Allow type promotion in conditions like while((a = b) != null) { ... }

[mmcdon20]

I'm not sure if there is an existing issue for this, I wasn't able to find one.

Consider the following program:

void main() {
  final items = [6, null, 1, 2, 3, 4, 5];
  int? item;
  while ((item = items.removeLast()) != null) {
    // item should be promoted to int, instead we get the following error:
    // The operator '*' can't be unconditionally invoked because the receiver can be 'null'.
    print(item * 100);
  }
  if ((item = items.removeLast()) != null) {
    // item should be promoted to int, instead we get the following error:
    // The operator '*' can't be unconditionally invoked because the receiver can be 'null'.
    print(item * 100);
  }
}

It would seem reasonable for a condition of the form (a = b) != null to promote a from SomeType? to SomeType, assuming a is a local variable.

Promotions should also work with (a = b) == null, (a = b) is SomeType, (a = b) is! SomeType.

Promotions should probably also apply when using ??= as in (a ??= b) != null.

And when doing multiple assignments as in (a = b = c = d) != null where a and b and c are all local variables they would all be promoted.

[mateusfccp]

Not exactly what you asked, but this could be solved if we had patterns in while.

void main() {
  final items = [6, null, 1, 2, 3, 4, 5];
  while (items.removeLast() case final item?) {
    print(item * 100);
  }
  if (items.removeLast() case final item?) {
    print(item * 100);
  }
}

[mmcdon20]

There is already an issue for patterns in while #2536.

This is a request about type promotion not patterns, so I think there is room for both proposals.

[lrhn]

See #3658 and #1420 (assignment promotion).

[mmcdon20]

@lrhn looks like this is a duplicate in that case.