Kotlin: Initialize class attribute in constructor - constructor

I create a Kotlin-class with a class attribute, which I want to initialize in the constructor:
public class TestClass {
private var context : Context? = null // Nullable attribute
public constructor(context : Context) {
this.context = context
}
public fun doSomeVoodoo() {
val text : String = context!!.getString(R.string.abc_action_bar_home_description)
}
}
Unfortunately I have to declare the attribute as Nullable with the "?" sign, although the attribute will be initialized in the constructor. Declaring this attribute as Nullable-attribute makes it always necessary to force an NonNull-value with "!!" or to provide a Null-check with "?".
Is there any way to avoid this, if the class attribute will be initialized in the constructor? I would like to appreciate a solution like this one:
public class TestClass {
private var context : Context // Non-Nullable attribute
public constructor(context : Context) {
this.context = context
}
public fun doSomeVoodoo() {
val text : String = context.getString(R.string.abc_action_bar_home_description)
}
}

As shown by D3xter you have the option of setting it in the constructor. You also have other options. Here they all are...
Create the property within the constructor (as per #D3xter), this is the most common case for simple properties initialized directly by the primary constructor:
class TestClass(private val context: Context) {
fun doSomeVoodoo() {
val text : String = context.getString()
}
}
You can declare the val property and not initialize it, assuming all possible constructors do actually initialize it (as per your second example in the question being asked). This is normal when you have more than one constructor that could initialize a value differently:
public class TestClass {
private val context: Context
public constructor(context : Context) {
this.context = context
}
// alternative constructor
public constructor(pre: PreContext) {
this.context = pre.readContext()
}
public fun doSomeVoodoo() {
val text : String = context.getString()
}
}
You can pass in constructor parameters that are not property declarations, and then use those within property initializations. This is common when you have more complex initializations or need to use delegated properties:
class TestClass(context: PreContext) {
private val context : Context by lazy { context.readContext() }
private val other: List<Items> = run {
context.items.map { it.tag }.filterNotNull()
}
private val simpleThing = context.getSimple()
fun doSomeVoodoo() {
val text : String = context.getString()
}
}
Using lateinit modifier when you cannot initialize the value during construction, but you are sure it will be done before your first read access. This is common when a dependency injection, IoC container, or something creates an empty version of your class and then initializes it immediately:
class TestClass() {
private lateinit var context : Context // set by something else after construction
fun doSomeVoodoo() {
val text : String = context.getString()
}
}
For lateinit the property must currently be a var and does not work with primitive types.
You can also declare a var property and not initialize it if you use a delegate designed for that purpose, such as Delegates.notNull(). This is similar to lateinit and common when you want a var that has no initial state, but is set later after construction at unknown point in time:
public class TestClass() {
private var context: Context by Delegates.notNull()
public fun doSomeVoodoo() {
// if context is not set before this is called, an exception is thrown
val text : String = context.getString()
}
}

If the only thing you are doing in the constructor is an assignment,
then you could use the Primary Constructor with a private Property.
e.g:
public class TestClass(private val context: Context) {
public fun doSomeVoodoo() {
val text = context.getString(R.string.abc_...)
}
}

I had a similar problem where I didn't want to hold onto the object after construction. Using lazy or lateinit resulted in inefficient bytecode so after some research I settled on this approach and returned to post the answer in case it helps:
Solution
class TestClass(context: Context) {
private val homeDescription: String
init {
homeDescription = context.getString(R.string.abc_action_bar_home_description)
}
fun doSomeVoodoo() {
val text : String = homeDescription
}
}
alternatively, the above can be further simplified into:
class TestClass(context: Context) {
private val homeDescription: String = context.getString(R.string.abc_action_bar_home_description)
fun doSomeVoodoo() {
val text : String = homeDescription
}
}
Decompiled Bytecode
And the decompiled java version of this feels a bit more acceptable than the other approaches and no reference to the context is held after construction:
public final class TestClass {
private final String homeDescription;
public final void doSomeVoodoo() {
String text = this.homeDescription;
}
public TestClass(#NotNull Context context) {
Intrinsics.checkParameterIsNotNull(context, "context");
super();
String var10001 = context.getString(2131296256);
Intrinsics.checkExpressionValueIsNotNull(var10001, "context.getString(R.striā€¦ion_bar_home_description)");
this.homeDescription = var10001;
}
}

Related

Options pattern, configuration, in Legacy .NET Application with Simple Injector

This article, https://medium.com/#dmitryzaets/legacy-net-applications-configuration-management-net-framework-4-5-1-68220335d9d8, describe how to use Options pattern together with Autofac. I have tried to translate this to use with Simple Injector. But I have no luck.
Here is my IOC code
public class IocBootstrap2
{
private Container Container { get; }
public IocBootstrap2()
{
Container = new Container();
var configurationBuilder = new ConfigurationBuilder()
.SetBasePath(Path.Combine(System.AppDomain.CurrentDomain.BaseDirectory, "Configuration"))
.AddJsonFile("settings.json", optional: false, reloadOnChange: true);
var configuration = configurationBuilder.Build();
//Register Options
Container.Register(typeof(IOptions<>), typeof(OptionsManager<>));
Container.Register(typeof(IOptionsMonitor<>), typeof(OptionsMonitor<>));
Container.Register(typeof(IOptionsFactory<>), typeof(OptionsFactory<>));
Container.Register(typeof(IOptionsMonitorCache<>), typeof(OptionsCache<>));
// Register ConfigurationOptions
Container.RegisterConfigurationOptions2<MailingOptions>(configuration.GetSection("mailing"));
#if DEBUG
Container.Verify();
#endif
}
}
public static class ConfigurationSetupExtensions2
{
public static void RegisterConfigurationOptions2<TOptions>(this Container container, IConfiguration config)
where TOptions : class
{
container.Register(typeof(IOptionsChangeTokenSource<TOptions>),
() => new ConfigurationChangeTokenSource<TOptions>(config), Lifestyle.Transient);
container.Register(typeof(IConfigureOptions<TOptions>),
() => new ConfigureFromConfigurationOptions<TOptions>(config), Lifestyle.Transient);
}
}
public class MailingOptions
{
public MailingOptions()
{
BatchSize = 1;
}
public int BatchSize { get; set; }
public int BatchDelay { get; set; }
}
settings.json
{
"mailing": {
"batchSize": 15,
"batchDelay": 1
}
}
Then I inject it in a ViewModel:s constructor like this
public class BlockViewModel
{
private readonly MailingOptions _options;
#region Constructor
public BlockViewModel(IOptions<MailingOptions> options)
{
_options = options.Value;
}
#endregion
}
When I run it I get Exceptions in Container.Verify.
The constructor of type OptionsFactory<MailingOptions> contains the parameter with name 'setups' and type IEnumerable<IConfigureOptions<MailingOptions>> that is not registered. Please ensure IEnumerable<IConfigureOptions<MailingOptions>> is registered, or change the constructor of OptionsFactory<MailingOptions>. There is, however, a registration for IConfigureOptions<MailingOptions>; Did you mean to depend on IConfigureOptions<MailingOptions>?
StackTrace:
at SimpleInjector.Container.ThrowParameterTypeMustBeRegistered(InjectionTargetInfo target)
How will I Register an IEnumerable<IConfigureOptions<MailingOptions>>?
Can someone tell my what I'm doing wrong, or more precise, what is it that I don't understand?
The short answer is: don't inject IOptions<T> into your application components. As explained here, that will only complicate your components, their unit tests, and, as you already noticed, your configuration.
Instead, let BlockViewModel depend on MailingOptions directly:
public class BlockViewModel
{
private readonly MailingOptions _options;
public BlockViewModel(MailingOptions options)
{
_options = options ?? throw new ArgumentNullException(nameof(options));
}
}
This allows you to simplify your configuration to the following:
Container = new Container();
var configutation = new ConfigurationBuilder()
.SetBasePath(Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "Configuration"))
.AddJsonFile("settings.json", optional: false);
.Build();
MailingOptions options = configuration.GetSection("mailing").Get<MailingOptions>();
Container.RegisterInstance<MailingOptions>(options);
// Register View Models
Container.Register<BlockViewModel>();
Container.Verify();

Unit test WCMUsePOJO class

I am writing unit test cases for following class which extends WCMUsePOJO. Now, this class is using a getSlingScriptHelper method shown below.
public class ConstantsServiceProvider extends WCMUsePojo {
private static final Logger logger = LoggerFactory.getLogger(ConstantsServiceProvider.class);
private String var1;
#Override
public void activate() throws Exception {
ConstantsService constantsService = getSlingScriptHelper().getService(ConstantsService.class);
if(constantsService != null) {
var1 = constantsService.getVar1();
}
}
public string getVar1() { return var1; }
}
The question is how do I mock getSlingScriptHelper method? Following is my unit test code.
public class ConstantsServiceProviderTest {
#Rule
public final SlingContext context = new SlingContext(ResourceResolverType.JCR_MOCK);
#Mock
public SlingScriptHelper scriptHelper;
public ConstantsServiceProviderTest() throws Exception {
}
#Before
public void setUp() throws Exception {
ConstantsService service = new ConstantsService();
scriptHelper = context.slingScriptHelper();
provider = new ConstantsServiceProvider();
provider.activate();
}
#Test
public void testGetvar1() throws Exception {
String testvar1 = "";
String var1 = provider.getVar1();
assertEquals(testvar1, var1);
}
}
The only thing that you should "have to"* mock is the SlingScriptHelper instance itself, so that it will mimic the dependency injection of the declared service.
Everything else (e.g. the Bindings instance) can be a concrete implementation, for example:
import org.apache.sling.api.scripting.SlingBindings;
import org.apache.sling.api.scripting.SlingScriptHelper;
import org.junit.Test;
import javax.script.Bindings;
import javax.script.SimpleBindings;
import static org.hamcrest.CoreMatchers.equalTo;
import static org.hamcrest.CoreMatchers.is;
import static org.hamcrest.MatcherAssert.assertThat;
import static org.mockito.Mockito.mock;
import static org.mockito.Mockito.when;
public class ConstantsServiceProviderTest {
private SlingScriptHelper mockSling = mock(SlingScriptHelper.class);
private ConstantsServiceProvider constantsServiceProvider = new ConstantsServiceProvider();
private Bindings bindings = new SimpleBindings();
#Test
public void testFoo() throws Exception {
//Arrange
final String expected = "Hello world";
final ConstantsService testConstantsService = new TestConstantsService(expected);
when(mockSling.getService(ConstantsService.class)).thenReturn(testConstantsService);
bindings.put(SlingBindings.SLING, mockSling);
//Act
constantsServiceProvider.init(bindings);
//Assert
final String actual = constantsServiceProvider.getVar1();
assertThat(actual, is(equalTo(expected)));
}
class TestConstantsService extends ConstantsService {
String var1 = "";
TestConstantsService(String var1) {
this.var1 = var1;
}
#Override
String getVar1() {
return var1;
}
}
}
The entry point here, as you said above, is via the init() method of the WCMUsePojo superclass (as this method is an implementation of the Use.class interface, this test structure also works for testing that via that interface, even if you don't use WCMUsePojo directly.)
*this could be any type of test-double, not necessarily a mock.
You shouldn't create a mock for ConstantsServiceProvider.class if you want to unit-test it. Instead, you should create mocks of its internal objects. So:
Create real instance of ConstantsServiceProvider with new
Mock objects that are returned by getSlingScriptHelper().getService(.) methods. Usually, dependencies are provided (injected) to classes by some container like Spring or simply provided by other classes of your app using setters. In both cases mocks creation is easy.
If your current implementation doesn't allow this - consider refactoring.
You are testing void activate() method which doesn't return anything. So, you should verify calling constantsService.getVar1() method.
I strongly advice you to study Vogella unit-testing tutorial
Here one of possible solution.
The main idea is to have a real object of your class but with overridden getSlingScriptHelper() to return mocked scriptHelper.
I mocked the ConstantsService as well but may be not needed, I don't know your code.
public class ConstantsServiceProviderTest {
#Mock
public SlingScriptHelper scriptHelper;
#Test
public void getVar1ReturnsActivatedValue() throws Exception {
// setup
final String expectedResult = "some value";
// Have a mocked ConstantsService, but if possible have a real instance.
final ConstantsService mockedConstantsService =
Mockito.mock(ConstantsService.class);
Mockito.when(
mockedConstantsService.getVar1())
.thenReturn(expectedResult);
Mockito.when(
scriptHelper.getService(ConstantsService.class))
.thenReturn(mockedConstantsService);
// Have a real instance of your class under testing but with overridden getSlingScriptHelper()
final ConstantsServiceProvider providerWithMockedHelper =
new ConstantsServiceProvider() {
#Override
SlingScriptHelper getSlingScriptHelper() {
return scriptHelper;
}
};
// when
String actualResult = providerWithMockedHelper.getVar1();
// then
assertEquals(expectedResult, actualResult);
}
}

Actionscript: Accessing private setter in own constructor

I am trying to set a variable whose setter is private in the constructor of the same class, but compiler keeps hitting me on the head with:
1059 Property is readonly
1178: Attempted access of inaccessible property gesamtBetrag through a
reference with static type ImmoObject.
package
{
public class Huhu
{
private var _prop1:int;
public function Huhu()
{
prop1 = 24;
}
public function get prop1():int
{
return _prop1;
}
private function set prop1(value:int):void
{
_prop1 = value;
}
}
}
Am I missing something here?
As stated, you cannot have a private setter and a public getter. The access modifier must be the same for getter and setter.
What you can do is only have a public getter and from within the class constructor or elsewhere you set the private variable directly:
class Foo {
private var _prop:int;
public function Foo() {
_prop = 24;
}
public function get prop():int { return _prop; }
}
(Note that in this case you could actually just initialize the private variable with a value, ex private _prop:int = 24.)
You could also expose other means of changing the private variable, such as a function or a setter with a different name:
class Foo {
private var _prop:int;
public function get prop():int { return _prop; }
protected function setProp(prop:int):void {
_prop = prop;
}
protected function set propValue(value:int):int {
_prop = value;
}
}
Edit: To be clear, access types must match for getter/setters of properties, but you can mix private/public access of getter/setter methods, in similar style as Java:
private prop:int;
public getProp():int { return prop; }
private setProp(value:int):void {
prop = value;
}
This is not as common of a style you see in AS3 because we have formal get/set property syntax, but it's still perfectly valid.
Getter/setter must have the same signature or else they are simply not related. You have a public getter and a private setter, as a result they are not related and the public getter has no setter so it is read only.

ninject factory constructor selection with runtime generics

How can I use a ninject factory, which creates an instance with constructor parameters, without relying on the argument names.
The problem is that the ToConstructor()-Method not works, because I bind it to a generic definition.
The following example works, if I use the factory method with the corresponding constructor argument name, but I don't like it rely on names.
Because the following solution is very fragil and breaks if someone chooses a wrong name or renames the ctor-argument in the derived class.
Any solution?
Here's the example code:
[TestFixture]
public class NinjectFactoryBindingsTest
{
[Test]
public void ConstructorSelectionWithArguments()
{
NinjectSettings ninjectSettings = new NinjectSettings();
ninjectSettings.LoadExtensions = false;
using (var kernel = new StandardKernel(ninjectSettings, new FuncModule()))
{
// IDependencyA will be passed to the factory, therefore it is not bounded
//kernel.Bind<IDependencyA>().To<DependencyA>();
kernel.Bind<IDependencyB>().To<DependencyB>();
kernel.Bind(typeof(IGenericBaseClass<>)).To(typeof(GenericDerivedClass<>));
kernel.Bind<IGenericClassFactory>().ToFactory();
IGenericClassFactory factory = kernel.Get<IGenericClassFactory>();
DependencyA dependencyA = new DependencyA();
IGenericBaseClass<GenericImpl> shouldWorkInstance = factory.Create<GenericImpl>(dependencyA);
Assert.NotNull(shouldWorkInstance);
}
}
}
public interface IGenericClassFactory
{
IGenericBaseClass<TGeneric> Create<TGeneric>(IDependencyA someName) where TGeneric : IGeneric;
// This works, but relies on ctor-param-names!!!
// IGenericBaseClass<TGeneric> Create<TGeneric>(IDependencyA otherNameThanInBaseClass) where TGeneric : IGeneric;
}
public class DependencyA : IDependencyA
{
}
public class DependencyB : IDependencyB
{
}
public class GenericDerivedClass<TGeneric> : GenericBaseClass<TGeneric> where TGeneric : IGeneric
{
public GenericDerivedClass(IDependencyA otherNameThanInBaseClass, IDependencyB dependencyB)
: base(otherNameThanInBaseClass, dependencyB)
{
}
}
public abstract class GenericBaseClass<TGeneric> : IGenericBaseClass<TGeneric> where TGeneric : IGeneric
{
protected GenericBaseClass(IDependencyA dependencyA, IDependencyB dependencyB)
{
}
}
public interface IGenericBaseClass<TGeneric> where TGeneric : IGeneric
{
}
public interface IDependencyB
{
}
public interface IDependencyA
{
}
public class GenericImpl : IGeneric
{
}
public interface IGeneric
{
}
The factory extension has the convention that arguments must have the same name as the constructor argument they will be passed to. There are no easy ways to do it differently. The only way I can think of is about the following:
Create a new IParameter implementation that can hold a reference to an IDependencyA.
Create either a hardcoded factory or a custom IInstanceProvider (see documentation) that creates an instance of your IParameter implementation so that it is passed to the Get<> request
Add a new binding for IDependencyA: Bind<IDependency>().ToMethod(ctx => extract and return your parameter from the context)

How to Override Constants in ActionScript 3

I have the two following classes:
public class Parent{
static internal const _name:String = "Parent";
public function get name():String{
return _name;
}
}
public class Child{
static internal const _name:String = "Child";
}
If I create an instance of class Child and call its name() getter, since it will call the name() method it inherits from Parent, it returns "Parent". I could, of course, override the name() method:
public class Child{
static internal const _name:String = "Child";
override public function get name():String{
return _name;
}
}
That returns "Child". However, it seems silly to have to copy the exact same code of the method from the parent. Is there any simpler way to do this?
I would take a different approach by making the "name" property a requirement for the parent's constructor:
public class Parent
{
static internal var _name : String;
public function Parent(name : String = "Parent") {
_name = name;
}
public function get name() : String {
return _name;
}
}
Child Class:
public class Child extends Parent
{
public function Child() {
super("Child");
}
}
Firstly, you cannot override static methods or properties - they are not inherited, so no override for them.
Secondly, if you declared a constant to be of a complex type, it is not really a constant. I.e. if it is an object, then you can change its keys / values, if it is an array, you can add / remove members and so on.
But the desire to make this functionality more generic is understandable. So, what I'd do:
Have some property outside both parent and child, let say in class X, or package Y. Let it be package Y. So, you'd create a dictionary in package Y, let it be Y.names and in your name getter you'd do:
import Y.names;
. . .
public function get name() {
return names[(this as Object).constructor];
}
your names variable would be:
package Y {
public var names:Dictionary = generateNames();
internal function generateNames():Dictionary {
var result:Dictionary = new Dictionary();
result[ChildClass] = "child";
result[ParentClass] = "parent";
. . .
return result;
}
}
This way it would be sufficient to only implement name getter in super-class, and all inheriting classes will be able to use super-class code as is, no need to change anything. However, this means that some (maybe important) information pertaining to this class will be stored elsewhere (may be difficult to find, this is not the common way people program in AS3).
your implementation of get name should look like this, then the getter is one and each of the new classes needs to have it's own public static var _name defined:
//in the base class
public function get name():String
{
var _sName:String;
if ((this as Object).constructor._name)
{
_sName = (this as Object).constructor._name;
}
else
{
try
{
var o:Object = getSuperClass(this);
while (o)
{
if (o._name)
{
_sName = o._name;
break;
}
o = getSuperClass(o);
}
}
catch (e:*)
{}
}
return _sName;
}
//as found here: http://www.actionscriptdeveloper.co.uk/getting-the-class-of-an-object-in-as3/
public static function getSuperClass(o: Object): Object
{
var n: String = getQualifiedSuperclassName(o);
if (n == null)
return(null);
return getDefinitionByName(n);
}
the static members can be accessed only via class reference which we can get from constructor object, "this" will point to the current class in the inheritance chain so you can call this in parent class and it will point to a Child in a Child class.
[EDIT]
I've modified it so it tests for existance of the public static property _name if not found on "this" instance then in a loop the parent class is checked until one is found - like inheritance:)
I'm using this feature to create clone method: constructor as helper in clone method implementation
best regards
Why don't you store such a constant within a corresponding function instead of declaring an inaccessible constant?
class Parent {
...
public function get name():String { return 'Parent'; }
}
class Child extends Parent {
...
override public function get name():String { return 'Child'; }
}
By the way, if your Parent class is a descendant of DisplayObject, you should be careful with name property, as it's needed sometimes by operating code, e.g. getChildByName().
I have found something that seems to work. Any feedback is greatly appreciated:
public class Parent{
prototype._name = "Parent";
public function get name():String{
return this["_name"];
}
}
public class Child{
prototype._name = "Child";
}