Int or float

Question

Hello,
In my J2ee server, I have a lots of computation of money. Usually, the primitive type float is satisfied for presenting money. But I have a question, is it better to use int to store it. Even in database, I can save more space by using INT than using DECIMAL(10,2). My MySql, INT requires 4 bytes, DECIMAL(10,2) requires 12 bytes. At I/O moudle, I just simply multiply 100 for input and divid 100 for output. Because I suspect that using float may change a tiny bit at the end of the value. Any suggestions to me?
Thanks!

Answer 1

How are we saying opposite things? I said that he should use an int unless it doesn't matter if the results of a calculation are actually correct. Floats/doubles are generally a bad idea to use unless you know what you are doing, since the approximating effect they have on calculations can be strange to a newcomer. For printing the results of money calculations, floats might work fine. However, you will get a certain amount of error, so you just have to watch out. If you are not storing the results of calculations where the operands are floats, then the error might be ok.
Rounding in accounting programs, like interest programs always have very specific business rules that specify to what degree of precision the calculation is to be performed and which style of rounding should be used. I most programs that need to deal with money, the smallest increment will usually be a cent (for US money anyway). And if they don't have specific rules, then maybe you should get some (or stash the error into an account in your name, Superman 2 thank you very much).
BigDecimal is fine to use if you don't need to worry about speed, and if you have the possibility of overflow with longs, BigDecimal is probably a good idea since it can represent huge numbers that long cannot. I would imagine that most programs don't need more precision then a long can provide.