Allow var in fixed statement

[sweemer]

Currently the pointer type used in a fixed statement must be explicitly declared, unlike normal variables, which may be declared using var.

I believe that it would be less redundant, and more consistent with the rest of the language, if var were allowed in fixed statements as well, such that the following code would become legal:

var s = "abc";
unsafe
{
    fixed (var* c = s)
    {
        // c has type char* in this block
    }
} 

I haven’t found any discussions or proposals for this yet. Can it be considered?

[huoyaoyuan]

The language doesn't allow multiple choice of fixed of a given type:

    unsafe
    {
        fixed (byte* ptr = c) // error, ambiguous
        {

        }
    }

    public class Foo
    {
    }

    public static class FooExtension1
    {
        public static ref byte GetPinnableReference(this Foo foo) => throw null;
    }

    public static class FooExtension2
    {
        public static ref int GetPinnableReference(this Foo foo) => throw null;
    }

This should be actionable.

[sweemer]

Hi @huoyaoyuan thanks for the reply but I'm a bit confused. My proposal isn't to allow multiple choice of type for fixed statements, but rather to allow inference of a single type using var. The language should allow that, no?

[huoyaoyuan]

If there are multiple applicable candidates decided by target type, using var can cause ambiguity.

[sweemer]

OK I see what you mean now, but is that enough of a reason to disallow var in all cases? Even with normal variables there are cases where ambiguous var results in a compiler error, such as declaring var x; without an assignment on the same line. To me it seems that ambiguous var in a fixed statement could also result in a compiler error while unambiguous var would not.

[huoyaoyuan]

It's not a veto that disallows the feature, but it can significantly impact the effort involved and also the interest towards the feature.