EDIT: There's a fourth option, i.e. to avoid all this nonsense and do what Jon Skeet suggests in his answer.
Something like this?
public static EventHandler ToEventHandler(this ClickMenuItem clickMenuItem)
{
if (clickMenuItem == null)
return null;
return (sender, e) => clickMenuItem(sender, e);
}
and the reverse:
public static ClickMenuItem ToClickMenuItem(this EventHandler eventHandler)
{
if (eventHandler == null)
return null;
return (sender, e) => eventHandler(sender, e);
}
Note that the compiler infers which delegate-types to convert the lamda-expressions to.
EDIT: If you prefer, you can use anonymous-delegates too.
EventHandler eventHandler = delegate(object sender, EventArgs e)
{
clickMenuItem(sender, e);
};
return eventHandler; // can be inlined, type-inference works fine
The third alternative of course, is to write a closure-class yourself. I wouldn't really recommend this, but it gives you an idea of what the compiler does with the anonymous methods. Something like:
public static class ClickMenuItemExtensions
{
public static EventHandler ToEventHandler(this ClickMenuItem clickMenuItem)
{
if (clickMenuItem == null)
return null;
// new EventHandler not required, included only for clarity
return new EventHandler(new Closure(clickMenuItem).Invoke);
}
private sealed class Closure
{
private readonly ClickMenuItem _clickMenuItem;
public Closure(ClickMenuItem clickMenuItem)
{
_clickMenuItem = clickMenuItem;
}
public void Invoke(object sender, EventArgs e)
{
_clickMenuItem(sender, e);
}
}
}