Why didn't you provide conversions to and from float/double types?

Although it might not be obvious, there is occasionally a need for such conversions. Unfortunately, unlike integers, there is no universal representation for floating-point values. Even IEEE 754, the most widely used floating-point specification, isn't completely nailed down -- it doesn't specify the endianness of the fields, making it difficult to provide code that would work on all systems. What's worse, the C/C++ float and double types don't have to be encoded in an IEEE 754 format at all.

I'd like to provide such conversions, and I'll keep looking for ways to do so. It may turn out to be possible to do them through a string format, for instance. But for now, the library only provides conversions to and from integral values.