void main()
{
    auto a = 0;
    auto b = 0;
    auto c = 1;

    auto ptr = &(c ? a : b);
    *ptr = 8;
    assert(a == 8);
    ptr = &(c ? b : a);
    *ptr = 2;
    assert(a + b == 10);
}

It's just beautiful

On Monday, 7 October 2024 at 20:03:46 UTC, Boaz Ampleman wrote:

void main()
{
    auto a = 0;
    auto b = 0;
    auto c = 1;

    auto ptr = &(c ? a : b);
    *ptr = 8;
    assert(a == 8);
    ptr = &(c ? b : a);
    *ptr = 2;
    assert(a + b == 10);
}

But why pointers?

void @safe main()
{
    auto a = 0;
    auto b = 0;
    auto c = 1;

    (c ? a : b) = 8;
    assert(a == 8);
    (c ? b : a) = 2;
    assert(a + b == 10);
}

On Tuesday, 8 October 2024 at 08:38:15 UTC, Dom Disc wrote:

void main()
{
    auto a = 0;
    auto b = 0;
    auto c = 1;

    auto ptr = &(c ? a : b);
    *ptr = 8;
    assert(a == 8);
    ptr = &(c ? b : a);
    *ptr = 2;
    assert(a + b == 10);
}

yeah sorry, that missed a bit of comments. Here there are: the point was that the conditional expression yields a lvalue, hence it's candidate to be a sub-expression of the address expression. That is just something I had never seen before. I just wanted to share the joy ;)

On Tuesday, 8 October 2024 at 08:38:15 UTC, Dom Disc wrote:

void main()
{
    auto a = 0;
    auto b = 0;
    auto c = 1;

    auto ptr = &(c ? a : b);
    *ptr = 8;
    assert(a == 8);
    ptr = &(c ? b : a);
    *ptr = 2;
    assert(a + b == 10);
}

void @safe main()
{
    auto a = 0;
    auto b = 0;
    auto c = 1;

    (c ? a : b) = 8;
    assert(a == 8);
    (c ? b : a) = 2;
    assert(a + b == 10);
}

That's kinda flexible actually

On Monday, 7 October 2024 at 20:03:46 UTC, Boaz Ampleman wrote:

void main()
{
    auto a = 0;
    auto b = 0;
    auto c = 1;

    auto ptr = &(c ? a : b);
    *ptr = 8;
    assert(a == 8);
    ptr = &(c ? b : a);
    *ptr = 2;
    assert(a + b == 10);
}

With next dmd you can use use ref variables aswell as

void main() {
    auto a = 0;
    auto b = 0;
    const c = 1;

    ref ptr = (c ? a : b);
    ptr = 8;
    assert(a == 8);
    ptr = (c ? b : a);
    ptr = 2;
    assert(b == 2); 			// why does this fail??
}

But for some reason the last statement doesn't behave as I believe it should. Is this behavior expected?

For reference see https://dlang.org/changelog/pending.html#dmd.reflocal.

On Friday, 11 October 2024 at 23:04:20 UTC, Per Nordlöw wrote:

void main() {
    auto a = 0;
    auto b = 0;
    const c = 1;

    ref ptr = (c ? a : b);
    ptr = 8;
    assert(a == 8);
    ptr = (c ? b : a);
    ptr = 2;
    assert(b == 2); 			// why does this fail??
}

This has to be a bug ;)

On Friday, 11 October 2024 at 23:04:20 UTC, Per Nordlöw wrote:

void main() {
    auto a = 0;
    auto b = 0;
    const c = 1;

    ref ptr = (c ? a : b);
    ptr = 8;
    assert(a == 8);
    ptr = (c ? b : a);
    ptr = 2;
    assert(b == 2); 			// why does this fail??
}

ref can’t change what it is pointing to. So the second ternary is just assigning through the reference, not re-binding the reference.

If you re-write the example in terms of pointers, it is equivalent to:

void main() {
    auto a = 0;
    auto b = 0;
    const c = 1;

    int* ptr = &(c ? a : b);
    *ptr = 8;
    assert(a == 8);
    *ptr = (c ? b : a);
    *ptr = 2;
}