Best practices for dealing with monetary amounts in C#

Question

I have read a couple of articles about the perils of certain data types being used to store monetary amounts. Unfortunately, some of the concepts are not in my comfort zone.

Having read these articles, what are the best practises and recommendations for working with money in C#? Should I use a certain data type for a small amount and a different one for a larger amount? Also, I am based in the UK which means we use , (e.g. £4,000, whereas other cultures represent the same amount differently).

Answer 1

Decimal is the most sensible type for monetary amounts.

Decimal is a floating point base 10 numeric type with 28+ decimal digits of precision. Using Decimal, you will have fewer surprises than you will using the base 2 Double type.

Double uses half as much memory as Decimal and Double will be much faster because of the CPU hardware for many common floating point operations, but it cannot represent most base 10 fractions (such as 1.05) accurately and has a less accurate 15+ decimal digits of precision. Double does have the advantage of more range (it can represent larger and smaller numbers) which can come in handy for some computations, particularly some statistics computations.

One answer to your question states that Decimal is fixed point with 4 decimal digits. This is not the case. If you doubt this, notice that the following line of code yields 0.0000000001:

Console.WriteLine("number={0}", 1m / 10000000000m);

Having said all of that, it is interesting to note that the most widely used software in the world for working with monetary amounts, Microsoft Excel, uses doubles. Of course, they have to jump through a lot of hoops to make it work well, and it still leaves something to be desired. Try these two formulas in Excel:

• =1-0.9-0.1
• =(1-0.9-0.1)

The first yields 0, the second yields ~-2.77e-17. Excel actually massages the numbers when adding and subtracting numbers in some cases, but not in all cases.

Answer 2

You should not use floating point numbers because of rounding errors. The decimal type should suit you.

Answer 3

Martin Fowler recommends using a Money class. See the link for the rationale. There are a number of implementations of his idea out there, or you could write your own. Fowler's own implementation is in Java, so he uses a class. The C# versions I have seen use a struct, which seems sensible.