Should NotNullWhen be working when assigning result to a variable?

Question

I have the following method which takes a Type as a parameter and checks if it is IEnumerable<T>. If it is, it returns true and sets the type of T in an out variable, otherwise it returns false.

private static bool TryGetEnumeratedType(Type type, [NotNullWhen(true)] out Type? enumeratedType)
{
    enumeratedType = null;
    bool isEnumerable = type.GetInterfaces().Any(i => i.GetTypeInfo() == typeof(IEnumerable));

    if (!isEnumerable)
    {
        return false;
    }

    Type? genericType = type.GetGenericArguments().FirstOrDefault();

    if (genericType is null)
    {
        return false;
    }

    enumeratedType = genericType;

    return true;
}

If I call the method this way, there is no issue:

Type fieldType;

if (TryGetEnumeratedType(inType, out Type? type))
{
    fieldType = type;
}
else
{
    fieldType = inType;
}

But if I assign the result to a variable and check against the variable, I get a compiler warning when assigning fieldType = type saying that type may be null.

Type fieldType;
bool isEnumerable = TryGetEnumeratedType(inType, out Type? type);

if (isEnumerable)
{
    fieldType = type;
}
else
{
    fieldType = inType;
}

Is this a bug, or am I missing something?

defaultUsernameN · Accepted Answer · 2022-03-11 23:52:54Z

2

I believe that has nothing to do with NotNullWhen. The compiler just doesn't do that kind of work - it doesn't go into deep checks of "where did that variable come from". Doing that would probably be 1) hard to implement and 2) you wouldn't know where to stop: how high up the call stack should compiler go to ensure there is no null assigned to your value? That would probably be very resource-intensive. This answer discusses it a bit.

Consider these two methods, for example:

public int GetInt() //no warning
{
    if (true)
        return 1; 
}

public int GetAnotherInt() //CS0161: not all code paths return a value
{
    var @true = true;
    if (@true)
        return 1; 
}

edited Mar 11, 2022 at 23:52

answered Mar 11, 2022 at 23:20

defaultUsernameN

8211 gold badge11 silver badges26 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Tore Aurstad · Accepted Answer · 2024-06-26 16:07:30Z

I tested out your code myself and in .NET 8.0 it still is an issue. It seems that the validation of these static analysis attributes such as [NotNullWhen] needs to be fixed for method parameters.

A search for these static analysis attributes reveals still a multitude of issues and it looks like they do not handle every case. This issue reported in is similar to the one you show here: https://github.com/dotnet/roslyn-analyzers/issues/6990

After all, the Roslyn team tries to create a mechanism for automated null checks, but there are many edge cases and issues takes years to resolve by looking through when open issues on this topic was reported.

For me, it seems that the fix for you for now is to use the null forgiving operator inside the if clause on the RHS.

if (isEnumerable)
{
    fieldType = type!;
}

Then you can head over to Github Issues page for Roslyn here and file an issue : https://github.com/dotnet/roslyn-analyzers/issues

I tried some of the other attributes such as [MaybeNullWhen] attribute looking at the documentation, but could not find a working alternative.

https://learn.microsoft.com/en-us/dotnet/csharp/language-reference/attributes/nullable-analysis#conditional-post-conditions-notnullwhen-maybenullwhen-and-notnullifnotnull

Collectives™ on Stack Overflow

Should NotNullWhen be working when assigning result to a variable?

2 Answers 2

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related