Static break and continue

[Bolpat]

A static foreach statement (and only that, not a foreach statement over a compile-time sequence) should permit static break and static continue statements, possibly labeled to discriminate which static foreach is meant. The unlabeled version is always admissible and refers to the innermost static foreach. Due to the static keyword, conflation with a non-static loop or other construct is unlikely.

In every loop iteration, it generates code. Essentially, the loop body is copied and the iteration variable replaced by its value. If it contains any static break or static continue that refer to the current loop, any static foreach and static if declarations must be be handled. If any static break or static continue remain, the first takes precedent. If it is a static break, anything after it must be discarded and the loop must not generate another instantiation. If it’s a static continue, anything after it must be discarded before the next loop iteration.

(Anything said for static foreach also applies to static foreach_reverse.)

[Herringway]

This was noted as a future expansion to DIP1010 as well, though not fleshed out there.

[dkorpel]

What do you propose this compiles to?

static foreach(i; 0 .. 1)
{
    alias xAlias = x;
    static if (x == 10)
        static break;
    immutable x = 10;
}

pragma(msg, xAlias); // ?
pragma(msg, xAlias.mangleof); // ?

Edit: better example. The question is, what happens to forward references to (or 'side effects' of) discarded symbols?

[zero-vector]

I dusted of my old toy implementation, and forward references like mentioned would go trough, but even if this resulted in some ambiguity error, or silent weird behavior, I think its in the realm of FAFO you get when overusing D meta-programing capabilities.
That said, there are use cases when break/continue would be helpful, but issues like breaking a nested static foreach has to be tackled as well.