r: support a precision factor for comparisons
Of course we should not test for exactly 0, we should use a precision factor. Have a look about what Knuth wrote about comparing floats, as an absolute difference is not such a good idea.
We can generalize R to include a dynamic precision member, with a default value of course. This is a very nice example of a weightset that needs a dynamic member (so far, Value/ValueSet was easier to explain for labelsets where the alphabet must be shared).