ValueWithError

C++98/11 header-only class for gaussian error propagation.

It allows to answer the question "What is the error of measured values with errors after an arithmetic operation?" for the general case.

For example

[5.0 +/- 0.3] + [2.0 +/- 0.2] = ?

can be implemented as

#include "ValueWithError.hpp"

int main()
{
  using error_propagation::make_value;
  auto a = make_value(5.0,0.3);
  auto b = make_value(2.0,0.2);

  std::cout << "a * b = " << a * b << std::endl;
}

and gives a * b = (10+-1.16619).

It is designed as a drop-in replacement for an arithmetic type as all arithmetic operations from the C++ standard library are implemented.

For the internal value type the following types have been tested:

• double
• float
• boost::multiprecision::cpp_dec_100/50
• Eigen::Array
• std::valarray

Requirements (minimum versions)

• C++ compiler (see below)
• Boost 1.55
• doxygen 1.8.5 (only for building the documentation)
• cmake 2.8.11 (only for building the test cases)

Usage

1. git clone https://github.com/t-b/value-with-error.git
2. Add value-with-error/src/cpp11 or value-with-error/src/cpp98 to your include path
3. Add #include "ValueWithError.hpp" and start using error_propagation::ValueWithError<T,P>

Tested C++ compiler

Windows

• C++98
  • VC 8-11
• C++11/C++98
  • VC 12
  • mingw gcc 4.8.1
  • mingw clang 3.3

Linux

• C++98
  • gcc 4.6
• C++11/C++98
  • gcc 4.7/4.8/4.9 trunk
  • clang 3.3/3.4 trunk

License

This software is released under the "Boost Software License 1.0" see http://www.opensource.org/licenses/bsl1.0.html.