Database rows are typically a fixed length. Why use 8 bytes per betid if you think they will always fit into 4 bytes. When you're storing 2 billion of them it makes a difference.
You sure?
Using postgres:
Quote
CREATE TABLE miniBets(
id int, -- this is a 32 bit number
user_id int,
payout float,
profit float
);
INSERT INTO minibets VALUES(1, 1, 23.0, 23.0);
CREATE TABLE bigBets(
id bigint, -- this is a 64 bit number
user_id bigint,
payout float,
profit float
);
INSERT INTO bigBets VALUES(1, 1, 23.0, 23.0);
SELECT pg_relation_size('miniBets'), pg_relation_size('bigBets')
id int, -- this is a 32 bit number
user_id int,
payout float,
profit float
);
INSERT INTO minibets VALUES(1, 1, 23.0, 23.0);
CREATE TABLE bigBets(
id bigint, -- this is a 64 bit number
user_id bigint,
payout float,
profit float
);
INSERT INTO bigBets VALUES(1, 1, 23.0, 23.0);
SELECT pg_relation_size('miniBets'), pg_relation_size('bigBets')
returns:
[8192, 8192]
AKA an int32 and int64 end up using the space, due to alignment. I would assume the same is true with other databases, assuming you're using a 64 bit machine .. which I can't imagine you're not. 32 bits is for barbarians
No, but seriously I almost never use 32 bits numbers.