I have generic class :
public class Test<T> {
private Test<? extends T> myInstance;
public Test<? extends T> getInstance () {
return myInstance;
}
public void setInstance (Test<? extends T> argType) {
this.myInstance = argType;
}
}
And I have two classes in my class hierarchy relations:
public abstract class Alphabet {
//code here
}
and
public class A extends Alphabet{
public A() {
super();
System.out.print("This is A call");
}
}
Finally I have a class where I want to store make generic class Test with particular type and set new Instance of Object -> A through setInstance() method:
public static void main(String[] args) {
List<Alphabet> list = new ArrayList<Alphabet>();
Test<Alphabet> tAlphabet = new Test<Alphabet>();
tAlphabet.setInstance(new A()); //Here is compilation ERROR
}
But I have got the compilation error in line tAlphabet.setInstance(new A());
What is the issue with my generic class?
Your instance is a Test object as it's currently written, and you are supplying it with an Alphabet object instead. You probably want your instance to be of type Alphabet:
public class Test<T> {
private T myInstance;
public T getInstance() {
return myInstance;
}
public void setInstance(T argType) {
myInstance = argType;
}
}
This way, your Test stores an Alphabet instead of another Test.
It seems you have made things more complicated than needed. You probably want this in your Test class instead of what you actually have:
private T myInstance;
public T getInstance () {
return myInstance;
}
public void setInstance (T argType) {
this.myInstance = argType;
}
With this arrangement you would be free to setInstance(new A()) on a Test<Alphabet> instance.
Related
I want to have a class to run other classes in java, like constructor parameterized with a class to run that class later on, similar to this
class MyClass{
Class classToRun;
public MyClass(Class c) {
super();
this.classToRun = c;
}
public void runClass(){
classToRun.someStaticMethod();
}
}
where classToRun possible classes doesn't have a common ancestor, but all have method someStaticMethod, and have no idea about MyClass, which runs them.
But there are problems, like inner classes cannot have static methods, classes cannot be cast Class, etc.
There are solutions for parameterized with class methods, like
How do I pass a class as a parameter in Java?
Passing a class as an argument to a method in java
but not for constructors.
What is the proper solution to do this?
Use lambdas and pass the method reference: they match on the method signature. For void someStaticMethod() you can use Runnable.
class MyClass{
private final Runnable methodToRun;
public MyClass(Runnable someStaticMethod) {
methodToRun = someStaticMethod;
}
public void runClass(){
methodToRun.run();
}
}
new MyClass(SomeClass::someStaticMethod).runClass();
You cannot enforce that the method passed has the right name, but looks even neater IMHO.
You need to understand what generics are.
interface
public interface SomeInterface {
void someStaticMethod();
}
use
class MyClass<T extends SomeInterface>{
T classToRun;
public MyClass(T c) {
super();
this.classToRun = c;
}
public void runClass(){
classToRun.someStaticMethod();
}
}
As 2 of 3 answers were not to the point, I decided to publish fixed versions of both answers as far as they can be fixed.
The f1sh version from the above should like follows:
public class ClassToRunOthers {
Class classToRun;
public ClassToRunOthers(Class c) {
this.classToRun = c;
}
public void runClass() throws Exception {
Optional<Method> method = Arrays.stream(classToRun.getDeclaredMethods()).filter(m -> m.getName().equals("someStaticMethod")).findFirst();
if(!method.isPresent()) {
throw new RuntimeException();
}
method.get().invoke(null);
}
public static void main(String[] args) throws Exception {
ClassToRunOthers mc = new ClassToRunOthers(SomeClass.class);
mc.runClass();
}
}
class SomeClass {
static void someStaticMethod() {
System.out.println("test");
}
}
The zwei solution above can not be fixed without reflection, as generics is not to the point. Evan if you try to parametrize not with SomeInerface (because SomeClass does not extend a common SomeInterface), but with Object, it is still won't solve the problem:
public class MyClass<T extends Object> {
T classToRun;
public MyClass(T c) {
super();
this.classToRun = c;
}
public void runClass() {
// classToRun.someStaticMethod(); // Cannot resolve method 'someStaticMethod' in 'T'
}
public static void main(String[] args) {
MyClass mc = new MyClass(SomeClass.class);
}
}
class SomeClass {
static void someStaticMethod() {
System.out.println("test");
}
}
This can be fixed like the above, via reflection.
I believe, it can be done with annotations in some elegant way, and may be someone will share us with such a solution or I will do it by myself as time permits.
By now for myself, a solution with saving class name in the String in constructor next day after the question been asked did the trick.
You will have to use reflection if you want to execute a method when you only have the Class instance.
In the code below, runClass finds the method of the class using it's name as a String, then executes it. This code assumes that the method is static, also ignoring any Exception handling.
The following code prints "test":
class MyClass {
Class classToRun;
public MyClass(Class c) {
this.classToRun = c;
}
public void runClass() throws Exception {
Optional<Method> method = Arrays.stream(classToRun.getDeclaredMethods()).filter(m -> m.getName().equals("someStaticMethod")).findFirst();
if(!method.isPresent()) {
throw new RuntimeException();
}
method.get().invoke(null);
}
}
class Main {
public static void main(String[] args) throws Exception {
MyClass mc = new MyClass(Main.class);
mc.runClass();
}
static void someStaticMethod() {
System.out.println("test");
}
}
I get this compile error "required T, found UseCaseTest" when I try to return INSTANCE
at the method keepStateBetweenTests of the class UseCaseTest. I know I can easily fix it casting to T, but
I want to do it right.
Plus, I don't know why the line
if (INSTANCE == null) INSTANCE = getInstance(activityTestRule);
can compile, given the previous error.
I supply two more classes to provide certain context.
UseCaseTest
public abstract class UseCaseTest {
private static UseCaseTest INSTANCE;
public abstract static class Builder<T extends UseCaseTest> {
private final ActivityTestRule activityTestRule;
protected Builder(ActivityTestRule activityTestRule) {
this.activityTestRule = activityTestRule;
}
public T keepStateBetweenTests() {
if (INSTANCE == null) INSTANCE = getInstance(activityTestRule);
return INSTANCE; //compile error: required T, found UseCaseTest
}
public T releaseStateBetweenTests() {
return getInstance(activityTestRule);
}
protected abstract T getInstance(ActivityTestRule activityTestRule);
}
private final ActivityTestRule activityTestRule;
public UseCaseTest(ActivityTestRule activityTestRule) {
this.activityTestRule = activityTestRule;
}
}
SessionUseCaseTest
public final class SessionUseCaseTest extends UseCaseTest {
public static Builder<SessionUseCaseTest> with(ActivityTestRule activityTestRule) {
return new Builder<SessionUseCaseTest> (activityTestRule) {
#Override protected SessionUseCaseTest getInstance(ActivityTestRule activityTestRule) {
return new SessionUseCaseTest(activityTestRule);
}
};
}
private SessionUseCaseTest(ActivityTestRule activityTestRule) {
super(activityTestRule);
}
public void signUp() {}
public void logout() {}
}
SessionTest
#FixMethodOrder(MethodSorters.NAME_ASCENDING)
public class SessionTest extends BaseTest {
#Test public void _1_SingUp() {
SessionUseCaseTest.with(mActivityRule)
.keepStateBetweenTests()
.signUp();
}
#Test public void _2_Logout() {
SessionUseCaseTest.with(mActivityRule)
.keepStateBetweenTests()
.logout();
}
}
The topic of static method/fields in conjunction with generics is very well described in this post
Nevertheless, it would seem a simple casting makes the compiler happy while allowing to keep some level of generics in place:
public T keepStateBetweenTests() {
if (INSTANCE == null)
INSTANCE = getInstance(activityTestRule);
return (T)INSTANCE;
}
UPDATE
Part of your question was why this can compile although there is the error:
if (INSTANCE == null) INSTANCE = getInstance(activityTestRule);
To my understanding, the compiler knows the INSTANCE itself is actually UseCaseTest, which fulfills the restriction defined on the Builder class: <T extends UseCaseTest>
You need to make your class aware of type T i.e.
public abstract class UseCaseTest <T> {
You can do like below:-
class UseCaseTest<T> {
final Class<T> useCaseTestType;
public UseCaseTest(Class<T> useCaseTestType) {
this.useCaseTestType = useCaseTestType;
}
//Then write your methods which are generic
}
Say you have a super-class. In that super class you want to pass runtime object of itself (this) as a parameter to an overloaded method. Trick is, this overloaded method is overloaded by sub-class type. When you try to do it, you'll get a
method ... is not applicable(actual argument
... cannot be converted to ... by method invocation
conversion)
Instead you would need to implement the method separately in each subtype (just to get the correct runtime class), which is a lot of duplicate effort when the contents of the method are identical.
e.g:
public class InferTypeTest {
public static void main(String[] args) {
SubClass1 s1 = new SubClass1();
s1.sayHi();
}
public static void sayHi(SubClass1 clz) {
System.out.println("clz 1");
}
private abstract static class SuperClass{
public void sayHi() {
InferTypeTest.sayHi(this);
}
}
private static class SubClass1 extends SuperClass{
}
}
Yes, this is how double dispatch works, you have to override the accept method in each subclass like this:
private static abstract class NodeWithChildren implements DomNode {
/* snip */
public void accept(DomNodeVisitor visitor) {
for (DomNode child : children) {
child.accept(visitor);
}
}
}
private static class BodyNode extends NodeWithChildren {
public void accept(DomNodeVisitor visitor) {
visitor.visit(this);
super.accept(visitor);
visitor.visited(this);
}
}
private static class DivNode extends NodeWithChildren {
public void accept(DomNodeVisitor visitor) {
visitor.visit(this);
super.accept(visitor);
visitor.visited(this);
}
}
}
BodyNode is a NodeWithChildren
DivNode is a NodeWithChidren
NodeWithChildren is a DomNode
DomPrinterVisitor is a DomNodeVisitor
DomNodeVisitor's visitor can visit "BodyNode" or "DivNode, But you are passing "NodeWithChildren" to visit.
Here BodyNode is a NodeWithChildren but NodeWithChildren is not BodyNode/
Theary is if B extends A, you can say B is a A/ not A is a B.
Why does this java code produce StackOverflowError? I understand that this somehow connected with recursive generic type parameter. But I don't understand clear the whole mechanism.
public class SomeClass<T extends SomeClass> {
SomeClass() {
new SomeClassKiller();
}
private class SomeClassKiller extends SomeClass<T> {
}
public static void main(String[] args) {
new SomeClass();
}
}
The generic part doesn't matter - nor does it really matter that the class is nested. Look at this mostly-equivalent pair of classes and it should be more obvious:
public class SuperClass
{
public SuperClass()
{
new SubClass();
}
}
public class SubClass extends SuperClass
{
public SubClass()
{
super();
}
}
So the subclass constructor calls the superclass constructor - which then creates a new subclass, which calls into the superclass constructor, which creates a new subclass, etc... bang!
Here it is invoking one constructor from another and from it the previous one, cyclic constructor chain, see the comments below
public class SomeClass<T extends SomeClass> {
SomeClass() {//A
new SomeClassKiller();// calls B
}
private class SomeClassKiller extends SomeClass<T> {//B
//calls A
}
public static void main(String[] args) {
new SomeClass(); //calls A
}
}
This is because of the Recursive constructor calls happening between the classes SomeClass and
SomeClassKiller.
public class SomeClass<T extends SomeClass> {
SomeClass() {
new SomeClassKiller();
}
private class SomeClassKiller extends SomeClass<T> {
public SomeClassKiller()
{
super(); //calls the constructor of SomeClass
}
}
public static void main(String[] args) {
new SomeClass();
}
}
The code produced by the compiler is something like this, so when u create an object it recursivly calls the SomeClass and SomeClassKiller for ever.
Constructors are invoked top-to-bottom, that is if a class A derives from B, A's constructors will first invoke the parent constructor (B).
In you case, new SomeClassKiller() recursively calls the constructor of SomeClass which in turn constructs another SomeClassKiller … there it is.
The main() method is creating a new instance of SomeClass which calls the SomeClass constructor that creates a new instance of SomeClassKiller that by default calls the parent constructor and the stackoverflow occurs.
To avoid the stackoverflow. Change the code to look as follows:
public class SomeClass<T extends SomeClass> {
SomeClass() {
new SomeClassKiller();
}
private class SomeClassKiller extends SomeClass<T> {
public SomeClassKiller(){
//super(); does this by default, but is now commented out and won't be called.
}
}
public static void main(String[] args) {
new SomeClass();
}
}
I wanted to create an interface for copying an object to a destination object of the same class. The simple way is to use casting:
import org.junit.Test;
import org.junit.internal.runners.JUnit4ClassRunner;
import org.junit.runner.RunWith;
#RunWith(JUnit4ClassRunner.class)
public class TestGenerics {
public static interface Copyable {
public void copy(Copyable c);
}
public static class A implements Copyable {
private String aField = "--A--";
protected void innerCopy(Copyable c) {
A a = (A)c;
System.out.println(a.aField);
}
public void copy(Copyable c) {
innerCopy(c);
}
}
public static class B extends A {
private String bField = "--B--";
protected void innerCopy(Copyable c) {
B b = (B)c;
super.innerCopy(b);
System.out.println(b.bField);
}
}
#Test
public void testCopy() {
Copyable b1 = new B();
Copyable b2 = new B();
b1.copy(b2);
}
}
But also i've found a way it can be done using generics:
import org.junit.Test;
import org.junit.internal.runners.JUnit4ClassRunner;
import org.junit.runner.RunWith;
#RunWith(JUnit4ClassRunner.class)
public class TestGenerics {
public static interface Copyable<T> {
public void copy(T t);
}
public static class A<T extends A<?>> implements Copyable<T> {
private String a = "--A--";
public void copy(T t) {
System.out.println(t.a);
}
}
public static class B<T extends B<?>> extends A<T> {
private String b = "--B--";
public void copy(T t) {
super.copy(t);
System.out.println(t.b);
}
}
#Test
#SuppressWarnings("unchecked")
public void testCopy() {
Copyable b1 = new B();
Copyable b2 = new B();
b1.copy(b2);
}
}
Though the only way i've found to get rid of warnings is the annotation. And it feels like something is wrong.
So what's wrong? I can accept that something is wrong in the root of the problem. So any sort of clarification is welcome.
Your interface definition:
public interface Copyable<T extends Copyable<T>> {
void copy(T copyFrom);
}
Your implementation:
public class Example implements Copyable<Example> {
private Object data;
void copy(Example copyFrom) {
data = copyFrom.data;
}
//nontrivial stuff
}
That should take care of your warnings.
Assuming you don't want to subclass further you just need:
public static /*final*/ class AClass implements Copyable<AClass> {
For an abstract class, you do the "enum" thing:
public static abstract class AClass<T extends AClass<T>> implements Copyable<T> {
In testCopy, one of the warnings is because you're instantiating a "raw type" of Copyable rather than some concrete Copyable<T>. Once you instantiate a Copyable, it can only be applied to Ts (which include subtypes of T). In order to instantiate with a formal type, the class definitions will need to be changed slightly:
public static class A<T extends A> implements Copyable<T>
public static class B<T extends B> extends A<T>
The next issue is that a Copyable<B> can only be passed a compile-time type of B (based on the definition of Copyable). And testCopy() above is passing it a compile-time type of Copyable. Below are some examples of what will work, with brief descriptions:
public void testExamples()
{
// implementation of A that applies to A and subtypes
Copyable<A> aCopier = new A<A>();
// implementation of B that applies to B and subtypes
Copyable<B> bCopier = new B<B>();
// implementation of A that applies to B and subtypes
Copyable<B> bCopier2 = new A<B>();
}
I keep trying to figure out a way to get rid of the warnings in your first approach and I can't come up with anything that works. Even so, I think the first approach is the lesser of two evils. An unsafe cast is better than needing to give your classes such a complicated api.
A completely separate approach would be to override Object.clone() and implement Cloneable.
This is the best possible code of second approach. It compiles without any warnings.
import static org.junit.Assert.fail;
import org.junit.Test;
import org.junit.internal.runners.JUnit4ClassRunner;
import org.junit.runner.RunWith;
#RunWith(JUnit4ClassRunner.class)
public class TestGenerics {
public static interface Copyable<T> {
public void copy(T t);
}
public static class A<T extends A<T>> implements Copyable<T> {
private String a = "--A--";
public void copy(T t) {
System.out.println(t.a);
}
#SuppressWarnings("unchecked")
public static Copyable<Object> getInstance() {
return new A();
}
}
public static class B<T extends B<T>> extends A<T> {
private String b = "--B--";
public void copy(T t) {
super.copy(t);
System.out.println(t.b);
}
#SuppressWarnings("unchecked")
public static Copyable<Object> getInstance() {
return new B();
}
}
#Test
public void testCopy() {
Copyable<Object> b1 = B.getInstance();
Copyable<Object> b2 = B.getInstance();
Copyable<Object> a = A.getInstance();
b1.copy(b2); // this works as intended
try {
b1.copy(a); // this throws ClassCastException
fail();
} catch (ClassCastException cce) {
}
}
}
And also i figured out all that happens in this program with help of reflection:
for (Method method : A.class.getMethods()) {
if (method.getName().equals("copy")) {
System.out.println(method.toString());
}
}
for (Method method : B.class.getMethods()) {
if (method.getName().equals("copy")) {
System.out.println(method.toString());
}
}
Here is the output:
public void com.sbp.core.TestGenerics$A.copy(com.sbp.core.TestGenerics$A)
public void com.sbp.core.TestGenerics$A.copy(java.lang.Object)
public void com.sbp.core.TestGenerics$B.copy(com.sbp.core.TestGenerics$B)
public void com.sbp.core.TestGenerics$B.copy(com.sbp.core.TestGenerics$A)
public void com.sbp.core.TestGenerics$A.copy(java.lang.Object)
It means that:
The copy(...) methods in A and B make compiler generate "bridges" -
2 different methods for each, one with reifed argument type from
ancestor (reified T from Copyable becomes Object, reified "T extends
A" from A becomes A) and that is why it's override and not overload,
and the other one with reified argument type for defining class. First
method (with autogenerated body) downcasts its argument to call the
second (they call it a bridge). Because of this downcasting we get
ClassCastException in runtime if we call b1.copy(a).
It looks like direct type casting is cleaner and better tool for my
problem and generics are better used in their direct purpose - to
enforce compile time type checking.
I've learned Scala and now i know that the thing i wanted 2 years ago could have been achieved with contravariant type parameter and Scala's type system:
trait CopyableTo[-T] {
def copyTo(t: T)
}
class A(private var a: Int) extends CopyableTo[A] {
override def copyTo(t: A) {
println("A:copy")
t.a = this.a
}
}
class B(private var a: Int, private var b: Int) extends A(a) with CopyableTo[B] {
def copyTo(t: B) {
println("B:copy")
super.copyTo(t)
t.b = this.b
}
}
#Test
def zzz {
val b1 = new B(1, 2)
val a1 = new A(3)
val b2 = new B(4, 5)
b1.copyTo(a1)
a1.copyTo(b1)
b1.copyTo(b2)
}
Java type system is too weak for this.