Big datatype in java

Question

I have been working on an application. It is related to shopping business.
There is a model that has:

 private Float count;
 private Long price;

I want to know the best java datatype for count * price because of huge amount of price and count.

Another hand the overflow is not occured whenprice * count operation.

Answer 1

JSR 354: Money and Currency API

As a solution you can consider an approach with the JavaMoney.org library on GitHub. This library implements the JSR 354: Money and Currency API specification.

The goals of that API:

1. To provide an API for handling and calculating monetary amounts
2. To define classes representing currencies and monetary amounts, as well as monetary rounding
3. To deal with currency exchange rates
4. To deal with formatting and parsing of currencies and monetary amounts

If you don't want use any library of course you should use BigDecimal.

Answer 2

The maximum value of:

• Float is Float.MAX_VALUE, (2-2^23)*2^127, something like 3.40282346638528860e+38 or 340,282,346,638,528,860,000,000,000,000,000,000,000.000000.
• Long is Long.MAX_VALUE, 2^63-1, or 9,223,372,036,854,776,000.

Just what kind of shopping business are you running that does not fit in those types?

Floating-point

Actually, you do not want Float for the simple reason that it is based on floating-point technology, and you never use floating-point for money. Never use floating-point for any context where accuracy is important. Floating-point trades away accuracy for speed of execution. Usually, your customers will care about money. So you would never use the float, Float, double, or Double types in Java for such purposes. This workaround is prone to confusion and mistakes obviously, so it requires careful documentation and coding.

BigDecimal

Where accuracy matters, such as money, use BigDecimal. Not because it can handle large numbers, but because it is accurate. Slow, but accurate.

The advice to use BigDecimal applies only where you have a fractional amount such as tracking the pennies on a dollar. If you are using only integer numbers, you have no need for BigDecimal.

Integer workaround for fractional amounts

Indeed a workaround for the floating-point problem in languages lacking an alternative like BigDecimal is to multiple all fractional amounts until they become integers. For example, if doing bookkeeping to the penny on a dollar (United States), then multiple all amounts by 100 to keep a count of whole pennies as an integer rather than a count of fractional dollars.

Integers

As for working with integer numbers in Java, you have multiple simple choices.

• For numbers 127 and less, use byte or Byte, using 8 bits.
• For numbers 32,767 and less, use short or Short, using 16 bits.
• For numbers 2^31-1 and less (about 2 billion), use int or Integer, using 32 bits.
• For numbers 2^63-1 and less (about umpteen gazillion), use long or Long, using 64 bits.
• For even larger numbers, use BigInteger.

Generally best to use the smallest integer type that can comfortably fit your current values as well as fit foreseeable future values.

The 32-bit or 64-bit types are your main choices for modern hardware. You needn't worry about the smallest types unless working with massive amounts of these values, or are quite constrained on memory. And using BigInteger is overkill for most any business-oriented app. (Science and engineering apps might be a different story.)

---

See also the Answer by i.merkurev on the JSR 354: Money and Currency API library.

Answer 3

For huge values there are BigDecimal or BigInteger classes. I will use BigDecimal in your case. You never get overflow with this classes.