Why does this happen?
An answer that cannot be argued with is "because the spec says so":
A protected
member of a base class is accessible in a derived class
only if the access occurs through the derived class type.
But let's explore this restriction behind the scenes.
Explanation
What happens here is the same thing that Eric Lippert describes in the blog post that you linked to. Your code does the equivalent of this:
public abstract class MenuItem
{
protected string m_Title;
}
public class ContainerItem : MenuItem
{
void Foo()
{
var derivedItem = new ContainerItem();
derivedItem.m_Title = "test"; // works fine
var baseItem = (MenuItem)derived;
baseItem.m_Title = "test"; // compiler error!
}
}
The problem here stems from the fact that this might happen. For the moment, please disregard the fact that this example uses a method instead of a field -- we 'll come back to it.
public abstract class MenuItem
{
protected void Foo() {}
}
public class SomeTypeOfItem : MenuItem
{
protected override void Foo() {}
}
public class ContainerItem : MenuItem
{
void Bar()
{
var baseItem = (MenuItem)something;
baseItem.Foo(); // #1
}
}
Look at line #1: how does the compiler know that baseItem
is not actually a SomeTypeOfItem
? If it is, you certainly must not be able to access Foo
! So, as Eric describes, the compiler is unable to statically prove that the access is always legal and because of that it has to disallow this code.
Note that in some cases, for example if
baseItem = (MenuItem)new ContainerItem();
or even
baseItem = (MenuItem)this;
the compiler does have enough information to prove that the access is legal but it still will not allow the code to compile. I imagine that's because the compiler team is not convinced that implementing such special-case handlers is worth the trouble (a point of view which I am sympathetic to).
But... but...
That's all well and good for methods (and properties, which are really methods) -- what about fields? What about this:
public abstract class MenuItem
{
protected string m_Title;
}
public class SomeTypeOfItem : MenuItem
{
protected new string m_Title;
}
public class ContainerItem : MenuItem
{
void Foo()
{
var baseItem = (MenuItem)something;
baseItem.m_Title = "Should I be allowed to change this?"; // #1
}
}
Since fields cannot be overridden, there should be no ambiguity here and the code should compile and set MenuItem.m_Title
irrespective of what the type of something
is.
Indeed, I cannot think of a technical reason why the compiler couldn't do this, but there is a good reason in any case: consistency. Eric himself would probably be able to provide a richer explanation.
So what can I do?
How would you access the protected members like m_Title while holding
a reference to MenuItem (because of Polymorphism design reasons)?
You simply cannot do that; you would have to make the members internal
(or public
).