Consider the following sample-code:

class TestTypeInference
{
  Str[] fieldStrs := [,]

  Void main()
  {
    Str[] localStrs := [,]
    echo(fieldStrs.typeof)
    echo(localStrs.typeof)
  }
}

When I run it, I get the following result:

sys::Str[]
sys::Obj?[]

The type-inference does not work the same-way with fields and with local-vars. Is there a reason for this or is this a bug?

It is by design. We made a special exception for fields in ticket 986 so that all fields have an explicit type signature for readability. Everything else uses normal type inference rules.

Ok, I got it. Is there a reason why we don't apply the same special treatment for local-vars? I know I can use type-inference with local-vars but it would be more consistent if it would behave the same way.

Ok, I got it. Is there a reason why we don't apply the same special treatment for local-vars?

You don't need it in local variables because they support type inference. So I think it is better to have just have one way:

x := Str[,]     // the one-way  
Str[] x := [,]  // if we supported both ways

Really it is a question of trade-offs: which is better to have one way versus two ways to use type inference in local variables, or to be consistent with fields?