3 ms·
Yes. MapReduce's model is basically: 1. Map (key, value) -> (new_key, tmp_value) 2. Group by new_key 3. Reduce (new_key, all tmp_values for that key) ->
by Sesse__ 2y ago
Yes. MapReduce's model is basically:
1. Map (key, value) -> (new_key, tmp_value)
2. Group by new_key
3. Reduce (new_key, all tmp_values for that key) -> (new_key, new_values)
In that respect, it's not that far from SQL with custom aggregates. I guess the most precise SQL representation would be
SELECT REDUCE(MAP(t)) FROM foo GROUP BY KEY(MAP(t))
(I've both been on the MapReduce team, and worked on an SQL database. I don't honestly think they're that comparable.)