Champion "tuple projection initializers" (C# 7.1, VB 15.3)

[jcouv]

• Proposal added infer-tuple-names.md
• Discussed in LDM (2017-04-05)
• Decision in LDM (and reviewed with compat council)
• Finalized (done, rejected, inactive)
• Spec'ed

See also #370
A working implementation is at dotnet/roslyn#18374

[NickCraver]

Can you please fix the link in https://github.com/dotnet/roslyn/blob/master/docs/Language%20Feature%20Status.md when the proposal is added?

[jcouv]

@NickCraver Will do. I think the link will actually be correct, but I need to write the document.

[DavidArno]

@jcouv,

I have run into an issue with this feature in Preview 7.0 of VS2017 Update 3. Consider the following example code:

public static void Foo()
{
    (T1, T2) Bar<T1, T2>((T1, T2) value) => value;

    var tuple = (a:1, b:2);
    var newTuple = Bar(tuple);
    var a = newTuple.a;
}

I get an error, "(int, int)' does not contain a definition for 'a' and no extension method 'a' accepting a first argument of type '(int, int)' could be found..." on the last line.

When it comes to generics, this tuple name inference appears to be limited to having to express the whole tuple as, eg T. If I change the code to:

public static void Foo()
{
    T Bar<T>(T value) => value;

    var tuple = (a:1, b:2);
    var newTuple = Bar(tuple);
    var a = newTuple.a;
}

then the code compiles. But then, my Bar isn't restricted to a two item tuple; it'll take any type. As ValueTuple<T1, T2> is a struct, I can't constrain T using generic constraints either.

Is this a bug/oversight, or is what I want to do completely outside the scope of this new feature?

[HaloFour]

@DavidArno

I think that behavior is correct. In your first example Bar is returning a new unnamed tuple so it doesn't follow that the names would carry over through the individual generic type arguments. Bar could be defined to instead return (T1 a, T2 b) instead. In the second example the entire generic type argument T could be determined to be (int a, int b) so the names can carry over.

[DavidArno]

@HaloFour,

I think you are right: it's not a bug. It's a (very necessary, for me at least) extension to this feature. As it stands, if I method-chain with T being substituted with (x, y), (x, y) is available throughout the fluent expression. If I method-chain with (T1, T2) being substituted with (x, y), (x, y) is not accessible to the expression. So v7.1 is an improvement on v7, but it's not quite there yet.

I shall create a new proposal issue therefore.