manu Manjunath
Manu Manjunath

Reputation: 364

Explode on multiple columns in Hive

I'm trying to explode records in multiple columns in Hive.

For example, if my dataset looks like this - 

COL_01  COL_02     COL_03
1       A, B       X, Y, Z
2       D, E, F    V, W

I want this as the output - 

COL_01  COL_02  COL_03
1       A        X
1       B        Y
1       NULL     Z
2       D        V
2       E        W
2       F        NULL

Is there a way to do this in Hive?

I saw some posts on exploding for a single column but not for multiple columns like in this case.

Upvotes: 2

Views: 2879

Answers (2)

leftjoin
leftjoin

Reputation: 38325

Explode separately in subqueries and join them using full join.

with your_data as (
select stack(2,
1, 'A, B',     'X, Y, Z',
2, 'D, E, F',  'V, W'
) as (col_01, col_02, col_03)
) 

select nvl(s1.col_01,s2.col_01) as col_01, --do the same nvl() for all not exploded columns
       s1.col_02, s2.col_03
from
(select d.col_01, c2.pos2, c2.col_02 --explode col_02
  from your_data d
       lateral view outer posexplode(split(col_02,', ?')) c2 as pos2, col_02
)s1

full join

(select d.col_01, c3.pos3, c3.col_03 --explode col_03
  from your_data d
       lateral view outer posexplode(split(col_03,', ?')) c3 as pos3, col_03
)s2
on s1.col_01=s2.col_01 
   and s2.pos3=s1.pos2 --match position

Result:

col_01  s1.col_02   s2.col_03   
1          A           X    
1          B           Y    
1          NULL        Z    
2          D           V    
2          E           W    
2          F           NULL

Upvotes: 3

Sagar Morakhia
Sagar Morakhia

Reputation: 797

@Manu - You can do a lateral view on 2 columns but that would be a cross product. but I see what you need is a one to 1 mapping between the columns.

Any change you can create a map field with columns col02 and col03?

Upvotes: 0

Related Questions