TITLE: why initialize static data outside class declaration

(Newsgroups: comp.lang.c++, 18 Jul 98)


NATALIE: Ron Natalie <ron@sensor.com>

 > This still doesn't do anything to explain why C++ requires
 > 
 > class foo {
 >    static int x;
 > };
 > 
 > foo::x = 1;
 > 
 > in lieu of
 > 
 > class foo {
 > 	static int x = 1;
 > };

STRUSTRUP: bs@research.att.com (Bjarne Stroustrup)

As a result of #include, class definitions often/typically appear in many
translation units. Had in-class initializers been allowed the implementor
would have had to face a few problems:

	Where should initialization code be generated?

	Is the value used as initializer the same in all translation units?

These problems could be dealt with, but requiring a single unique definition
(as for every other statically allocated object) simplifies matters considerably.

Standard C++ allows one very limited special case of in-class initialization:

	class X {
		static const int max = 7;

		// ...

		int v[max];

		// ...
	};

See "The C++ Programming Language (3rd edition) for details.