9 нояб. 2017 г. 21:34 пользователь "Peter Dimov via Boost" < boost@lists.boost.org> написал: Antony Polukhin wrote:
You got me wondering here. Is it not possible to pass an array of size_t to the size_t* parameter, instead of a single size_t, and then encode > the layout of X1 in it?
Yes, that is what the flat reflection does. But during this operation the X1 type flattens/decays to the the basic types. You can not get X1 from a record of {int, int}.
Ah right. You can only get an anonymous tuple_like<int, int> type, not X1. But isn't this still better than flattening? You could print it as { { 1, 2 }, { 3, 4} }, instead of { 1, 2, 3, 4 }. That's one of the options. Some time ago I thought that is it less usefull than returning flattened type. Nowadays there's a precise reflection that does a lot of things much better, so I see no big need in improving flat reflection.