John McFarlane

@hbe72 could you try it now? Thanks

Absolutely. Generic solutions are always better. Don't forget `int`->`int` and `int`->`float` also.

In answer to your previous question about `cnl::convert` vs `cnl::_impl::convert`, `convert` is primarily a customization point. The `_impl` function is an internal helper which you're welcome to reimplement until I...

Also on the TODO list: `convert` with `rounding_integer` input and result.

Proved. Now add to a paper.

In other words, prove that an special purpose intrinsic, e.g. from an embedded systems chipset, could be injected into `fixed_point`.

I think I've mentioned this in person but `ASSERT_DEATH` is the _only_ thing keeping using GTest both at work and at home. It's vital to testing contract tests.

As discussed previously, non symmetric binary arithmetic operators (except ``) are typically ill-advised. From item 26 of [C++ Coding Standards](http://www.gotw.ca/publications/c++cs.htm) entitled *Preserve natural semantics for overloaded operators*: > Programmers expect...

Requirements for currencies is subtlety different to units. The constant scaling factor is normally a power of ten. It isn't always the power you'd expect: sometimes it's tenth of a...

+1 for a low-level linear algebra library for real-time simulation! I'd really like to fork [GLM](https://github.com/g-truc/glm) and [fix it](https://github.com/g-truc/glm/issues/532) so that it worked with numeric types like those in units...