MySQL two insert statements in one transaction with a dependency - mysql

Question

Say there are three tables for your bakery business:
online_order
invoice
customer_id
Your boss calls you and says "I need to order a chocolate cake for all of our clients, invoice them all separately as well". So now you have to do two things:
Create an online_order entry for each client
Then invoice them for that order.
However, each client needs to have their OWN online_order number that they can reference on their invoice.
How can you do this in one transaction? Below is what I have (psuedocode) but I'll show you my output and expected output.
BEGIN;
-- First lets create the order
INSERT INTO online_order (online_order_id, date, cake_type, cake_flavor)
VALUES
(AUTO_INCREMENT, NOW(), round, chocolate);
-- Then take that online_order_id and create an invoice out of it for one of our customers
INSERT INTO invoice (date, online_order_id, invoice_amt, customer_id)
SELECT
NOW(),
LAST_INSERT_ID(),
10.00,
CUSTOMER_ID
FROM CUSTOMER;
ROLLBACK;
END;
My output
-- Let's say the next auto-incremented online_order_id is 10011:
INSERT INTO invoice VALUES ("2016-01-01", 10011, 10.00, "1");
INSERT INTO invoice VALUES ("2016-01-01", 10011, 10.00, "2");
INSERT INTO invoice VALUES ("2016-01-01", 10011, 10.00, "3");
INSERT INTO invoice VALUES ("2016-01-01", 10011, 10.00, "4");
INSERT INTO invoice VALUES ("2016-01-01", 10011, 10.00, "5");
Desired output (different online_order_id for each invoice)
INSERT INTO invoice VALUES ("2016-01-01", 10011, 10.00, "1");
INSERT INTO invoice VALUES ("2016-01-01", 10012, 10.00, "2");
INSERT INTO invoice VALUES ("2016-01-01", 10013, 10.00, "3");
INSERT INTO invoice VALUES ("2016-01-01", 10014, 10.00, "4");
INSERT INTO invoice VALUES ("2016-01-01", 10015, 10.00, "5");
I need to insert an order, then insert an invoice, then insert an order, then insert an invoice.... for each client. How can I write this in MySQL? Thank you!

Answer 1

You can create a table to maintain the IDs for different entity and use SELECT.. FOR UPDATE to get the atomic values. Table would look like this:
Entity | Value
Order | 1
Blah | 2
You can then use the following block to get the new id:
set #id := select value from Ids where Entity = 'Order' FOR UPDATE;
update Ids set value = value + 1 Enity = 'Order';
commit;
This would give you the id and increment it by 1. This id then can be used in invoide table.
As you will be using select.. for update, corresponding row will be locked and no other transaction will be able to read/modify the value, resulting in atomicity.
This approach might prove to be a bottleneck if you have too many insertions executing in parallel.

Answer 2

Your pseudocode doesn't make much sense - you should only rollback the transaction if it fails. But leaving that aside for now, inserting one row at a time is very innefficient. I would go with something like....
INSERT INTO invoice
(date, online_order_id, invoice_amt, customer_id, reference)
SELECT
NOW(), LAST_INSERT_ID(), 10.00, CUSTOMER_ID, CONNECTION_ID()
FROM CUSTOMER;
INSERT INTO online_order
(date, cake_type, cake_flavor)
SELECT NOW(), 'round', 'chocolate'
FROM invoice
WHERE reference=CONNECTION_ID();
UPDATE invoice SET reference=NULL
WHERE reference=CONNECTION_ID();
The above ensures most of the consistency without the use of a transaction at all, although with a procedure you could create a variable and assign its value from uuid() those obviating the need for the final update and keeping an index on reference well balanced.