Just another one CS0165: Use of unassigned variable

[gurustron]

Was not able to find exactly the same one. Consider the following method:

    bool aBool = SomePredicate();
    if (!aBool && !OutFunction(out int value))
    {
        return;
    }
    int result = -1; //other code here...
    if (!aBool && SomeCondition(result, value)) // CS0165: Use of unassigned variable 'value'.
    {
    }

    bool SomePredicate() => false;
    bool OutFunction(out int i) => false;
    bool SomeCondition(int i, int j) => false;

While due to the same check !aBool and short-circuiting at point of SomeCondition(result, value) logically value is definitely assigned the compiler fails to determine that.

Sharplab

[bernd5]

if aBool is true OutFunction is not executed - value is not assigned.

In the past there were already some improvements of the Definite Assignment Analysis.

But I think it is not only hard for the compiler to see this - for humans it is hard, too.

I would refactor the code to:

using System;
public class C {
    public void M() {
        if (!SomePredicate())
        {
            if (!OutFunction(out int value))
            {
                return;
            }
            int result = -1; //other code here...
            if (SomeCondition(result, value)) // CS0165: Use of unassigned variable 'value'.
            {

            }
        }
        
        bool SomePredicate() => false;
        bool OutFunction(out int i) => throw new Exception();

        bool SomeCondition(int i, int j) => false;
    }
}

[gurustron]

@bernd5 that is what I wrote in this SO answer =) I was wondering if there are some additional insights here (hence like some specs or links where compiler developers explain why it expected here).

[bernd5]

See the rules: https://github.com/dotnet/csharpstandard/blob/draft-v7/standard/variables.md#94-definite-assignment

[bernd5]

See especially: https://github.com/dotnet/csharpstandard/blob/draft-v7/standard/variables.md#94426--expressions

[RikkiGibson]

dotnet/roslyn#36927 (comment)

[gurustron]

@RikkiGibson and @bernd5 thank you!