bill@proxftl.UUCP (T. William Wells) (05/28/88)
I noticed a discussion that seems to be about declaring constants
of other than the simple types. While I do not advocate adding
them to the language (though I would not necessarily oppose
adding them), I do have a suggestion for a syntax for them.
'(' type-name '=' initializer ')'
This should be considered as creating an anonymous variable of an
appropriate type which is initialized according to the
initializer and then used in the expression the constant is
contained in. The existence and modifiability of the anonymous
variable would be implementation dependent.
For example, an expression of the form:
typedef struct foo { int x, y; } foo;
foo z, *p;
z = (foo = {1,2});
might be compiled as if it were written:
z.x = 1;
z.y = 2;
And
p = &(foo = {1,2});
would require that the constant exist (so that p could point to
it) but would not require that the constant be modifiable.
Anybody got a better idea? Does anyone believe that the
resulting grammer or semantics would be ambiguous?