Analysing simple codesnips

dheerendra

I'm not very clear for which one I need to answer. Let me try to answer.

QWidget is not a rectangle. It is visual elements whose size is defined with rectangle.
Since you are writing the code in main.cpp you don't see much difference. Use some method and use QScopedPointer. You will see the difference.

tomy

@dheerendra
The first two texts in italic are quotes from Docs which seemingly are unlike what you said.
The next two ones were questions that you sent replies as answers.
Thanks. But I haven't understood that simple code completely yet.

mrjj

Hi
Lets first talk about QScopedPointer.
its used to auto delete a class (one has newed) automatically as soon as the QScopedPointer
runs out of scope. Scope is often the function its located in. So scope ends at last }
However, this example is not really good as there is no need to actually new the widget since
app.exec(); is blocking and first ends when app ends.

Also, often a widget is assigned a parent which will then delete it when parent is deleted. (ownership).

Also, if you did this in a function, you would never see the actual widget as it would be deleted as soon as function ends.
So QScopedPointer is more often used with own data class that is not widgets.
It would make more sense to do
int main(int argc, char** argv)
{
QApplication app(argc, argv);
CustomWidget widget;
widget.resize(600, 400);
widget.show();
return app.exec();
}

For your other questions.

paintEvent is an overwritten method here for repainting part or all of its widget which is initially left blank
This virtual function is called by the framework and its expected to paint the Widget in its current state.
For QWidget is not much code, but say for QPushButton it has code to draw the button.
QPainter acts as a brush to to draw/fill something
Its the class that can do drawing onto the screen. It contains tons of function to draw
various shapes on screen.
rect().adjusted adds a new rectangle on (?)
Yes it returns a new QRect by copy with its value adjusted.
Note, there is also Adjust function which does alter the QRect in place.
I imagine QWidget is a rectangle we using resize make is smaller.
The Geometry of any widget is a rectangle. While you can use a mask to paint say a round
window, any hit testing, like mouse clicks still operates on the rectangular definition.
So yes, you can say all widgets are rectangular.
In painEvent if we move this statement painter.fillRect(r2, QColor("#FFBB33")); up before if, the program's behaviour will astonishingly be completely different! (Why)

well the if will alter r2 always.
if you move before if, its will draw a rect that is 10 less on all sides that full rect.
the if will move the rect to center around a point. that part is missing if moved before if.

[edit: fixed typo code SGaist]

tomy

@mrjj

Hi
Lets first talk about QScopedPointer.
its used to auto delete a class (one has newed) automatically as soon as the QScopedPointer
runs out of scope. Scope is often the function its located in. So scope ends at last }
However, this example is not really good as there is no need to actually new the widget since
app.exec(); is blocking and first ends when app ends.

Also, often a widget is assigned a parent which will then delete it when parent is deleted. (ownership).

Also, if you did this in a function, you would never see the actual widget as it would be deleted as soon as function ends.
So QScopedPointer is more often used with own data class that is not widgets.
It would make more sense to do
int main(int argc, char** argv)
{
QApplication app(argc, argv);
CustomWidget widget;
widget.resize(600, 400);
widget.show();
return app.exec();
}

Hi,

I, too, thought that way and it was the real rationale I asked that question about QScopePointer because it has no difference with that description you described to the commented replacement.

This virtual function is called by the framework and its expected to paint the Widget in its current state.
For QWidget is not much code, but say for QPushButton it has code to draw the button.

Now what is/are its painting(s) in the code? I know rect2 is filled but rect1 is only outlined. And we didn't need to use that virtual function, we could use something like below:

void CustomWidget::painting() {
...
}

Because it's QPainter which does paint.

Its the class that can do drawing onto the screen. It contains tons of function to draw
various shapes on screen.

Yeah, thanks.
I also found something else in mind. Can it draw a circle a paint it?

Yes it returns a new QRect by copy with its value adjusted.

What do you mean by that, please? For example, I think it creates a new over a prior rect. But where is the prior one?

well the if will alter r2 always.
if you move before if, its will draw a rect that is 10 less on all sides that full rect.
the if will move the rect to center around a point. that part is missing if moved before if.

I think you misunderstood my talk. I meant this:

...
    QRect r2(QPoint(0, 0), QSize(40, 40));
    painter.fillRect(r2, QColor("#FFBB33"));

    if(m_lastPos.isNull())
        r2.moveCenter(r1.center()); 
    else
        r2.moveCenter(m_lastPos);    
}

painter.fillRect(r2, QColor("#FFBB33"));'s act is only painting rect2; it musn't have anything to do with moveCenter.

mrjj

Hi
You are only allowed to use QPainter in the paintEvent function.
You cannot use other name as Qt would not call your paint function and nothing would be drawn.
you could do


void CustomWidget::paintEvent(QPaintEvent *)
{
  QPainter p(this); 
  painting(&p);

but
void CustomWidget::painting() alone would not work.

and yes it can draw a circle
painter.drawEllipse(QPointF(x,y), radius, radius);
and fill it with QBrush.

I think it creates a new over a prior rect. But where is the prior one?
QRect r1 = rect().adjusted(10, 10, -10, -10);
the rect function returns a QRect and calling adjusted returns a new one with altered values.
so the prior one is rect();

When you do like this

 QRect r2(QPoint(0, 0), QSize(40, 40));
    painter.fillRect(r2, QColor("#FFBB33"));

    if(m_lastPos.isNull())
        r2.moveCenter(r1.center()); 
    else
        r2.moveCenter(m_lastPos);    
}

you paint the rect BEFORE its alterd. the next moveCenter have no effect as its already painted.
Hence it looks different as r2 is no longer changed.

tomy

@mrjj Hi, and thanks.

Qt would not call your paint function

Does it mean that those three virtual functions in the code are automatically called by the event loop? It might be true and I just realized that because there is no statement in the code to call them! (We just create an object of the class)

QRect r1 = rect().adjusted(10, 10, -10, -10);
the rect function returns a QRect and calling adjusted returns a new one with altered values.
so the prior one is rect();

So, there's a rectangle created by rect() with a reasonable size, then using adjusted we narrow it by 10 points from both up-left and down-right corners. Is it right?

If right, it sounds redundant, because we could create a rect with the desired size without need to adjusting that. Couldn't we?

ou paint the rect BEFORE its alterd

My problem is with that "altered", I say it's moved and it's not changed/altered.
For example, you take a white vase to the living room then there paint it. Another day, you paint another white vase and take it to the living room.

mrjj

Hi

Does it mean that those three virtual functions in the code are automatically called by the event loop? It might be true and I just realized that > because there is no statement in the code to call them! (We just create an object of the class)

yes the paintEvent ( among many others) are called by the Qt framework and its part of the system that makes it all work.

So, there's a rectangle created by rect() with a reasonable size, then using adjusted we narrow it by 10 points from both up-left and down-right corners. Is it right? If right, it sounds redundant, because we could create a rect with the desired size without need to adjusting that. Couldn't we?

Yes you could also create the r1 rect without using "adjusted" but would be more code for same result if you want to
have a rect that is 10 less that whole widgets rect. but you can set the x,y and width, height of a QRect in many ways.
This is just one way.

My problem is with that "altered", I say it's moved and it's not changed/altered.
In any case the the x,y of the r2 is altered and hence its painted differently.
Its centers around a point. if painted before moveCenter. its not centered.

tomy

@mrjj Hi and thanks so much for your help.

( among many others)

So here in this code, only paintEvent is called by Qt automatically, and not other two. mousePressevent/mouseMoveevent is invoked when the mouse is pressed/moved. Yeah?

be more code for same result if you want to
have a rect that is 10 less that whole widgets rect

So the rect() function here, knows the size of the application/widget's window and creates a new but rather smaller one over that. Right?

In any case the the x,y of the r2 is altered and hence its painted differently.
Its centers around a point. if painted before moveCenter. its not centered.

You say that when the coordinates of r2 are changed it's just as it itself is changed, hence it's altered. OK I got it. But why if it's painted before moveCenter it's not centered? Why should a movement task has that thing to do with the color of an object to act that differently?

mrjj

Hi

So here in this code, only paintEvent is called by Qt automatically, and not other two. mousePressevent/mouseMoveevent is invoked when - the mouse is pressed/moved. Yeah?

Well Qt calls paintEvent when a widgets needs to be painted. Qt calls the (also virtual ) function for mouse events if
a mouse event come from the OS and the mouse cursor is over the widget.

So the rect() function here, knows the size of the application/widget's window and creates a new but rather smaller one over that. Right?
the rect() function returns a widgets area as a QRect. and adjusted will then read values and return a new QRect. (by copy)
You say that when the coordinates of r2 are changed it's just as it itself is changed, hence it's altered. OK I got it. But why if it's painted before -moveCenter it's not centered? Why should a movement task has that thing to do with the color of an object to act that differently?
before moveCenter, its values are 0,0 to 40,40. so if while widget rect() is 0,0, 500,500 , it would paint up in the corner.
After move center its x,y is different. ( depending on the point its been centered around)
it would alter the x,y of the Rect so the point given would be center of the rect. and hence, its positions(x,y) are changed.

tomy

@mrjj

it would alter the x,y of the Rect so the point given would be center of the rect. and hence, its positions(x,y) are changed.

I still can't comprehend your perspective!
We have a rectangle named r2. Its position firstly is (0,0), (40,40). The function painter.fillRect(r2, QColor("#FFBB33")); fills r2 using its name not its new position. So it's firstly one the top-left corner and painted, then it goes somewhere else on the area. So as an independent widget, it must keep its colour wherever it is.

mrjj

@tomy
well
painter.fillRect(r2, QColor("#FFBB33"));
would indeed be same as
painter.fillRect(QRect(0,0,40,40), QColor("#FFBB33"));
and would paint in corner. and never move.

however, if you modify the r2 BEFORE painting with it, (with r2.moveCenter(r1.center()); )
the r2 rect would be changed. and paint in other location.

But im not sure what you doubts still are ?

if you omit moveCenter, r2 is unchanged.
If you apply moveCenter, its changed.
if you paint with r2, BEFORE you change it,
nothing will happen with moveCenter as the fillRect was already run.

tomy

@mrjj
Got it, thank you.

Each time the mouse is clicked, all the statements of the paintEvent method are run, while only the if condition and painter are needed to be re-run each time.
The code looks raw or, as Stroustrup says, ugly. Is it not?

mrjj

@tomy
Hi
Well its not for sure a mousePress event will cause a widget to repaint.
But for say for a QPushButton it will so button can draw as pressed.
Its pretty basic painting with one color rectangle.
Much widgets have much more complicated paint code.

tomy

@mrjj

Well its not for sure a mousePress event will cause a widget to repaint.

For example, we click on different areas of the widget for five times and the r2 rectangle traces us, how many times will the method paintEvent be called?

If five times, so in each call, all the statements inside the method are called. Isn't it?

mrjj

@tomy
Hi
if the r2 follows it does mean for each click, the paint is called.
i assume there is an update() in mousePressEvent/release.
yes all code in painteEvent is run each time.

tomy

@mrjj
Hi, good morning. :)

Yes, r2 follows the click presses and there's an update() which makes it happen by re-calling painEvent each time.
My assumption was that, the part of paintEvent below is useful only for the first time when the program runs.

QPainter painter(this);
   QRect r1 = rect().adjusted(10, 10, -10, -10);
   painter.setPen(QColor("#33B5E5"));
   painter.drawRect(r1);
   QRect r2(QPoint(0, 0), QSize(40, 40));

Afterwards when the user clicks different locations on the widget, only the following section is needed and the above one will be excessive.

if(m_lastPos.isNull())
        r2.moveCenter(r1.center()); 
    else
        r2.moveCenter(m_lastPos);
      painter.fillRect(r2, QColor("#FFBB33"));
}

Disagree?

mrjj

@tomy
Hi and good morning
it doesn't really work that way.
You need all of the paint code each time.

tomy

@mrjj

One reason is that QPainter must be in the following code,

    QPainter painter(this);
    QRect r1 = rect().adjusted(10, 10, -10, -10);
    painter.setPen(QColor("#33B5E5"));
    painter.drawRect(r1);
    QRect r2(QPoint(0, 0), QSize(40, 40));

because it doesn't work somewhere else, as you said.
What remains is the rectangles, r1, r2. Couldn't we define them somewhere else and only use them in that code?

mrjj

@tomy
Hi
yes Painter only works in paintEvent ( you can also draw on pixmap with it as only other case)

You could have r1 and r2 as members of the class but if the window can be resized,
then you want to call rect() each time anyway to make sure to use right size.
So im not sure there would be any benefit to store the rects.
However, color and fonts and images and such things should be stored in class and
not loaded/created each time.

tomy

@mrjj
OK, thanks so much.