isspace() relies on the locale() - we currently document the CEL parser as assuming the “C” locale - we should note the same requirement in the XML parser. Later it would be good to have a library providing this functionality which is wired to utf encoding - most likely this would be along the lines of the current parser syntax - something like the following:

namespace implementation {

template<std::size_t S>
struct is_space;

template<>
struct is_space<1>
{
template<typename I>
bool operator()(I& x) const
{
if (space_table[*x]) { ++x; return true; }
return false;
}
};

template<>
struct is_space<2>
//...

} // namespace implementation

template<typename I> // I models InputIterator where value_type(I) == UTF encoded character