On 29/08/2017 04:12, Paul A. Bristow wrote:
On 29/08/2017 02:54, Robert Ramey wrote:
Just one more uninformed opinion, but wouldn't this work better as an addition to the boost units library?
I'd suggest not, on the grounds that physical properties are inexact and accepted values can and do change, while math constants really are constant (though their representation may change depending on the type used to represent them - as 32, 64 or 128 bit doubles for example). I'd prefer to keep these two different things separate.
Of course, the algorithms should not change, but generally involve some inconstant constants, and so they do change.
It sounds like you are arguing against inclusion in Boost.Math, which is not what was suggested. There is already precedent for Boost.Units containing physical constants -- see boost/units/systems/si/codata/*.