how to do set union operations on arrays? - distinct

Question

Here's what I want to accomplish -
I want to have an array[string] column in a singlestore columnar table
Do a set union operation on this column over some time range/conditions.
I don't know how to express this in Singlestore sql.
Singlestore doesn't support array columns, so I've resorted to using json col.
Any pointers ?
Sample table
| key | locations |
| ----- | --------------------------------- |
| alice | ["sanjose", "sancarlos", "miami"] |
| alice | ["sanjose","milpitas"] |
| alice | ["miami"] |
| alice | ["sanmateo","sanfrancisco"] |
| alice | ["redwoodshores","sanfrancisco"] |
| bob | ["sanjose","milpitas"] |
| bob | ["freemont","onioncity"] |
| bob | ["onioncity"] |
| bob | ["newark","milpitas"] |
| bob | ["sanjose"] |
| bob | ["santacruz"] |
Expected output
| key | array_agg(distinct elem) OR array_distinct(flatten(array_agg(locations))) |
| ----- | ------------------------------------------------------------------------------ |
| alice | [ miami, milpitas, redwoodshores, sancarlos, sanfrancisco, sanjose, sanmateo ] |
| bob | [ freemont, milpitas, newark, onioncity, sanjose, santacruz ] |
Any pointers on how I can accomplish this with Singlestore ?
How I would do this in other DBs/Frameworks:
Postgres - It was relatively simple to construct this query. I can create an array column and do array_agg(distinct elem) - http://www.sqlfiddle.com/#!17/58628/1
Spark - as simple as array_distinct( array_agg( col ) )

Answer 1

You can use JSON_AGG function to achieve this.
This function does not support distinct like json_agg(distinct(key_col)), so using the distinct clause in CTE itself.
Docs: https://docs.singlestore.com/managed-service/en/reference/sql-reference/json-functions/json_agg.html
Sample Code
create table json_array_example (key_col text, locations json);
insert into json_array_example values('alice', '["sanjose", "sancarlos", "miami"]');
insert into json_array_example values('alice', '["sanjose","milpitas"]');
insert into json_array_example values('alice', '["miami"]');
insert into json_array_example values('alice', '["sanmateo","sanfrancisco"]');
insert into json_array_example values('alice', '["redwoodshores","sanfrancisco"]');
insert into json_array_example values('bob' , '["sanjose","milpitas"]');
insert into json_array_example values('bob' , '["freemont","onioncity"]');
insert into json_array_example values('bob' , '["onioncity"]');
insert into json_array_example values('bob' , '["newark","milpitas"]');
insert into json_array_example values('bob' , '["sanjose"]');
insert into json_array_example values('bob' , '["santacruz"]');
WITH t AS (SELECT distinct key_col, table_col AS locations
FROM json_array_example
JOIN TABLE(JSON_TO_ARRAY(locations)))
select key_col, json_agg(locations) as locations
from t
group by key_col;