Hi, I hope someone can offer some advice for the following query example. Or … can you point me to a good source for simple model. I am now hearing things like ‘double entry ledgers’ and wondering if there was a much simpler way to do all this.
So, the balance for a user (after each transaction) is contained in the user_balance column. And the current balance is contained in the last txn for that user.
(I will agree with rudy that a datetime column would be helpful, considering the future, but for the purposes of this specific query, it’s moot, as the order of events doesnt matter to a flat total. If you need to make sure the person never went into negative balance, for example, datetimes are a must.)
That said. This is a… confusing table structure.
What is the purpose of ‘type’? Surely it’s redundant; if FROM = you, money went out from you, if TO = you, money went in to you.
What does the type field hold if user 11 gives money to user 22?
Two ways to change it:
Type = In/Out, ID field.
(ID = party involved, Type is whether to add or subtract from balance).
Cons to this table: Just because User 11 sent out 10 dollars, and User22 received 10 dollars, does not mean you can say user 11 paid user 22.
No type field, 2 ID fields (From/To)
(ID = 0 = ‘unspecified user’)
Cons to this table: Possibly harder to construct a balance query due to having to do conditional statements.
My bad, actually I do have a datetime in there, should have put it in the example. If you could just imagine datetime as a column called ‘date’.
The problem is the tos and froms are not all the same type of object. They reference entries in different tables, so there will be duplicate ids and thus I can’t simply SELECT … WHERE id=to_id
I reckon I over-simplified my example, maybe this will help
id date type from to amount giver_balance receiver_balance
1 001 deposit NULL 11 10.00 10.00 NULL
2 002 txn1 11 333 1.00 09.00 1.00
3 003 txn1 11 344 2.00 07.00 2.00
4 004 deposit NULL 11 10.00 17.00 NULL
4 004 payout NULL 334 1.00 NULL 1.00
So the transactions are:
Deposit - Givers load money
txn - Givers give money to receivers
payout - Receivers take money
At the time (a long time ago) I needed a quick solution and it seemed perfectly logical. Now I’m trying to pull answers out such as daily totals etc, I’m not so sure.
Yes, quicker, but quicker is not always better. It’s not proper modeling of the data being stored. A balance is not an element of data. It’s calculated, and therefore should never be persisted in your DB.
It also presents a dimension of potential conflict. If, for example, a ledger item was missed… from last month… inputting that record today requires that you recalculate all ledger entries’ “balance” records that were previously stored, and run a subsequent process to verify that all balances have been properly adjusted. If you’ve ever used a financial application, you know this is not an uncommon occurrence.
It’s just a mess waiting to happen, with a lot of potential pitfalls.
If speed is a requirement in your financial application, e.g. for reporting purposes, you should have a separate reporting database that “flattens” your data periodically.
If you have less than 10 million records in your table, discussions of performance are really moot, though…