Compound trigger: collecting mutating rows into nested table - sql

Question

I have two tables in my project: accounts and transactions (one-to-many relationship). In every transaction I store the balance of the associated account (after the transaction is executed). Additionally in every transaction I store a value of the transaction.
So I needed a trigger fired when someone adds new transaction. It should check whether new account balance will be correct (old account balance + transaction value = new account balance stored in transaction).
So I was suggested, I should use a compound trigger which would:
in before each row section: save a row's PK (made of two columns) somewhere,
in after statement section: check whether all inserted transactions where correct.
Now I can't find anywhere how could I implement the first point.
What I already have:
CREATE OR REPLACE TRIGGER check_account_balance_is_valid
FOR INSERT
ON Transactions
COMPOUND TRIGGER
TYPE Modified_transactions_T IS TABLE OF Transactions%ROWTYPE;
Modified_transactions Modified_transactions_T;
BEFORE STATEMENT IS BEGIN
Modified_transactions := Modified_transactions_T();
END BEFORE STATEMENT;
BEFORE EACH ROW IS BEGIN
Modified_transactions.extend;
Modified_transactions(Modified_transactions.last) := :NEW;
END BEFORE EACH ROW;
AFTER STATEMENT IS BEGIN
NULL; -- I will write something here later
END AFTER STATEMENT;
END check_account_balance_is_valid;
/
However, I got that:
Warning: execution completed with warning
11/58 PLS-00049: bad bind variable 'NEW'
Could someone tell me, how to fix it? Or maybe my whole "compound trigger" idea is wrong and you have better suggestions.
Update 1
Here is my ddl script: http://pastebin.com/MW0Eqf9J

Answer 1

Maybe try this one:
TYPE Modified_transactions_T IS TABLE OF ROWID;
Modified_transactions Modified_transactions_T;
BEFORE STATEMENT IS BEGIN
Modified_transactions := Modified_transactions_T();
END BEFORE STATEMENT;
BEFORE EACH ROW IS BEGIN
Modified_transactions.extend;
Modified_transactions(Modified_transactions.last) := :NEW.ROWID;
END BEFORE EACH ROW;
or this
TYPE PrimaryKeyRecType IS RECORD (
Col1 Transactions.PK_COL_1%TYPE, Col2 Transactions.PK_COL_2%TYPE);
TYPE Modified_transactions_T IS TABLE OF PrimaryKeyRecType;
...
Modified_transactions(Modified_transactions.last) := PrimaryKeyRecType(:NEW.PK_COL_1, :NEW.PK_COL_2);

Answer 2

Your immediate problem is that :new is not a real record so it is not of type Transactions%ROWTYPE. If you're really going to go down this path, you would generally want to declare a collection of the primary key of the table
TYPE Modified_transactions_T IS TABLE OF Transactions.Primary_Key%TYPE;
and then put just the primary key in the collection
BEFORE EACH ROW IS BEGIN
Modified_transactions.extend;
Modified_transactions(Modified_transactions.last) := :NEW.Primary_Key;
END BEFORE EACH ROW;
The fact that you are trying to work around a mutating table exception in the first place, however, almost always indicates that you have an underlying data modeling problem that you should really be solving. If you need to query other rows in the table in order to figure out what you want to do with the new rows, that's a pretty good indication that you have improperly normalized your data model and that one row has some dependency on another row in the same table rather than being an autonomous fact. Fixing the data model is almost always preferrable to working around the mutating table exception.