Discussion: extend the scope of parameter names to attributes

[alrz]

[SomeAttribute(nameof(param))]
void M(object param, [CallerArgumentExpression(nameof(param))] string arg = null) {}

PS: This is different from #373 as it proposes to allow parameters in nameof from any method.

[gafter]

This would be an incompatible change. Instead of binding to a constant from the enclosing scope, an attribute argument would now bind to a (non-constant) parameter, which would make a currently compiling program an error.

[jamesqo]

@gafter Do you mean something like

private const string foo = "";

void M(string foo, [AttributeAcceptingString(foo)] string blah) {}

would break, because it would bind to the parameter foo instead of the constant foo? In that case, why not check to see if there is a constant named foo before checking for parameters?

[alrz]

@jamesqo this change would not affect your example, because foo as a parameter has no meaning in an attribute. However, it affects a nameof(constant) if there is also a parameter named constant, with no observable difference (it's the string "constant" either way). Anywho, as you said, with a specific lookup order this issue will be resolved.

[taori]

Accessing the name of argument names with this version is very inconvenient in comparison to a modification of the nameof expression as described in #373 i would say. (though bringing it in scope in gernal certainly would not hurt)

[alrz]

Closing in favor of championed issue #373