Added Accepted label.

This comment was originally written by @mhausner

I proposed to write Symbol literals as #"abc" instead of #abc. This would make #"abc=" a regular case, and we could also express operator symbols as literals. I think it also expresses better what a symbol literal is: a string (with some special semantics for compilers). Lastly, it is similar syntax to raw strings.

Thinking about the last point some more, we could also define symbol literals as s"abc". But I won't push my luck any further here.

This comment was originally written by @seaneagan

String syntax might make people think they can put any String, not just identifiers, and that it returns an actual String, as opposed to a Symbol.

What do you mean by "regular case"?

Why couldn't operator symbols be exposed with (an extension of) the current syntax?

#()
#<
#<=
#==
#+
#-
// ...

unary - is always tricky, maybe:

#-1

This comment was originally written by @seaneagan

Oh, my bad, looks like the spec already allows operator symbol literals!

Still valid, there is no way to express the symbol for a private setter. You can get a private getter name as #_foo, an a non-private setter name as const Symbol("foo="), but there is no way to create _foo= as a constant symbol.

You need an elaborate noSuchMethod-based approach to create such a symbol at runtime, which makes it ineligible for being used in a switch case.

void main() {
  print(#_foo);
  print(const Symbol("foo="));
  print(_fooSetSymbol);
}

final Symbol _fooSetSymbol = symbolOf((d) => d._foo = 0);

// Imagine it's in another library, so `_last` won't get in the way.
Symbol symbolOf(void Function(dynamic target) access) {
  var s = _Symbolizer();
  access(s);
  return s._last ?? (throw StateError("No member access"));
}

class _Symbolizer {
  Symbol? _last;
  noSuchMethod(Invocation invocation) {
    _last = invocation.memberName;
  }
}

Right, but then I can see that we need an issue for dart2js: #54603.