------- Comment #5 from damian at rouson dot net 2010-05-04 14:25 ------- (In reply to comment #4) > What a horrible rule... >
I'm not sure why you don't like it, but the reason for the rule is to have the ability to overload the intrinsic structure constructors. The intrinsic structure constructors are virtually useless for object-oriented programming (OOP) because they can't be used if the type has private data (required by the information-hiding philosophy of OOP) unless all components have default initializations (which only helps in trivial cases). It's standard in C++ (and I'd guess in most other OOP languages) for constructors to have the same name as the type. What could be more clear to the reader than type(vector) :: f f = vector(1.0,2.0) Once one adopts this approach, it touches every type you develop, so a compiler that doesn't support it becomes useless unless you're willing to change every constructor name to suit that particular compiler. It's already supported by IBM XL Fortran (my primary development platform), by Cray, by NAG (I think), and Intel has promised to support it in its next release. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39427
