Proposal/Discussion: "Damnit" Postfix Operator (!)

[HaloFour]

This proposal/discussion is an addendum to #36 to make official the proposal of a "damnit" operator which can be used to forcibly unwrap a nullable value into a non-nullable value.

In the current non-nullable reference proposal it would cause a warning to be emitted if a nullable reference is passed/assigned to a non-nullable reference value where flow analysis cannot guarantee that the reference is not null. In those cases the developer could use the postfix ! operator to explicitly instruct the compiler that the value would not be null in that circumstance which would suppress the warning:

string? GetName() => ...;
void PrintName(string name) => Console.WriteLine(name);

string? name = GetName();
PrintName(name);   // emits warning
PrintName(name!);  // suppresses warning

For the sake of symmetry I also propose that this operator be available on nullable value types. It would be equivalent to calling the Value property:

int? length = name?.Length;

int x = length!;  //  equivalent to length.Value or (int)length

Open Questions:

1. Should the ! operator cause an immediate exception if the value in question is null?

[jnm2]

Should the ! operator cause an immediate exception if the value in question is null?

How would it make sense to not throw immediately?

[HaloFour]

@jnm2

That's specifically in the case of a nullable reference type. The compiler could simply accept that the developer knew what they were doing and pass the reference, even if it happens to be null. It saves the cost of the comparison, but not really much else.

For nullable value types there really isn't an option; it will throw.

[jnm2]

If I declare a non-nullable reference parameter, I'd rather have the invariant enforced outside the method. Failing fast when an assumption is breached is the safest and most merciful option in the long run.

[whoisj]

While I still prefer to ban the whole concept of not nullable (sorry, I'm an old fashioned C guy and believe null is very useful), this is going to be the next best thing.

[whoisj]

I'd rather have the invariant enforced outside the method.

How? From the time the value is validated, and the time everything is pushed to the stack, copied, and executed, another thread could have swept through and null'd the value.

[YaakovDavis]

Just having fun:

bool? a, b;
...
var x = !a! != b! ? a! : !b!;

[jnm2]

@whoisj

From the time the value is validated, and the time everything is pushed to the stack, copied, and executed, another thread could have swept through and null'd the value.

Nulled the stack of another thread?

[lachbaer]

var x = !a! != b! ? a! : !b!;

Mh, despite how ridiculous it looks I must say that I understood it instantaneously. 😏

[lachbaer]

Should the ! operator cause an immediate exception if the value in question is null

Yes, ! states that programmer is knowing what she is doing - therefore the '!' character. In case of a Nullable GetValueOrDefault() could be called, but what method should be called on a reference type that is 'null' instead? null throws. Period! 😃

[jjvanzon]

var x = !a! != b! ? a! : !b!;

I can read that. By the way: the last two dammit operators are redundant.

[lachbaer]

@janjoostvanzon

the last two dammit operators are redundant

No, they aren't. The condition !a! != b! is independent of the assignment to var x. To be sure that x is of type bool and not bool? both expressions must be surely of type bool, therefore the 'damnit' cast. Otherwise either the true or false part is of type bool? and thus the resulting type should be of type bool?.

If either of a or b is null, the condition will throw. Only in that regard they are redundant.

[jjvanzon]

@lachbaer:
I get it, it turns var from bool? into bool.
I guess I am not fluent yet in C# 7.?

[lachbaer]

I guess I am not fluent yet in C# 7.

Neither am I 😃

[whoisj]

Nulled the stack of another thread?

Depends on what they point to, so yeah. Also, there's no reason the expectation cannot be value.nonNullValue != null, then you're really hosed.

[jnm2]

@whois So same as the general problem with (new string![1])[0].