boost::polygon has a nice facility for generic numeric types for coordinate values via coordinate_traits, etc. But there are a number of places in the library code where a hard-coded type, usually a double, is used instead of referring to the appropriate type in this trait struct. Should this be considered a bug in b::p? Background: I'd like to use a custom class as a type which manages the scaling of real-world decimal coordinates to the integral values that b::p likes to use. I understand the whole thing about snap-rounding to an integral grid and all that, so need to scale up the real-world decimal values to an appropriate grid size. But it makes sense to carry this paradigm through to all of the typedef's in coordinate_traits and high_precision_type, but this is where things break with the hard-coded double's. T -- View this message in context: http://boost.2283326.n4.nabble.com/boost-polygon-type-genericity-tp4651484.h... Sent from the Boost - Dev mailing list archive at Nabble.com.