Newbie Syntax Question (Start of subroutine)

I'm sure I should know this, but I can't remember and can't find the answer. I'm well used to Java and Perl, which, of course, have quite a different syntax than C++. I'm embarrassed that I have to ask this.

I'm reading a tutorial on using QT from C++ (here: Qt4 tutorial), and I see this in almost every example. Here's some code:

Code:

---

Absolute::Absolute(QWidget *parent)    : QWidget(parent)
{
  QTextEdit *edit = new QTextEdit(this);
  edit->setGeometry(5, 5, 200, 150);
}

---

I know about classes and the double colon, but after that, what is the single colon and the text after that? (": QWidget(parent)")

(I'm not asking about QT at all - just the notation and the purpose.)

I assume you understand what you have there is a definition (some people call it "implementation", but that's the wrong word) of a constructor for a class named Absolute.

The bit after the single colon ("QWidget(parent)" is called an initialiser list (or initializer to Amer-icans). From the argument list of the constructor, we know that QWidget is the name of a user-defined type. The fact that a type name (QWidget) is mentioned directly in an initialiser list says that it is a base class of Absolute. The "QTWidget(parent)" means that the value of parent supplied as an argument to the constructor is used to initialise the base class.

So, if we have

Code:

---

    QWidget *parent_widget = some_function_that_creates_a_widget();

    Absolute ab(parent_widget);

---

the constructor you described will be invoked. The QWidget base class of Absolute will be initialised (i.e. its constructor invoked) using the pointer parent_widget. Then the body of the constructor (starting at line 2) will be executed. It happens in that order because the standard says so .....


Initialisers lists aren't only about initialising base classes though. The can (if a class has them) be used to initialise members of that class, as well as any bases.

Okay, so the single colon is similar to "extends" in Java? (I know I'm leaving out a lot of details - but is that the general quick summary?)

Quote:

---

Originally Posted by TangoOversway

so the single colon is similar to "extends" in Java?

---

No, not in this context. This is a constructor definition, so as grumpy explained the single colon indicates that the follwing is the initializer list to initialise the base class sub-object(s) and member variables of the object.

In the context of a class definition, then yes, a single colon (optionally followed by an access specifier then) followed by a class name is similiar to "extends" in Java.

You realize that you can create an object without using new, right?
So instead of

QTextEdit *edit = new QTextEdit(this);
edit->setGeometry(5, 5, 200, 150);

You can do

QTextEdit edit(this);
edit.setGeometry(5, 5, 200, 150);

You also realize that if you use new, then you must use delete to free it later, right (or use a smart pointer)?

Quote:

---

Originally Posted by Elysia

You also realize that if you use new, then you must use delete to free it later, right (or use a smart pointer)?

---

You don't realize that in QT windows destroy their children automatically, right?

Quote:

---

Originally Posted by kmdv

You don't realize that in QT windows destroy their children automatically, right?

---

Of course I don't, unless the interface uses a smart pointer of some sort. How am I supposed to know?

Quote:

---

Originally Posted by Elysia

Of course I don't, unless the interface uses a smart pointer of some sort. How am I supposed to know?

---

it's a fairly common trait of GUI frameworks, including GTK+, wxWidgets, and the one we have here: Qt. I suspect that you're not especially familiar with Qt, and as such, you probably aren't aware of its behavior with respect to resource management. Qt doesn't play well with stack objects, so new is more or less required.

O_o

(Elaborating on the style of ownership seen in such libraries.)

It isn't really using a smart pointer, but the ownership of the object is given to the parent in exactly the same way you would give ownership of an object to a smart pointer.

The practical upshot is that destroying a parent destroys children in a hierarchical fashion so you, as the client, don't need to "unhook" layers of ownership.

Soma

Code:

---

class Base
{
    // ...
    virtual Base(Base * fParent = 0)
    {
        // fParent->registerChild(this)
    }
    // ...
    virtual ~Base()
    {
        // for each child in Children
        // delete child
    }
    // ...
    void registerChild(Base * fChild)
    {
        // add Child to Children list
    }
    // ...
    List<Base *> mChildren;
    // ...
};

---

I'm not saying it's wrong. I'm saying it's not self-documenting.
A raw pointer could mean anything:
- Is it because a reference is not possible/practical in this case, and hence we just want a pointer to it (and not store it for forward reference)?
- Is it because we must store a copy for future reference?
-- If so, are we (the library) responsible for deleting the pointer?
-- Or is it the programmer who created it who is responsible?

What makes matters worse is that this is often not documented in the API reference. Sometimes it tends to be documented in "how to use" the library, though.

Quote:

---

Originally Posted by Elysia

I'm not saying it's wrong. I'm saying it's not self-documenting.

What makes matters worse is that this is often not documented in the API reference. Sometimes it tends to be documented in "how to use" the library, though.

---

top page -> QT Object model -> object hierarchy

which reads as follows in the first paragraph:

Quote:

---

QObjects organize themselves in object trees. When you create a QObject with another object as parent, it's added to the parent's children() list, and is deleted when the parent is. It turns out that this approach fits the needs of GUI objects very well. For example, a QShortcut (keyboard shortcut) is a child of the relevant window, so when the user closes that window, the shorcut is deleted too.

---

What is your point? I know it documents the behaviour somewhere, but that wasn't the point.

Quote:

---

Originally Posted by Elysia

What is your point? I know it documents the behaviour somewhere, but that wasn't the point.

---

I put your observation to the test and found information that you didn't have quite easily.

There never was any question as to whether the information could be found easily or not.

I have no interest in debating what your point really was. At worst, your backpedaling now because you don't know how QT really works.