OOP confusion in classes

Question

I am from a C# background and have been doing programming for quite some time now. But only recently i started giving some thoughts on how i program. Apparently, my OOP is very bad.

I have a few questions maybe someone can help me out. They are basic but i want to confirm.

1- In C#, we can declare class properties like

private int _test;

and there setter getters like

public int Test {get; set;}

Now, lets say i have to use this property inside the class. Which one will i use ? the private one or the public one ? or they both are the same ?

2- Lets say that i have to implement a class that does XML Parsing. There can be different things that we can use as input for the class like "FILE PATH". Should i make this a class PROPERTY or should i just pass it as an argument to a public function in the class ? Which approach is better. Check the following

I can create a class property and use like this

public string FilePath {get; set;}

public int Parse()
{
    var document = XDocument.Load(this.FilePath);
    .........//Remaining code
}

Or

I can pass the filepath as a parameter

public int Parse(string filePath)

On what basis should i make a decision that i should make a property or i should pass something as argument ?

I know the solutions of these questions but i want to know the correct approach. If you can recommend some video lectures or books that will be nice also.

Answer 1

There are really two questions wrapped in your first question.

1) Should I use getters and setters (Accessors and Mutators) to access a member variable.

The answer depends on whether the implementation of the variable is likely to change. In some cases, the interface type (the type returned by the getter, and set by the setter) needs to be kept consistent but the underlying mechanism for storing the data may change. For instance, the type of the property may be a String but in fact the data is stored in a portion of a much larger String and the getter extracts that portion of the String and returns it to the user.

2) What visibility should I give a property?

Visibility is entirely dependent on use. If the property needs to be accessible to other classes or to classes that inherit from the base class then the property needs to be public or protected.

I never expose implementation to external concerns. Which is to say I always put a getter and setter on public and protected data because it helps me ensure that I will keep the interface the same even if the underlying implementation changes. Another common issue with external changes is that I want a chance to intercept an outside user's attempt to modify a property, maybe to prevent it, but more likely to keep the objects state in a good or safe state. This is especially important for cached values that may be exposed as properties. Think of a property that sums the contents of an array of values. You don't want to recalculate the value every time it is referenced so you need to be certain that the setter for the elements in the array tells the object that the sum needs to be recalculated. This way you keep the calculation to a minimum.

I think the second question is: When do I make a value that I could pass in to a constructor public?

It depends on what the value is used for. I generally think that there are two distinct types of variables passed in to constructors. Those that assist in the creation of the object (your XML file path is a good example of this) and those that are passed in because the object is going to be responsible for their management. An example of this is in collections which you can often initialize the collection with an array.

I follow these guidelines.

If the value passed in can be changed without damaging the state of the object then it can be made into a property and publicly visible.

If changing the value passed in will damage the state of the object or redefine its identity then it should be left to the constructor to initialize the state and not be accesible again through property methods.

A lot of these terms are confusing because of the many different paradigms and languages in OO Design. The best place to learn about good practices in OO Design is to start with a good book on Patterns. While the so-called Gang of Four Book http://en.wikipedia.org/wiki/Design_Patterns was the standard for many years, there have since been many better books written.

Here are a couple resources on Design Patterns:

http://sourcemaking.com/design_patterns

http://www.oodesign.com/

And a couple on C# specific.

http://msdn.microsoft.com/en-us/magazine/cc301852.aspx

http://www.codeproject.com/Articles/572738/Building-an-application-using-design-patterns-and

Answer 2

Fields vs Properties

Seems like you've got a few terms confused.

private int _test;

This is an instance field (also called member). This field will allow direct access to the value from inside the class.

Note that I said "inside the class". Because it is private, it is not accessible from outside the class. This is important to preserve encapsulation, a cornerstone of OOP. Encapsulation basically tells us that instance members can't be accessed directly outside the class.

For this reason we make the member private and provide methods that "set" and "get" the variable (at least: in Java this is the way). These methods are exposed to the outside world and force whoever is using your class to go trough your methods instead of accessing your variable directly.

It should be noted that you also want to use your methods/properties when you're inside the current class. Each time you don't, you risk bypassing validation rules. Play it safe and always use the methods instead of the backing field.

The netto result from this is that you can force your logic to be applied to changes (set) or retrieval (get). The best example is validation: by forcing people to use your method, your validation logic will be applied before (possibly) setting a field to a new value.

public int Test {get; set;}

This is an automatically implemented property. A property is crudely spoken an easier way of using get/set methods.

Behind the scenes, your code translates to

private int _somevariableyoudontknow;

public void setTest(int t){
    this._somevariableyoudontknow = t;
}

public int getTest(){
    return this._somevariableyoudontknow;
}

So it is really very much alike to getters and setters. What's so nice about properties is that you can define on one line the things you'd do in 7 lines, while still maintaining all the possibilities from explicit getters and setters.

Where is my validation logic, you ask?
In order to add validation logic, you have to create a custom implemented property.

The syntax looks like this:

private int _iChoseThisName;

public int Test {
   get {
       return _iChoseThisName;
   }

   set {
       if(value > 5) { return _iChoseThisName; }
       throw new ArgumentException("Value must be over 5!");
   }
}

Basically all we did was provide an implementation for your get and set. Notice the value keyword!

Properties can be used as such:

var result = SomeClass.Test; // returns the value from the 'Test' property
SomeClass.Test = 10; // sets the value of the 'Test' property

Last small note: just because you have a property named Test, does not mean the backing variable is named test or _test. The compiler will generate a variablename for you that serves as the backing field in a manner that you will never have duplication.

XML Parsing

If you want your second answer answered, you're going to have to show how your current architecture looks.

It shouldn't be necessary though: it makes most sense to pass it as a parameter with your constructor. You should just create a new XmlParser (random name) object for each file you want to parse. Once you're parsing, you don't want to change the file location.

If you do want this: create a method that does the parsing and let it take the filename as a parameter, that way you still keep it in one call.

You don't want to create a property for the simple reason that you might forget to both set the property and call the parse method.

Answer 3

I can possibly answer your first question. You asked "I have to use this property inside the class." That sounds to me like you need to use your private variable. The public method which you provided I believe will only do two things: Allow a client to set one of your private variables, or to allow a client to "see" (get) the private variable. But if you want to "use this property inside the class", the private variable is the one that should be your focus while working with the data within the class. Happy holidays :)

Answer 4

The following is my personal opinion based on my personal experience in various programming languages. I do not think that best practices are necessarily static for all projects.

---

When to use getters, when to use private instance variables directly

it depends.

You probably know that, but let's talk about why we usually want getters and setters instead of public instance variables: it allows us to aquire the full power of OOP.

While an instance variable is just some dump piece of memory (the amount of dumbness surely depends on the language you're working in), a getter is not bound to a specific memory location. The getter allows childs in the OOP hirarchy to override the behaviour of the "instance variable" without being bound to it. Thus, if you have an interface with various implementations, some may use ab instance variable, while others may use IO to fetch data from the network, calculate it from other values, etc.

Thus, getters do not necessarily return the instance variable (in some languages this is more complicated, such as c++ with the virtual keyword, but I'll try to be language-independent here).

Why is that related to the inner class behaviour? If you have a class with a non-final getter, the getter and the inner variable may return different values. Thus, if you need to be sure it is the inner value, use it directly. If you, however, rely on the "real" value, always use the getter.

If the getter is final or the language enforces the getter to be equal (and this case is way more common than the first case), I personally prefer accessing the private field directly; this makes code easy to read (imho) and does not yield any performance penalty (does not apply to all languages).

---

When to use parameters, when to use instance variables/properties

use parameters whereever possible.

Never use instance variables or properties as parameters. A method should be as self-contained as possible. In the example you stated, the parameterized version is way better imo.

Intance variables (with getters or not) are properties of the instance. As they are part of the instance, they should be logically bound to it.

Have a look at your example. If you hear the word XMLParser, what do you think about it? Do you think that a parser can only parse a single file it is bound to? Or do you think that a parser can parse any files? I tend to the last one (additionally, using an instance variable would additionally kill thread-safety).

Another example: You wish to create an XMLArchiver, taking multiple xml documents into a single archive. When implementing, you'd have the filename as a parameter of the constructor maybe opening an outputstream towards the file and storing a reference to it as an instance variable. Then, you'd call archiver.add(stuff-to-add) multiple times. As you see, the file (thus, the filename) is naturally bound to the XMLArchiver instance, not to the method adding files to it.