What is the data type for a company's profit?

Answer 1

Although it may seem obvious that currency should be represented by some floating point data type, floating point values are merely approximations, and these approximations can introduce rounding errors that can be difficult to predict. For instance, when dealing with Sterling, we physically deal with whole pounds (£) and hundredths of a pound (pennies). However, tax and interest rates are typically expressed as percentages, so we inevitably need to introduce fractions of a penny into our calculations, rounding up or down to the nearest penny. But a penny is already a fraction (one hundredth of a pound), so we immediately lose two digits of precision before we've begun.

In order to retain the precision we require, we need to introduce a scalar, such that we deal in whole pence rather than whole pounds. A scalar of 100 would allow us to treat £1.23 as if it were really 123 pence. Now we can store monetary values as integers instead of floating point values and that alone can have a significant benefit in terms of performance (integer arithmetic is significantly quicker than floating point arithmetic).

We still have the problem of how to deal with unpredictable rounding errors that might occur within the fractions of a penny we create when working with percentages. However, this is easily resolved by raising our scalar by a higher power of 10. Thus a scalar of 1000 would mean our 123 pence then becomes 1230 tenths of a pence, while a scalar of 10,000 would create 12,300 hundredths of a pence. That's about as much precision as we require for currency values. We could go further, but we have to keep in mind that the greater the scalar, the lower the range of values we can represent.

Assuming twos-complement signed notation, a 32-bit signed integer can accommodate whole values in the range -2,147,483,648 through 2,147,483,647. With a scalar of 10,000, this would accommodate all monetary values in the range £-214,748.3648 through £214,748.3647. While adequate for many smaller businesses, this would be highly unsuitable for large corporations with a profit/loss in excess of this range. However, each additional bit doubles the range such that a 64-bit integer could easily accommodate all monetary values in the range £-922,337,203,685,477.5808 through £922,337,203,685,477.5807.

So, the data type we actually use will be an integer, typically a 64-bit integer. However, in order to make the integer more intuitive and easier to work with, it will ideally be encapsulated by a class that can also encapsulate a user-defined scalar and currency enumerator (such as GBP or USD). In this way, users can work with actual currency in real numbers, such as pounds and pence (including fractions of a penny where appropriate), while the class performs its calculations upon the scaled integer and presents the result back to the user in pounds and pence.