Currently I am working on the Racing-Car-Katas. The goal is to refactor a peach of code so that it follows the solid-principles.
I try to add the Dependency Inversion Principle. Where I can pass a dependency through the constructor.
Initial Situation
Insid the class Alarm is the dependency Sensor which generates a psiPressureValue.
public class Alarm {
Sensor sensor = new Sensor();
public void check()
{
double psiPressureValue = sensor.popNextPressurePsiValue();
/* ... */
}
}
Idea
public class Alarm {
Sensor sensor;
public Alarm() {
this.sensor = new Sensor();
}
public Alarm(ISensor sensor) {
this.sensor = sensor;
}
public void check()
{
double psiPressureValue = sensor.popNextPressurePsiValue();
/* ... */
}
}
If this would be a real application I don't want to break any dependency between Alarm and Sensor. There for I would created the following constructor
public Alarm() {
this.sensor = new Sensor();
}
But my gut feeling says this is code smell..
How to handle such dependencies in real world applications?
The Dependency Inversion Principle (DIP) states:
High-level modules should not depend on low-level modules.
But by defining the default constructor that creates a Sensor implementation, however, you are violating the DIP, because Sensor is a low-level module and Alarm (your high-level module) takes a dependency on it:
public Alarm() {
this.sensor = new Sensor();
}
If there is a need to let the high level module depend on an abstraction (as your additional constructor shows), there is no need to add a default constructor. Doing so only drags along the dependency to the low-level module. Since both your final application and your tests should make use of the overloaded constructor, the default constructor has no meaning and only causes tight coupling and therefore a violation of the DIP.
This is not a theoretical exercise. The DIP should be followed in real-world applications. A well-designed real-world application applies the SOLID principles and uses Dependency Injection as way to achieve loose coupling and the DIP. This decouples modules and allows complete object graphs to be composed in the Composition Root.
In order to not break any dependency between Alarm and Sensor, you need to:
Take all the methods of Sensor that are used by Alarm;
Put them in one (or more) interface(s); and then,
Have Sensor implement those interfaces.
The second constructor approach is not the right way to secure the dependencies. It is violating the DIP by creating an object of Sensor class.
Instead of that, add validation code to the other constructor and make sure that Alarm is always using a valid Sensor through an ISensor (and not create an instance of a Sensor directly).
Ok you want to break the direct coupling between Alarm and Sensor.
Your proposed solutions shows two constructors, one injecting the Sensor object (created externally) and one creating Sensor object directly. You should drop:
public Alarm() {
this.sensor = new Sensor();
}
because its not needed and because it prolongs the tight coupling you are trying to avoid with SOLID. Dependency is not broken by doing this.
Broadly the most common options for creating dependencies are as below.
Dependent objects are created:
1) Directly by the Client object (tight coupling)
2) Indirectly by the Client using Factory Methods of Client
3) Indirectly by the Client using Abstract Factory object injected into the Client (the factory injection is DI). You can inject different factories.
4) Externally manually and injected into the Client (DI)
5) Externally via a DI Container and injected into the Client (DI)
Its a key point of DI that the Client does not control how its dependants are created. This is explains the term "Inversion of Control".
However, the client retains control with "Factory Method", "Abstract Factory".
DI containers have become extremely popular in recent years especially with modern frameworks.
Related
A question I've been struggling a lot with lately is how, in my opinion, Inversion of Control
breaks Encapsulation and can easily lead to side effects in a program. However, at the same time, some of the
big advatages of IoC is loose coupling/modularity as well as Test Driven Design making Unit testing a class much easier (I think TDD is really pushing IoC in the industry).
Here is my argument againt IoC.
If the injected types are Immutable and Pure then IoC is acceptable, for example primitive types. However, if they are impure
and can modify the state of the program or hold their own state then side effects can easily
occur.
Take the following example C#/Pseudo:
public class FileSearcher: IFileSearcher
{
private readonly string searchPath;
public void SetSearchPath(string path)
{
searchPath = path;
}
public List<string> FindFiles(string searchPattern)
{
//...Search for files with searchPattern starting at searchPath
}
}
public class PlayListViewer
{
public PlayListViewer(string playlistName, IFileSearcher searcher)
{
searcher.SetSearchPath($"playlists/{playlistName}")
}
public List<string> FindSongNames()
{
return searcher.FindFiles(
"*.mp3|*.wav|*.flac").Select(f => Path.GetFileName(f))
}
//....other methods
}
public class Program
{
public static void Main()
{
var searcher = FileSearcher();
var viewer = PlayListViewer("Hits 2021", searcher);
searcher.SetSearchPath("C:/Users") //Messes up search path
var pictures = searcher.FindFiles("*.jpg") //Using searcher for something else
viewer
.FindSongNames()
.ForEach(s => Console.WriteLine(s)) //WRONG SONGS
}
}
In the (very uncreative) example above, The PlaylistViewer has a method for finding songs within a playlist. It
attempts to set the correct search path for the playlist on the injected IFileSearcher, but the User of the class
overwrote the path. Now when they try to find the songs in the playlist, the results are incorrect.
The Users of a class do not always know the implementation of the class they're using and don't know the side
effects they're causing by mutating the objects they passed in.
Some other simple examples of this:
The Date Class in Java is not immutable and has a setDate (deprecated now) method. The following could occur:
date = new Date(2021, 10, 1)
a = new A(date)
a.SomethingInteresting() //Adds 1 year to the Date using setDate
b = new B(date) //No longer the correct date
I/O abstractions such as streams:
audioInput = new MemoryStream()
gainStage = new DSPGain(audioInput)
audioInput.write(....)
audioInput.close()
gainStage.run() //Error because memory stream is already closed
etc...
Other issues can come up too if the Object gets passed to multiple classes that use it across different threads concurrently. In these cases
a User might not know/realize that class X internally is launching/processing on a different thread.
I think the simple, and functional, answer would be to only write pure functions and immutable classes but that isn't always practical in the real world.
So when should IoC really be used? Maybe only when the injected types are immutable and pure and anything else should be composed and encapsulated? If that's the answer, then what does that mean for TDD?
First, Inversion of Control is not the same as Dependency Injection. DI is just one implementation of IoC. This question makes more sense if we limit it to just DI.
Second, Dependency Injection is orthogonal to Test Driven Development. DI can make writing unit tests easier, which may encourage you to write more unit tests; but that does not necessitate TDD. You can certainly use DI without TDD, and I suspect that's the way the vast majority of developers use it. TDD is not a widespread practice.
Conversely, practicing TDD may encourage you to implement DI; but that is far from a requirement. Don't confuse statements like, "TDD and DI work well together," with "TDD and DI require each other." They can be used together or separately.
Finally, if you want to use your DI container as a repository of global variables, you certainly can. This approach of storing mutable state and injecting it across your application brings the same caveats and pitfalls as sharing mutable state anywhere else.
That should be the main takeaway from this question: not the downside of DI or TDD, but the downside of mutable state in general. You don't need DI to run afoul of mutable state. Trouble with mutable state is virtually guaranteed in the practice of imperative programming, which is by far the most common programming paradigm.
Consider that the functional programmers might really be onto something with their declarative approach.
I'm currently learning the API for Autofac, and I'm trying to get my head around what seems to me like a very common use case.
I have a class (for this simple example 'MasterOfPuppets') that has a dependency it receives via constructor injection ('NamedPuppet'), this dependency needs a value to be built with (string name):
public class MasterOfPuppets : IMasterOfPuppets
{
IPuppet _puppet;
public MasterOfPuppets(IPuppet puppet)
{
_puppet = puppet;
}
}
public class NamedPuppet : IPuppet
{
string _name;
public NamedPuppet(string name)
{
_name = name;
}
}
I register both classes with their interfaces, and than I want to resolve IMasterOfPuppets, with a string that will be injected into the instance of 'NamedPuppet'.
I attempted to do it in the following way:
IMasterOfPuppets master = bs.container.Resolve<IMasterOfPuppets>(new NamedParameter("name", "boby"));
This ends with a runtime error, so I guess Autofac only attempts to inject it to the 'MasterOfPuppets'.
So my question is, how can I resolve 'IMasterOfPuppets' only and pass parameter arguments to it's dependency, in the most elegant fashion?
Do other ioc containers have better solutions for it?
Autofac doesn't support passing parameters to a parent/consumer object and having those parameters trickle down into child objects.
Generally I'd say requiring the consumer to know about what's behind the interfaces of its dependencies is bad design. Let me explain:
From your design, you have two interfaces: IMasterOfPuppets and IPuppet. In the example, you only have one type of IPuppet - NamedPuppet. Keeping in mind that the point of even having the interface is to separate the interface from the implementation, you might also have this in your system:
public class ConfigurablePuppet : IPuppet
{
private string _name;
public ConfigurablePuppet(string name)
{
this._name = ConfigurationManager.AppSettings[name];
}
}
Two things to note there.
First, you have a different implementation of IPuppet that should work in place of any other IPuppet when used with the IMasterOfPuppets consumer. The IMasterOfPuppets implementation should never know that the implementation of IPuppet changed... and the thing consuming IMasterOfPuppets should be even further removed.
Second, both the example NamedPuppet and the new ConfigurablePuppet take a string parameter with the same name, but it means something different to the backing implementation. So if your consuming code is doing what you show in the example - passing in a parameter that's intended to be the name of the thing - then you probably have an interface design problem. See: Liskov substitution principle.
Point being, given that the IMasterOfPuppets implementation needs an IPuppet passed in, it shouldn't care how the IPuppet was constructed to begin with or what is actually backing the IPuppet. Once it knows, you're breaking the separation of interface and implementation, which means you may as well do away with the interface and just pass in NamedPuppet objects all the time.
As far as passing parameters, Autofac does have parameter support.
The recommended and most common type of parameter passing is during registration because at that time you can set things up at the container level and you're not using service location (which is generally considered an anti-pattern).
If you need to pass parameters during resolution Autofac also supports that. However, when passing during resolution, it's more service-locator-ish and not so great becausee, again, it implies the consumer knows about what it's consuming.
You can do some fancy things with lambda expression registrations if you want to wire up the parameter to come from a known source, like configuration.
builder.Register(c => {
var name = ConfigurationManager.AppSettings["name"];
return new NamedPuppet(name);
}).As<IPuppet>();
You can also do some fancy things using the Func<T> implicit relationship in the consumer:
public class MasterOfPuppets : IMasterOfPuppets
{
IPuppet _puppet;
public MasterOfPuppets(Func<string, IPuppet> puppetFactory)
{
_puppet = puppetFactory("name");
}
}
Doing that is the equivalent of using a TypedParameter of type string during the resolution. But, as you can see, that comes from the direct consumer of IPuppet and not something that trickles down through the stack of all resolutions.
Finally, you can also use Autofac modules to do some interesting cross-cutting things the way you see in the log4net integration module example. Using a technique like this allows you to insert a specific parameter globally through all resolutions, but it doesn't necessarily provide for the ability to pass the parameter at runtime - you'd have to put the source of the parameter inside the module.
Point being Autofac supports parameters but not what you're trying to do. I would strongly recommend redesigning the way you're doing things so you don't actually have the need to do what you're doing, or so that you can address it in one of the above noted ways.
Hopefully that should get you going in the right direction.
I'm not so sure the title is a good match for this question I want to put on the table.
I'm planning to create a web MVC framework as my graduation dissertation and in a previous conversation with my advisor trying to define some achivements, he convinced me that I should choose a modular design in this project.
I already had some things developed by then and stopped for a while to analyze how much modular it would be and I couldn't really do it because I don't know the real meaning of "modular".
Some things are not very cleary for me, like for example, just referencing another module blows up the modularity of my system?
Let's say I have a Database Access module and it OPTIONALY can use a Cache module for storing results of complex queries. As anyone can see, I at least will have a naming dependency for the cache module.
In my conception of "modular design", I can distribute each component separately and make it interact with others developed by other people. In this case I showed, if someone wants to use my Database Access module, they will have to take the Cache as well, even if he will not use it, just for referencing/naming purposes.
And so, I was wondering if this is really a modular design yet.
I came up with an alternative that is something like creating each component singly, without don't even knowing about the existance of other components that are not absolutely required for its functioning. To extend functionalities, I could create some structure based on Decorators and Adapters.
To clarify things a little bit, here is an example (in PHP):
Before
interface Cache {
public function isValid();
public function setValue();
public function getValue();
}
interface CacheManager {
public function get($name);
public function put($name, $value);
}
// Some concrete implementations...
interface DbAccessInterface {
public doComplexOperation();
}
class DbAccess implements DbAccessInterface {
private $cacheManager;
public function __construct(..., CacheManager $cacheManager = null) {
// ...
$this->cacheManager = $cacheManager;
}
public function doComplexOperation() {
if ($this->cacheManager !== null) {
// return from cache if valid
}
// complex operation
}
}
After
interface Cache {
public function isValid();
public function setValue();
public function getValue();
}
interface CacheManager {
public function get($name);
public function put($name, $value);
}
// Some concrete implementations...
interface DbAccessInterface {
public function doComplexOperation();
}
class DbAccess implements DbAccessInterface {
public function __construct(...) {
// ...
}
public function doComplexQuery() {
// complex operation
}
}
// And now the integration module
class CachedDbAcess implements DbAccessInterface {
private $dbAccess;
private $cacheManager;
public function __construct(DbAccessInterface $dbAccess, CacheManager $cacheManager) {
$this->dbAccess = $dbAccess;
$this->cacheManager = $cacheManager;
}
public function doComplexOperation() {
$cache = $this->cacheManager->get("Foo")
if($cache->isValid()) {
return $cache->getValue();
}
// Do complex operation...
}
}
Now my question is:
Is this the best solution? I should do this for all the modules that do not have as a requirement work together, but can be more efficient doing so?
Anyone would do it in a different way?
I have some more further questions involving this, but I don't know if this is an acceptable question for stackoverflow.
P.S.: English is not my first language, maybe some parts can get a little bit confuse
Some resources (not theoretical):
Nuclex Plugin Architecture
Python Plugin Application
C++ Plugin Architecture (Use NoScript on that side, they have some weird login policies)
Other SO threads (design pattern for plugins in php)
Django Middleware concept
Just referencing another module blows up the modularity of my system?
Not necessarily. It's a dependency. Having a dependencies is perfectly normal. Without dependencies modules can't interact with each other (unless you're doing such interaction indirectly which in general is a bad practice since it hides dependencies and complicates the code). Modular desing implies managing of dependencies, not removing them.
One tool - is using interfaces. Referencing module via interface makes a so called soft dependency. Such module can accept any implementation of an interface as a dependency so it is more independant and as a result - more maintainable.
The other tool - designing modules (and their interfaces) that have only single responcibility. This also makes them more granular, independant and maintainable.
But there is a line which you should not cross - blindly applying these tools may leed to a too modular and too generic desing. Making things too granular makes the whole system more complex. You should not solve universe problems, making generic modules, that all developers can use (unless it is your goal). First of all your system should solve your domain tasks and make things generic enough, but not more than that.
I came up with an alternative that is something like creating each component singly, without don't even knowing about the existance of other components that are not absolutely required for its functioning
It is great if you came up with this idea by yourself. The statement itself, is a key to modular programming.
Plugin architecture is the best in terms of extensibility, but imho it is hard to maintenance especially in intra application. And depending the complexity of plugin architecture, it can make your code more complex by adding plugin logics, etc.
Thus, for intra modular design, I choose the N-Tier, interface based architecture. Basically, the architecture relays on those tiers:
Domain / Entity
Interface [Depend on 1]
Services [Depend on 1 and 2]
Repository / DAL [Depend on 1 and 2]
Presentation Layer [Depend on 1,2,3,4]
Unfortunately, I don't think this is achieveable neatly in php projects as it need separated project / dll references in each tier. However, following the architecture can help to modularize the application.
For each modules, we need to do interface-based design. It can help to enhance the modularity of your code, because you can change the implementation later, but still keep the consumer the same.
I have provided an answer similiar to this interface-based design, at this stackoverflow question.
Lastly but not least, if you want to make your application modular to the UI, you can do Service Oriented Architecture. This is simply make your application as bunch of services, and then make the UI to consume the service. This design can help to separate your UI with your logic. You can later use different UI such as desktop app, but still use the same logic. Unfortunately, I don't have any reliable source for SOA.
EDIT:
I misunderstood the question. This is my point of view about modular framework. Unfortunately, I don't know much about Zend so I will give examples in C#:
It consist of modules, from the smallest to larger modules. Example in C# is you can using the Windows Form (larger) at your application, and also the Graphic (smaller) class to draw custom shapes in the screen.
It is extensible, or replaceable without making change to base class. In C# you can assign FormLoad event (extensible) to the Form class, inherit the Form or List class (extensible) or overridding form draw method to create a custom window graphic (replaceable).
(optional) it is easy to use. In normal DI interface design, we usually inject smaller modules into a larger (high level) module. This will require an IOC container. Refer to my question for detail.
Easy to configure, and does not involve any magical logic such as Service Locator Pattern. Search Service Locator is an Anti Pattern in google.
I don't know much about Zend, however I guess that the modularity in Zend can means that it can be extended without changing the core (replacing the code) inside framework.
If you said that:
if someone wants to use my Database Access module, they will have to take the Cache as well, even if he will not use it, just for referencing/naming purposes.
Then it is not modular. It is integrated, means that your Database Access module will not work without Cache. In reference of C# components, it choose to provide List<T> and BindingList<T> to provide different functionality. In your case, imho it is better to provide CachedDataAccess and DataAccess.
In the Head First Design Patterns book, the author often says that one should program to interface rather than implementation?
What does that mean?
Let's illustrate it with the following code:
namespace ExperimentConsoleApp
{
class Program
{
static void Main()
{
ILogger loggerA = new DatabaseLogger();
ILogger loggerB = new FileLogger();
loggerA.Log("My message");
loggerB.Log("My message");
}
}
public interface ILogger
{
void Log(string message);
}
public class DatabaseLogger : ILogger
{
public void Log(string message)
{
// Log to database
}
}
public class FileLogger : ILogger
{
public void Log(string message)
{
// Log to File
}
}
}
Suppose you are the Logger developer and the application developer needs a Logger from you. You give the Application developer your ILogger interface and you say to him he can use but he doesn't have to worry about the implementation details.
After that you start developing a FileLogger and Databaselogger and you make sure they follow the interface that you gave to the Application developer.
The Application developer is now developing against an interface, not an implementation. He doesn't know or care how the class is implemented. He only knows the interface. This promotes less coupling in code and it gives you the ability to (trough configuration files for example) easily switch to another implementation.
Worry more about what a class does rather than how it does it. The latter should be an implementation detail, encapsulated away from clients of your class.
If you start with an interface, you're free to inject in a new implementation later without affecting clients. They only use references of the interface type.
It means that when working with a class, you should only program against the public interface and not make assumptions about how it was implemented, as it may change.
Normally this translates to using interfaces/abstract classes as variable types instead of concrete ones, allowing one to swap implementations if needed.
In the .NET world one example is the use of the IEnumerable/IEnumerator interfaces - these allow you to iterate over a collection without worrying how the collection was implemented.
It is all about coupling. Low coupling is a very important property* of software architecture. The less you need to know about your dependency the better.
Coupling can be measured by the number of assumptions you have to make in order to interact/use your dependency (paraphrasing M Fowler here).
So when using more generic types we are more loosely coupled. We are for example de-coupled from a particular implementation strategy of a collection: linked list, double linked list, arrays, trees, etc. Or from the classic OO school: "what exact shape it is: rectangle, circle, triangle", when we just want to dependent on a shape (in old school OO we apply polymorphism here)
I am going over some OO basics and trying to understand why is there a use of Interface reference variables.
When I create an interface:
public interface IWorker
{
int HoneySum { get; }
void getHoney();
}
and have a class implement it:
public class Worker : Bee, IWorker
{
int honeySum = 15;
public int HoneySum { get { return honeySum; } }
public void getHoney()
{
Console.WriteLine("Worker Bee: I have this much honey: {0}", HoneySum);
}
}
why do people use:
IWorker worker = new Worker();
worker.getHoney();
instead of just using:
Worker worker3 = new Worker();
worker3.getHoney();
whats the point of a interface reference variable when you can just instatiate the class and use it's methods and fields that way?
If your code knows what class will be used, you are right, there is no point in having an interface type variable. Just like in your example. That code knows that the class that will be instantiated is Worker, because that code won't magically change and instantiate anything else than Worker. In that sense, your code is coupled with the definition and use of Worker.
But you might want to write some code that works without knowing the class type. Take for example the following method:
public void stopWorker(IWorker worker) {
worker.stop(); // Assuming IWorker has a stop() method
}
That method doesn't care about the specific class. It would handle anything that implements IWorker.
That is code you don't have to change if you want later to use a different IWorker implementation.
It's all about low coupling between your pieces of code. It's all about maintainability.
Basically it's considered good programming practice to use the interface as the type. This allows different implementations of the interface to be used without effecting the code. I.e. if the object being assigned was passed in then you can pass in anything that implements the interface without effecting the class. However if you use the concrete class then you can only passin objects of that type.
There is a programming principle I cannot remember the name of at this time that applies to this.
You want to keep it as generic as possible without tying to specific implementation.
Interfaces are used to achieve loose coupling between system components. You're not restricting your system to the specific concrete IWorker instance. Instead, you're allowing the consumer to specify which concrete implementation of IWorker they'd like to use. What you get out of it is loosely dependent components and better flexibility.
One major reason is to provide compatibility with existing code. If you have existing code that knows how to manipulate objects via some particular interface, you can instantly make your new code compatible with that existing code by implementing that interface.
This kind of capability becomes particularly important for long-term maintenance. You already have an existing framework, and you typically want to minimize changes to other code to fit your new code into the framework. At least in the ideal case, you do this by writing your new code to implement some number of existing interfaces. As soon as you do, the existing code that knows how to manipulate objects via those interfaces can automatically work with your new class just as well as it could with the ones for which it was originally designed.
Think about interfaces as protocols and not classes i.e. does this object implement this protocol as distinct from being a protocol? For example can my number object be serialisable? Its class is a number but it might implement an interface that describes generally how it can be serialised.
A given class of object may actually implement many interfaces.