aliasing for fields does not work

[WalterBright]

struct T {
    int k,i = 2;
}

struct S {
    int x;
    T t;
    alias ti = t.i;
}

int main() {
    T t = T(1, 2);
    S s;
    assert(s.ti == 2); // Error: accessing non-static variable `i` requires an instance of `T`
    return 0;
}

s.ti should be equivalent to s.t.i

[WalterBright]

fix: #20611

[TurkeyMan]

Wonderful!

Will this also address elements of an AliasSeq?

alias seq = AliasSeq!(x, y, z);
alias x = seq[0];

[WalterBright]

Unfortunately, not. Wanna submit a bug report?

[TurkeyMan]

In general, I think it might be really useful to make alias work generally with expressions.
Since I think the reason that alias x = seq[0] doesn't work, is because it is effectively aliasing a [] operator expression? There are so many situations that fall into this general category where alias fails; is it out of the question for alias to be able to carry an expression?

Although, my understanding was that the . operator which accesses a member was also a member-access-expression, and that's why alias didn't work on that... which is why I thought this might have addressed the AliasSeq situation too...?

[TurkeyMan]

Unfortunately, not. Wanna submit a bug report?

#20612

[WalterBright]

Ehhxcellent!

[WalterBright]

There are already ambiguities that I struggled to account for in fixing this issue. More aliasing means more trouble with this. Aliasing tries to do too much!

Aliasing an expression is better done with a lambda, where it is well defined. The lambda will get inlined, and all will be well.

ImportC creates lambdas out of macros that look like expressions. This works quite well.

[TurkeyMan]

Aliases are more often used to refer to types or symbols... In the cases of interest if a lambda was workable, there wouldn't be anything to talk about.