Twitch Gamers data in DuckDB with SQL/PGQ extensions

duckpgq-twitch-gamers.sql

-- From https://motherduck.com/blog/duckdb-puppygraph-graph-model-on-motherduck/

install zipfs from community;
load zipfs;
load httpfs; -- TODO: autoload

CREATE TABLE account AS SELECT * FROM read_csv(
  'zip://https://snap.stanford.edu/data/twitch_gamers.zip/large_twitch_features.csv',
  types={mature: boolean, dead_account: boolean, affiliate: boolean}
);

CREATE TABLE follows AS SELECT * FROM read_csv(
  'zip://https://snap.stanford.edu/data/twitch_gamers.zip/large_twitch_edges.csv',
  names=[follower, followee]
);

-- SQL queries

select count(*) from follows;
select count(*) from account;
select count(*) from account where dead_account;
select * from account order by updated_at limit 5;

-- Property Graph

install duckpgq from community;
load duckpgq;

CREATE PROPERTY GRAPH gamers
  VERTEX TABLES (
    account
  )
  EDGE TABLES (
    follows SOURCE KEY (follower) REFERENCES account (numeric_id)
            DESTINATION KEY (followee) REFERENCES account (numeric_id)
  );

-- Property Graph Queries

FROM GRAPH_TABLE (gamers MATCH (a:account)) ORDER BY updated_at LIMIT 5;

FROM GRAPH_TABLE (gamers
  MATCH (a:account)-[f:follows]->(b:account)
  COLUMNS (a.numeric_id, b.numeric_id)
) LIMIT 1;

-- List the top-5 viewed accounts among the 2-hop followers of
-- the least recently updated account.
FROM GRAPH_TABLE (gamers
  MATCH (a:account)-[f:follows]-(b:account)-[g:follows]-(c:account)
  WHERE c.numeric_id = (
    SELECT first(numeric_id ORDER BY updated_at) FROM account
  )
  COLUMNS (a.numeric_id, a.views)
) ORDER BY views DESC LIMIT 5;