[js inteorp] `asA` to correspond to `isA`

[parlough]

With isA, I always have to end up casting to the desired type as well if I want to use members on the specified T type, as we can't rely on Dart's type promotion.

if (node.isA<HTMLAnchorElement>()) {
  final anchor = node as HTMLAnchorElement;
}

I've found myself wanting to add an extension method like:

extension on JSAny? {
  T? asA<T extends JSAny?>() {
    if (isA<T>()) {
      return this as T;
    }

    return null;
  }
}

So I can easily store the result without casting separately within the if.

if (node.asA<HTMLAnchorElement>() case final anchor?) {
  // ...
}

Yes I can add this myself, but I feel this is a pretty common case, so some version of it might be useful in dart:js_interop (if it makes sense). tryAsA could also work as a name if it's desired to distinguish from Dart's as which can throw.

[srujzs]

Similar issue here: #54138

I avoided adding this because I was a little worried users might do both an isA and an asA e.g.

if (node.isA<HTMLAnchorElement>()) {
  var anchor = node.asA<HTMLAnchorElement>();
}

leading to duplicate interop type-checks. That being said, I do like the one-liner of node.asA<HTMLAnchorElement>() case final anchor?, so I'm torn. :)

The one potential pitfall here is if T is nullable here. Then it isn't obvious whether the returned value is null because the input was null or because the type-check failed. I suppose we could restrict T to be non-nullable, but that may be ugly.

[kevmoo]

Similar issue here: #54138

I avoided adding this because I was a little worried users might do both an isA and an asA e.g.

if (node.isA<HTMLAnchorElement>()) {
  var anchor = node.asA<HTMLAnchorElement>();
}

leading to duplicate interop type-checks. That being said, I do like the one-liner of node.asA<HTMLAnchorElement>() case final anchor?, so I'm torn. :)

The one potential pitfall here is if T is nullable here. Then it isn't obvious whether the returned value is null because the input was null or because the type-check failed. I suppose we could restrict T to be non-nullable, but that may be ugly.

Restrict to <T as JSAny> is great. We can document this well, I think.