How to get array/bag of elements from Hive group by operator?

Question

I want to group by a given field and get the output with grouped fields. Below is an example of what I am trying to achieve:-

Imagine a table named 'sample_table' with two columns as below:-

F1  F2
001 111
001 222
001 123
002 222
002 333
003 555

I want to write Hive Query that will give the below output:-

001 [111, 222, 123]
002 [222, 333]
003 [555]

In Pig, this can be very easily achieved by something like this:-

grouped_relation = GROUP sample_table BY F1;

Can somebody please suggest if there is a simple way to do so in Hive? What I can think of is to write a User Defined Function (UDF) for this but this may be a very time consuming option.

Cassius

Answer 1

The built in aggregate function collect_set (doumented here) gets you almost what you want. It would actually work on your example input:

SELECT F1, collect_set(F2)
FROM sample_table
GROUP BY F1

Unfortunately, it also removes duplicate elements and I imagine this isn't your desired behavior. I find it odd that collect_set exists, but no version to keep duplicates. Someone else apparently thought the same thing. It looks like the top and second answer there will give you the UDAF you need.

Scibert

In Hive 0.13 there's a collect_list function which would return duplicates. – Featly

Answer 2

collect_set actually works as expected since a set as per definition is a collection of well defined and distinct objects i.e. objects occur exactly once or not at all within a set.

Reverential