All queries to reproduce the technique from my blog post https://torsten.io/stdout/expanding-json-arrays-to-rows

expanding-json-arrays-to-rows.sql

-- Related blog post to this Gist:
-- https://torsten.io/stdout/expanding-json-arrays-to-rows

-- Run these commands on a interactive RedShift session:

CREATE TEMP TABLE clusters AS (
    SELECT 1 AS id, '[1, 2]' AS node_sizes UNION ALL
    SELECT 2 AS id, '[5, 1, 3]' AS node_sizes UNION ALL
    SELECT 3 AS id, '[2]' AS node_sizes
);


-- This is just a temp table to leave no trace after the session
-- In practice I am using a `CREATE VIEW` and more rows.

CREATE TEMP TABLE seq_0_to_100 AS (
    SELECT 0 AS i UNION ALL
    SELECT 1 UNION ALL
    SELECT 2 UNION ALL
    SELECT 3 UNION ALL
    SELECT 4 UNION ALL
    SELECT 5
    -- I am stopping here, you could easily generate this as a VIEW with 100+ real rows...
);


-- To see the intermediate relation:

SELECT id, JSON_EXTRACT_ARRAY_ELEMENT_TEXT(node_sizes, seq.i) AS size
FROM clusters, seq_0_to_100 AS seq
WHERE seq.i < JSON_ARRAY_LENGTH(node_sizes)
ORDER BY 1, 2;


-- To fetch the maximum size:

WITH exploded_array AS (
    SELECT id, JSON_EXTRACT_ARRAY_ELEMENT_TEXT(node_sizes, seq.i) AS size
    FROM clusters, seq_0_to_100 AS seq
    WHERE seq.i < JSON_ARRAY_LENGTH(node_sizes)
  )
SELECT max(size)
FROM exploded_array;

Thanks for this amazing hack 👍

Btw, you can create the sequence table with less manual work:

SELECT ROW_NUMBER() OVER (ORDER BY TRUE)::INTEGER- 1 as seq into seq_0_to_5000
FROM <table.with.at-least.5000.rows> LIMIT 5000;