Inverse operator for ??

[andrey-messi]

I think this is a duplicate request, but..
If we have this code^
W q = w != null ? w : new W();
then we can shorten it in this code:
W q = w ?? new W();
But I often need to make logic that inversed, I need this:
Y q = w == null ? null : new Y();
It is very long expression and need to be shortened. Every time I hope in new version C# we will have something like
Y q = w !? new Y();
but when I try it in my current net8 project it give compile error.

Whar the problem to realize this function
Y q = w !? new Y();
?

[Richiban]

Sorry, but I'm really struggling to see where this would ever be useful; you want reassign a variable only if it's not null?

Can you give a more concrete or real-world example of where this might be useful?

[TheUnlocked]

Their request is similar to Option.map in languages that have that. A common use of that is to get members of something that could be null (which is what the ?. operator already does), but sometimes it can be useful to run a function on the value, preserving nulls.

With that said, I haven't had a particular need for this in C#. I do often find myself wanting a similar operator in JS/TS, but extension methods paired with ?. cover a lot of use cases in C#, and C#'s nulls aren't nearly as flexible as in JS anyways.

[suugbut]

Please kindly see my proposal. #7855 (comment)

[dersia]

I on the other hand have a lot of use for this feature in recent development.

[jnm2]

The specific syntax a !? b would be complicated to do because it's the same as a! ? b which might already appear in C# as part of a conditional expression. Not impossible, but not appealing. (SharpLab)

[andrey-messi]

What does "!" in "a! ?? expr" mean? As I know, in that case "!" after "a" just tell a static analizer that "a" is not null to suppress showing warnings. But guaranteed not null "a" has no sense when you need
Y q = a !?? new Y(a);
So I think it is possible to use "!??"

[HaloFour]

It is the null forgiveness operator and does behave as you have described. But since it is already legal code today that cannot be changed.

[HaloFour]

Syntax aside, I think the value of having any language feature here has to be demonstrated. I don't think it's a remotely common pattern and the existing code via ternary conditional operator is plenty terse.

A map like operator could be interesting, but it's useless unless it has access to the unwrapped value.

[andrey-messi]

"existing code via ternary conditional operator is plenty terse." - so in the next version of C# we should remove ?? ...

" I think the value of having any language feature here has to be demonstrated. "
for example. i have logic with car and garage . If i have car i need garage, if i dont have car i dont need garage.

What can I do?
1)
Garage q = car == null ? null : new Garage();
2)
Garage q;
if(car != null)
q = new Garage();

What I want:
Garage q = car !?? new Garage();

Because in real coding it is not rare than "W q = w != null ? w : new W();" case that has ?? operator.

[HaloFour]

Wanting a default fallback for a null value is a very common occurrence and justified the language feature. I don't believe this request is remotely as common, thus it does not justify a language feature specifically to support. I don't see why either of the two alternatives you've presented would be considered onerously verbose.

[suugbut]

I propose the following syntax:

• x ?< y equals to x is null ? y : x equals to the existing x ?? y
• x ?> y equals to x is null ? x : y

How to memorize?

If the left expression x is null for

• x ?< y then y will be applied as Pacman's mouth is facing to y.
• x ?> y then x will be applied as Pacman's mouth is facing to x.

Non-trivial examples for x ?> y cases:

Case 1

Buy a car only if we have a garage.

Garage? garage = null;

Car? car = garage is null ? null: new Car(); // normal code
Car? car = garage ?> new Car(); // proposed alternate

class Car;
class Garage;

Case 2

isNut is

• true if fruit is Nut
• null if fruit is null
• false if fruit is neither Nut nor null

Fruits? fruit = null;

bool? isNut = fruit is null ? null : fruit == Fruits.Nut; // normal code
bool? isNut = fruit?.Equals(Fruits.Nut); // normal code
bool? isNut = fruit ?> fruit == Fruits.Nut; // proposed alternate

public enum Fruits
{
    Nut,
    Kiwi,
    Melon
}

Note: For my previous proposed syntax, see #7881 (comment)

Contemplation

For the sake of fairness, let's consider the following:

• Why do we have x ?? y when we can just write x is null ? y : x that is more expressive, readable, idiomatic?
• Why do we have x?.Something when we can just write x is null ? null : x.Something that is more expressive, readable, idiomatic?
• To be honest, all of these expressions x ?? y, x ?< y, x ?> y have the same degree of "obscurity" for new comers until they learn and become familiar with.

[suugbut]

But car is of type Car?. It is clear that null is possible for car.

[tacosontitan]

To be honest, both x ?? y or x ?< y have the same degree of "obscurity" for new comers until they learn and become familiar with.

That's not equivalent at all. In a = b ?? c it is explicitly clear that only b or c will be assigned to a, regardless of what ?? actually means. In a = b ?< c, it is explicit that b or c can be assigned but null can also be assigned, making it implicit behavior.

[suugbut]

It is not implicit behavior.

• The type of a in either a = b ?< c or a = b ?? c is not nullable.

  // normal scenario
  int c = 1;
  int? b = null;
  int a = b ?? c;
  int a = b ?< c; // equivalent

• The type of a in a = b ?> c is nullable.

  // normal scenario
  int c = 1;
  int? b = null;
  int? a = b ?> c;

[viceroypenguin]

I'm gonna be honest, Car? car = garage ?> new Car(); is problematic to me for an entirely different reason.

With a = b ?? c, you always have the explicit agreement that a = b and a = c are valid statements. You know that you could assign either b or c to a. The operator is shortening which one.

With Car? car = garage ?> new Car(); you break that assumption. You can, at no time, ever do Car? car = garage (if Car and Garage are independent types with no assignment relation). Even if garage == null, you still can't do car = garage, because they are independent types. Adding ?> capability of doing so would break all sorts of readability expectations.

[suugbut]

New comers who read int? id = person?.Id; also might get confused how person of type Person? can be assigned to id of type int? when person = null, there is a mismatch type here. It really breaks all sorts of readability expectations. It does not obey The Principle of Least Surprise (POLS). 😊

Person? person = null;

int? id = person?.Id;

class Person
{
    public int Id { get; set;}
}

So actually behind the scene, the compiler can do something as follows.

Garage? garage = null;
Car? car = garage ?> new Car();
Car? car = garage is null ? (Car?) garage: new Car();
Car? car = garage is null ?  null : new Car();

Getting familiar with new things can only be achieved by learning.

[tacosontitan]

I'm against this idea because it seems like a request to support writing shorter code for the sake of writing shorter code, especially given your statement:

But I often need to make logic that inversed, I need this:

Y q = w == null ? null : new Y();

It is very long expression and need to be shortened.

_{Emphasis mine.}

The use case q = w is null ? null : new Y() just screams "refactor needed", and without context (to me at least), doesn't really make any sense. I second @Richiban's question:

Can you give a more concrete or real-world example of where this might be useful?

I do like the fact that you made a comment to another poster using the car and garage example, but it still left me asking the question "why are you creating a variable you don't need?". Everything I came up with around this ended in me refactoring it to something more legible than creating an unnecessary variable. For example:

Car? car = await service.GetCar(id: 12345, cancellationToken);
if (car is null)
    return;

Garage garage = new(car);
...

[tacosontitan]

@andrey-messi, I feel as though I've really beaten my point into the ground at this point, so this will be my last response to this discussion as a whole because I have no interest in this feature.

Being terse, for the sake of being terse is never a good idea. This is a niche use case that really doesn't have a method (in my opinion) to explicitly represent intent and outcome in a more concise manner than a ternary, and it's better in those scenarios to simply be more verbose for the sake of legibility. A few other points that I've been really clear about avoiding are:

• Don't create variables you don't need.
• Don't assign to properties if you don't need to.
  • Why would you explicitly assign null if that's the default for your type?

My stance remains unchanged in that the many things I've mentioned are much more legible than even using a ternary to solve this problem.

• Apply inversion with a guard clause and early return.
• Assign to the property conditionally before returning.

I wish you the best of luck!

[andrey-messi]

I dont tell that "inverse ?? Operator" is a thing that i strongly need. What i want is just a syntax sugar for that logic.
And by my statistics i use case
A a = b == null ? Null : new A()
Not rare but maybe even often than
A a = a == null ? New A() : a
For that reason why should i think that my cases are "This is a niche use case"?

So if to want null is "a niche use case" then why we have this syntax sugar

Garage g = car?.ReturnGarageForThisCar();

We can always do that:
Garage g = car==null ? null : car.ReturnGarageForThisCar();

[suugbut]

@andrey-messi : Out of topic. To make your code well formatted, use the following template.
```csharp
// your code goes here
```
where ` is a backtick (a key to the left of 1 on our QWERTY keyboard).

[HaloFour]

then why we have this syntax sugar

The answer is going to always be that it was found to be a common enough scenario to warrant a language feature. The desire to navigate deep call chains without nested conditional statements is common enough that the null propagation operator exists in quite a few programming languages.

This case is much less common, and the suggested behavior of such a proposed operator to be very unintuitive.

[Daynvheur]

There is a difference in ?? vs !? (or whatever other inverted character I can't have without a numeric pad):
In favor of ??:

int a = GetIntFromDatabase() != null ? GetIntFromDatabase() : -1; // This needs a method call refactor

int? b = GetIntFromDatabase(); // This is a method call refactor
int aBis = b != null ? b : -1; // This is equivalent to int aBis = b ?? -1 // This is equivalent to a = GetIntFromDatabase() ?? -1;

In defavor of !?:

int? a = GetIntFromDatabase() != null ? -1 : null; // This doesn't need a method call refactor

int? b = GetIntFromDatabase() !? -1; // This isn't shortening that much, and hides the defaultening intent

[suugbut]

[viceroypenguin]

@suugbut I'm in the same boat. You are adding more items to make it appear as if the existing votes are in favor of the certain ideas, when they are not. Please stop attempting to manipulate results like this.

[CyrusNajmabadi]

My downvote is an actual downvote. not an actual vote for the second line. I don't like any of thse.

[suugbut]

The table was designed from the beginning for voting and I am updating this table gradually.

[dersia]

my 🫤 was not for the option that I just mentioned because there was a ask, I don't proposed that solution nor was my emote ment to be used for that. I just chose an emote that wasn't an option at that time to express that I don't vote for any of the options. so if you'd be so nice to remove me from the list or remove that option completely, I'd appreciate it

[suugbut]

I have made a new one to avoid misunderstanding. #7855 (comment) I hope the given information at the top is clear enough. Thank you.

[dersia]

long story short, it would be nice to have an operator for this, but I don't feel strongly about it. if maintainers think we don't need it, it's fine with me, there are more important features that I am looking forward to like extions, unions, expression bodies and HKT (I know it's unlikely that we get HKT, but I am still hoping)

[suugbut]

This "answer" is made only for ones supporting the proposal.
If you intent is for rejection, please react to the existing posts or make new ones.

Vote	x is null ? y : x	x is null ? x : y	by
👍	x ?| y	x |? y	suugbut
👎	x ?? y	x ¿¿ y	suugbut
❤️	x ?< y or x <? y	x ?> y or x >? y	suugbut
🎉	x ?? y	x !? y	andrey-messi (*)
🚀	x ?? y	x ⸮⸮ y	huoyaoyuan (*)
👀
🚀
😕
😂

(*) : Feel free to ask for attribution removal.

[HaloFour]

None of the above, this operation isn't remotely common enough to justify a special operator.

[Daynvheur]

Maybe reconsider the usage of inverted characters, as I don't (and don't plan to) have access to inverted characters,
I comment here to maybe narrow your proposal list.

[suugbut]

Yes. Both the inverted and mirrored ones are not in 7-bit ASCII characters. I keep putting them in the table only for the sake of documentation. In the future other people who come with the same idea should not propose them again.