I have a library here with some Java classes. One class has some protected static methods, which I realize is sorta an OOP no-no but I can't change its code. Assuming I have a Scala class that subclasses the aforementioned Java class, how can I call its protected static members?
See Frequently Asked Questions - Java Interoperability:
This is a known limitation of Scala:
there is no notion of 'static' members
in Scala. Instead, Scala treats
static members of class Y as members
of the singleton object Y (the
companion object of class Y). When
inheriting from this class, one can
access only protected members of class
Y but cannot access protected members
of object Y.
There's no way Scala can simulate
static protected without impairing the
integrity of Scala's object model in a
fundamental way, so this is not going
to change. To work around this
limitation, one has to create an
implementation of the enclosing class
with Java code which encapsulates all
accesses to the protected static inner
class.
See ticket #1806 for more
information and a concrete example of
the limitation and its workaround.
Related
The question may be a bit confusing, and is best illustrated by an example:
unit Test
interface
type
TestClass = class()
Splitter1: TcxSplitter;
procedure SomeMethod();
end;
implementation
uses
cxSplitter;
// Locally-declared child type
type
TcxSplitterAccess = class(TcxSplitter);
procedure TestClass.SomeMethod()
var
pos: integer;
begin
// Access to protected field FPositionBeforeClose by casting
pos := TcxSplitterAccess(Splitter1).FPositionBeforeClose;
end;
Notice in the implementation section that there is a type TcxSplitterAccess being declared as a child of the TcxSplitter class. In the method SomeMethod(), belonging to the class TestClass, a TcxSplitter object is cast to the locally-declared TcxSplitterAccess class, and then a protected field is accessed on that object.
This is surprising to me as someone coming from a background of languages like Java, C++, C#, etc. In those languages, it is possible to access protected data in an object so long as you are doing it from within that object's type or an inherited type. For example, a method inside of a class ClazzA can access the private fields of other ClazzA objects since access is enforced at the type level rather than the instance level. Declaring a class locally in these languages would not give the containing class access to the local class' protected data (edit: As pointed out in the comments, this is actually not true at least for Java).
In this example, however, the type TestClass is directly accessing a protected field on the TcxSplitter object by first casting to the TcxSplitterAccess type. I am having trouble finding documentation on why this "trick" works. Does Delphi handle access levels fundamentally differently to Java-like languages and allows this sort of thing? Regardless, why does this trick work?
While I stumbled onto this behavior by using a nested, inherited class in order to access fields on the parent class (which breaks encapsulation, and I shouldn't do), the use of inheritance here is unnecessary. If the nested class did not inherit from a class, but instead had its own protected fields defined, TestClass would still be able to access those protected fields.
A unit has implicit-friendship semantics within itself. Types declared in the same unit are "friends" of each other (similar to friend in C++), and so can access each other's private and protected members (but not strict private or strict protected members).
So, in this case:
TcxSplitterAccess derives from TcxSplitter, so TcxSplitterAccess inherits all of the protected (but not private) members of TcxSplitter via normal class inheritance.
TestClass is declared in the same unit as TcxSplitterAccess, so TestClass has access to all of the protected members of TcxSplitterAccess, including the protected members of TcxSplitter (and the protected members of its ancestors).
FPositionBeforeClose is protected in TcxSplitter.
So, that is why TestClass.SomeMethod() is able to access FPositionBeforeClose when type-casting the Splitter1 object to TcxSplitterAccess.
This is covered by the Delphi documentation:
Private, Protected, Public, and Published Declarations
Classes and Objects (Delphi): Visibility of Class Members
I'm using the Mixin pattern as illustrated below. Why does Typescript require you to provide stand-in properties for private properties of the mixin class in the target class (A)? It perfectly makes sense for the public properties, but for private properties it unnecessarily crufts up the target class with details of the internal implementation of the mixin class by requiring them to be stubbed-out in the target class. Seems like the Typescript transpiler should be able to not require this.
class Mixin {
private foo:string;
}
class A implements Mixin {
// stand-in properties, typescript requires even
// private properties to be stubbed-out
foo:string;
}
Private members contribute to the structure of a type in TypeScript, so if you don't implement them you are not compatible with the type. This actually makes it impossible to match a type structurally in TypeScript if it has a private member, because you are either:
a. Failing to provide the type
or
b. Providing a separate implementation of the private member
So you can only extend a type with a private member, not implement it.
With this in mind, you are better off not using private members with mixins. Provide the ghost-members in the implementation class and keep your fingers crossed that if mixins gain some traction the ghosting will become unnecessary (see TypeScript mixins part one).
Suppose MyClass is a class defined in Java, and has many static as well as non-static members. I tried to alias this class (and associated companion object) in a Scala object MyObject as shown below:
object MyObject {
import javastuff._
type MyAlias = MyClass
val MyAlias = MyClass
}
Scalac complains:
error: object MyClass is not a value
val MyAlias = MyClass
How do I work around this? Thanks.
Although this works in pure Scala for a class + companion object, it's not possible with Java's static methods, as these don't belong to any interface.
Scala could, in theory, create an object containing delegates to all the static methods of some class, but it doesn't do this currently. It's also possible to write a compiler plugin for this if you feel comfortable writing plugins.
Failing that, you'll either have to create an object full of delegates yourself, or just cherry-pick a few methods and pass them around as functions.
it's not possible with Java's static methods, as these don't belong to any interface.
Update 5 years later: PR 5131 mentions:
We used to disable generation of static forwarders when a object had a
trait as a companion, as one could not add methods with bodies to an
interface in JVM 6.
The JVM lifted this restriction to support default methods in interfaces,
so we can lift the restriction on static forwarders, too.
Fixes scala-dev issue 59
See commit 41c9a17 by Jason Zaugg (retronym).
In Programming in Scala: A Comprehensive Step-by-Step Guide, the author said:
One way in which Scala is more
object-oriented than Java is that
classes in Scala cannot have static
members. Instead, Scala has singleton
objects.
Why is a singleton object more object-oriented? What's the good of not using static members, but singleton objects?
Trying for the "big picture"; most of this has been covered in other answers, but there doesn't seem to be a single comprehensive reply that puts it all together and joins the dots. So here goes...
Static methods on a class are not methods on an object, this means that:
Static members can't be inherited from a parent class/trait
Static members can't be used to implement an interface
The static members of a class can't be passed as an argument to some function
(and because of the above points...)
Static members can't be overridden
Static members can't be polymorphic
The whole point of objects is that they can inherit from parent objects, implement interfaces, and be passed as arguments - static members have none of these properties, so they aren't truly object-oriented, they're little more than a namespace.
Singleton objects, on the other hand, are fully-fledged members of the object community.
Another very useful property of singletons is that they can easily be changed at some later point in time to not be singletons, this is a particularly painful refactoring if you start from static methods.
Imagine you designed a program for printing addresses and represented interactions with the printer via static methods on some class, then later you want to be able to add a second printer and allow the user to chose which one they'll use... It wouldn't be a fun experience!
Singleton objects behave like classes in that they can extend/implement other types.
Can't do that in Java with just static classes -- it's pretty sugar over the Java singleton pattern with a getInstance that allows (at least) nicer namespaces/stable identifiers and hides the distinction.
Hint: it's called object-oriented programming.
Seriously.
Maybe I am missing something fundamentally important, but I don't see what the fuss is all about: objects are more object-oriented than non-objects because they are objects. Does that really need an explanation?
Note: Although it sure sounds that way, I am really not trying to sound smug here. I have looked at all the other answers and I found them terribly confusing. To me, it's kind of obvious that objects and methods are more object-oriented than namespaces and procedures (which is what static "methods" really are) by the very definition of "object-oriented".
An alternative to having singleton objects would be to make classes themselves objects, as e.g. Ruby, Python, Smalltalk, Newspeak do.
For static members, there is no object. The class really just is a namespace.
In a singleton, there is always at least one object.
In all honesty, it's splitting hairs.
It's more object oriented in the sense that given a Scala class, every method call is a method call on that object. In Java, the static methods don't interact with the object state.
In fact, given an object a of a class A with the static method m(), it's considered bad practice to call a.m(). Instead it's recommended to call A.m() (I believe Eclipse will give you a warning). Java static methods can't be overridden, they can just be hidden by another method:
class A {
public static void m() {
System.out.println("m from A");
}
}
public class B extends A {
public static void m() {
System.out.println("m from B");
}
public static void main(String[] args) {
A a = new B();
a.m();
}
}
What will a.m() print?
In Scala, you would stick the static methods in companion objects A and B and the intent would be clearer as you would refer explicitly to the companion A or B.
Adding the same example in Scala:
class A
object A {
def m() = println("m from A")
}
class B extends A
object B {
def m() = println("m from B")
def main(args: Array[String]) {
val a = new B
A.m() // cannot call a.m()
}
}
There is some difference that may be important in some scenarios. In Java you
can't override static method so if you had class with static methods you would not be able to customize and override part of its behavior. If you used singleton object, you could just plug singleton created from subclass.
It's a marketing thing, really. Consider two examples:
class foo
static const int bar = 42;
end class
class superfoo
Integer bar = ConstInteger.new(42);
end class
Now, what are the observable differences here?
in a well-behaved language, the additional storage created is the same.
Foo.bar and Superfoo.bar have exactly the same signatures, access, and so on.
Superfoo.bar may be allocated differently but that's an implementation detail
It reminds me of the religious wars 20 years ago over whether C++ or Java were "really" Object Oriented, since after all both exposed primitive types that aren't "really" objects -- so, for example you can't inherit from int but can from Integer.
I've written a Scala trait, named Cache[A,B], to provide a caching API. The Cache has the following methods, asyncGet(), asyncPut(), asyncPutIfAbsent(), asyncRemove().
I'm going to have a few static methods, such as getOrElseUpdate(key: A)(op: => B). I don't want methods like this as abstract defs in the Cache trait because I don't want each Cache implementation to have to provide an implementation for it, when it can be written once using the async*() methods.
In looking at Google Guava and parts of the Java library, they place public static functions in a class that is the plural of the interface name, so "Caches" would be the name I would use.
I like this naming scheme actually, even though I could use a Cache companion object. In looking at much of my code, many of my companion objects contain private val's or def's, so users of my API then need to look through the companion object to see what they can use from there, or anything for that matter.
By having a object named "Caches" is consistent with Java and also makes it clear that there's only public functions in there. I'm leaning towards using "object Caches" instead of "object Cache".
So what do people think?
Scala's traits are not just a different name for Java's interfaces. They may have concrete (implemented) members, both values (val and var) and methods. So if there's a unified / generalized / shared implementation of a method, it can be placed in a trait and need not be replicated or factored into a separate class.
I think the mistake starts with "going to have a few static methods". Why have static methods? If you explain why you need static methods, it will help figure out what the design should be.