Store price in double?

Question

Is it safe to store prices in double variable in C++? In C# and Java you can use Decimal. (They also say that storing prices in double is very bad idea). What should I do in C++?

Answer 1

Depends on accuracy that you need. If you are going to work with numbers like 5.25, 9.14, etc., you don't even need double. float will be good enough.

If you are going to work with account balance and add 0.01% to number like 18 570.548911545, then precision of 15 decimal digits that double provides will be enough.

If you need a slightly higher precision, go for long double. If that's not enough, only then look for "more precise alternative".

Answer 2

If the price range is suitable (probably it is), you can use long long with an implied dot before the last two digits. So, 12000 will mean 120.00. It's called "fixed point" arithmetic. But you need to take care of multiplications of two currency values (why would you need it anyway).

Answer 3

You can have a look at Intel® Decimal Floating-Point Math Library for storing price. Also you can check the Boost.Multiprecision library which can be very useful, you can use decimal based floating point template class called cpp_dec_float

Answer 4

I'd be tempted store prices in terms of the smallest fraction of currency you care about using int or unsigned int. So if an item costs $5.99 and you don't track values smaller than one cent, store it in an int as 599.

Answer 5

Unfortunately, you have the same Problem in C++ as everywhere else. You can't trust double to store the values correctly, although its much better than a float.

There's no data-structure like Decimal avaiable in C++, so you either need to look for a Library that does this for you, you you use an other format.

A simple solution would be, to store the value "42.37" as "4237" - this is how you can trust you values. But I would still recommend using double, if you are calculating with this numbers.

If you decide to use bigger variable-types like long double, you can reduce the possibilty that something goes wrong - but it doesn't guarantee it.

Answer 6

You can use long double to store the price

// get_money manipulator example
#include <iostream>     // std::cin, std::cout
#include <iomanip>      // std::get_money

int main ()
{
  long double price;
  std::cout << "Please, enter the price: ";
  std::cin >> std::get_money(price);

  if (std::cin.fail()) std::cout << "Error reading price\n";
  else std::cout << "The price entered is: " << price << '\n';

  return 0;
}

Answer 7

From many years of ERP development experience...

Never store pricing in anything but integers (int, long, long int, int64, etc.) as a function of qty with implied decimal places (i.e. storing $4.00 as 400)

An example. Think of a price break on a product.

1. qty 1 for $1 --- unit price $1
2. qty 25 for $13.33 --- unit price $.393939393939 repeating for ever.

Storing the unit of line 2 in a double would cause rounding. Instead of storing as a double, have one int variable for price 1333 and one for qty 25 and calculate that unit price whenever it needs to be displayed.

When stored as double, float, etc., you always risk rounding/losing precision in converting that decimal value to binary.