-----Original Message----- From: boost-bounces@lists.boost.org [mailto:boost-bounces@lists.boost.org] On Behalf Of Tobias Schwinger Sent: Wednesday, March 02, 2005 9:21 AM To: boost@lists.boost.org Subject: [boost] Empty sequence elements
Paul Mensonides wrote:
BOOST_PP_IF(HAS_COLOR(t), and color is COLOR(t) , );
^^^ This is undefined behavior in C++. Should be:
May a PP-Sequence have '()'-elements ?
Not until C++ gets placemarkers (i.e. empty arguments) from C99. Without them, "emptiness" is not an element. '()' is just nullary parentheses. Without variadics, a mechanism can only handle sequences in which each element is a specific arity. In the pp-lib, they must be unary (i.e. not nullary). It is certainly possible (in current C++) to define primitives for (e.g.) a binary sequence, but I'd have to replicate the entire mechanism. Even if binary sequence primitives where there, you still couldn't have (,) or (a,) or (,b) because each "element" would be a comma-separated pair of elements, and emptiness is not an element.
The documentation says that "a sequence cannot be empty" which seems to be something different than containing empty elements...
Yes, when the documentation says the above it means there is no such thing as a nil sequence. E.g. (a)(b)(c) // 3-element, unary sequence (a)(b) // 2-element, unary sequence (a) // 1-element, unary sequence // 0-element sequence, n-ary sequence The 0-element sequence has a representation (emptiness), but that representation is not yet valid C++. In other words, the reason that you can't have empty sequences and can't have empty elements is the same reason. This applies generally: you can't have empty arguments, you can't have empty tuple elements, you can't have empty list elements, etc.. With placemarkers, that will change, because placemarkers are a well-defined way of saying that an entity can be nothing. Until that happens, it is undefined behavior. :( Regards, Paul Mensonides