bronek/juno

Some C++17 metaprogramming

juno's Issues

Function naming in type_set

#Member functions in type_set have confusing names and ideally (some of them) should be renamed. Here is how function currently map to mathematical concepts, starting with creation of sets.

• let A be an empty set : using A = type_set<>
• let A be set containing elements E and F : using A = type_set<E, F>
• is A a subset of B ? B.is_in<A>()
• is E an element of B ? B.is_in_setof<E>() (it is a shortcut for B.is_in<type_set<E>>())
• are all of E, F, G elements of B ? B.is_in_setof<E, F, G>() (a shortcut for B.is_in<type_set<E, F, G>>())
• is an empty set an element of B ? (guaranteed to be true, also when B is an empty set) B.is_in<type_set<>>() or B.is_in_setof<>()
• is there a non-empty intersection of sets A and B ? (or in other words, is any of elements of B also in A) B.is_any<A>()(or reverse, obviously; note B.is_any<type_set<>>() is guaranteed to be false)
• is any of elements E, F, G in B ? B.is_any_setof<E, F, G>() (it is a shortcut for B.is_any<type_set<E, F, G>>() ; note B.is_any_setof<>() are guaranteed to be false)
• let C be an union of sets A and B : using C = decltype(A.join<B>()) (or typedef decltype(A.join<B>()) C, or reverse, obviously)
• let C be an union of set A and an unnamed set with elements E and F: using C = decltype(A.join_setof<E, F>()) (it is a shortcut for using C = decltype(A.join<type_set<E, F>>()))
• let C be an intersection of sets A and B : using C = decltype(A.cross<B>()) (or reverse, obviously)
• let C be an intersection of set A and an unnamed set with elements E and F: using C = decltype(A.cross_setof<E, F>()) (it is a shortcut for using C = decltype(A.cross<type_set<E, F>>()))
• let C be a relative complement of B in A : using C = decltype(A.less<B>())
• let C be a relative complement of B in an unnamed set with elements E and F: using C = decltype(A.less_setof<E, F>()) (it is a shortcut for using C = decltype(A.less<type_set<E, F>>()))

Other function names seem OK, as they are consistent with other C++ libraries, for example:

• are sets A and B equal: B.is_same<A>() (std::is_same is similar although lower-level concept)
• is set A empty: A.empty()
• number of elements in set A: A.size()

The challenge here is that I insist on names which are:

1. consistent with other libraries (hence is_same , size and empty should stay) and
2. which are concise (i.e. little typing involved) and
3. internally consistent with other functions in type_set.

Functions which produce a new set should be verbs or mathematical terms. Last but not least, the function names should make sense mathematically, even if strict terminology is not followed (currently some of them do not make sense, e.g. is_in is in fact reverse to implied meaning).

There is no helping to verbosity of syntax when using the library, i.e. sharp brackets or keywords using ... decltype to return result from a function. This an unfortunate side effect of compile-time programming being an ugly duckling in the C++ language.