Exact correct size needed

JonB

@mrjj
This comes out view->geometry().size() == event->size() always, when first called and as I drag window smaller/larger.

Playing with resizing now, I see that the smaller I make the window the larger the non-visible area becomes, i.e. I get scrollbars on both vertical and horizontal with the "empty" areas of the bar to scroll into getting larger. So, I make that.... event->size() is too large and gets proportionately larger as the window gets smaller? Which might well correspond to a fixed amount which needs to be subtracted from event->size()? Like some margin or border? Or I might be getting my math confumbulated over this....

P.S.
A useful discovery? If I comment out KeepAspectRatio that does not solve but the scrollbars are slightly different. For example, where it starts out fitting with my numbers with KeepAspectRatio I actually get scrollers with a bit missing when not. That implies to me that we have some issue about the QPixmap/scaling to take into account??

Pl45m4

@JonB

So you don't want scrollbars?! Try a fixed scene (that fits in your view and widget)

https://doc.qt.io/qt-5/qgraphicsscene.html#sceneRect-prop

Then resize your pixmap according to your window/view

JonB

@Pl45m4
I don't think I understand you at all! Or, I'm already doing that?

I do not want scrollbars, but only in the sense that they should not be needed because everything fits precisely, not that there is to be anything out of view.

The actual pixmap turns out to be 6057 x 3824. I want it to always fit in the gfx view perfectly (scaled, keeping ratio), i.e. nothing scrolled out of view, as I resize the window. The window is supposed to be a widget, which is totally filled by the one QGraphicsView, which in turn is totally filled by the one QGraphicsPixmapItem.

You can see I am acting on the window widget's resizeEvent, and rescaling the pixmap (with keep ratio), which is what I understood I need to do, perhaps from reading around.

But, as I said, I am new to Qt graphics so are you referring me to something I do not understand/know is available?

Pl45m4

@JonB

From my experience, you do get scrollBars as soon as your content is close to the border of your scene.
And because you don't seem to set any (fixed) sceneRect, you always will. (AFAIK, 1 or 2 px difference is still enough to trigger the scrollBars)

From sceneRect doc:
"If unset, or if set to a null QRectF, sceneRect() will return the largest bounding rect of all items on the scene since the scene was created (i.e., a rectangle that grows when items are added to or moved in the scene, but never shrinks)."

-> Because your sceneRect never shrinks, your non-visible area will grow when you make your widget bigger and then decrease the size again. (Your Pixmap shrinks, but your empty, blank scene doesn't. This is why you have scrollBars)

So I don't know if this will help you, but I would try to resize the scene with your widget as well.

JonB

@Pl45m4 said in Exact correct size needed:

AFAIK, 1 or 2 px difference is still enough to trigger the scrollBars

Indeed. But while it starts out like that, when the window is "large", as I make the window "small" the unscrolled area increases to like half the total width or height. And that unscrolled area is just empty space, it's not my pixmap that fills it, the pixmap looks right and fits. We are not talking about "a few pixels".

-> Because your sceneRect never shrinks, your non-visible area will grow when you make your widget bigger and then decrease the size again.

So I don't know if this will help you, but I would try to resize the scene with your widget as well.

I'm looking into this. I know about the "sceneRect never shrinks". Yep, I'm thinking about this now, I do see what you mean. It won't solve why it's not quite right initially --- though, as you say, that might be the "few pixels accuracy" --- but that would make sense that any shrink would then have scrollbars.... Lemme have a play :)

JonB

@Pl45m4
Looks like I owe you a beer :)

/*virtual*/ void MapWindow::resizeEvent(QResizeEvent *event) /*override*/
{
    const QSize &size(event->size());
    // to avoid scrollbars, scene must be resized, and needs to be "a little bit" smaller
    int width = size.width() - 2;
    int height = size.height() - 2;
    scene->setSceneRect(0, 0, width, height);
    // scale the pixmap to fit, keeping aspect ratio
    const QPixmap &pixmap(pixmapOriginalMap->scaled(width, height, Qt::KeepAspectRatio));
    pixmapMap->setPixmap(pixmap);
}

This works much better! I was confusing the pixel issue and the need to resize scene smaller on shrink.

Those - 2s are magic :) Without I get scrollbar. There should be a Qt constant for this :)

ollarch

Hi,

You have to understand what is the scene and what is the view. In your case, the scene is in pixel units. What the view does is show the scene (or part). The view can show the pixmapItem zoomed but the item size in scene units is still the same.

So, as I posted on the other topic:

Add the pixmapItem to the scene.
Call "fitInView(pixmapItem,Qt::KeepAspectRatio)
Override the "resizeEvent" of your Main Widget or your Main Window and call the "fitInView" there.
If you want to get the mouse position: create a derived QGraphcisView, and you can do something like this:

void MyGraphicsView::mouseMoveEvent(QMouseEvent *pqEvent)
{
	QGraphicsView::mouseMoveEvent(pqEvent);
	QPointF qScenePos = mapToScene(pqEvent->pos());
	qint32 qX = qScenePos.x();
	qint32 qY = qScenePos.y();
	emit position(qX,qY);
}

This takes the GraphicsView position and maps it into the scene coordinates, in this case your image pixel coordinate.
Note that you have to use the MyGraphicsView class instead of QGraphicsView class. Also I have added the SIGNAL "position"

JonB

@ollarch
Thank you for joining this discussion!

QGraphcisView::fitInView(), yes I remembered that vaguely.

I have questions about what you are saying:

I looked through articles and they said for best quality scaling of pixmaps I should do it via the QPixmap::scaled() I am using [I am struggling to find references for this now...]. Isn't your fitInView(pixmapItem,Qt::KeepAspectRatio) way going to "lose me quality"?
You are doing it by resizing the view. Yet I am not doing that, I am resizing the whole scene to fir the pixmap. As I think advised by @Pl45m4. What I have works (though there are those - 2s...). Doesn't it? is it better to do your way via view instead of scene?

ollarch

Hi,

The view is the one that takes care of the display part. What it does is show the scene with the indicated transformations. So, let the view do its job.
The view is not resizing, it only paints the scene and does not modify the scene. So, as I told you, the pixmap is still the same, what have changed is where are you looking it.

Think on a CAD software. The schematic have to keep the units regardless of the zoom applied or the rotation of the view. You can rotate the view, zoom it, ... but when you ask the view to map the mouse position into scene coordinates, this coordinates will still be the same pixel position.

Start with points 1 and 2. When you have it working, do the point 3. Finally point 4.
When you finish all this steps, call "rotate(90)" on your view and see what happens (what you see and the mouse positions).

JonB

@ollarch
Thank you. I will save what I have now and rewrite your way tomorrow, and report back :)

JonB

@ollarch
Firstly, I do understand what you are saying/your approach. I had been expecting to have to take the current scaling into account and do my own conversion of mouse coords to scene/pixmap coords. I agree that your way means the QGraphicsView will do this for me, and that makes me a happy bunny :)

I followed your instructions above exactly. It almost worked, but there is a little wrinkle!

When the window is first shown the pixmap is "small", does not fit whole area --- even though I had fitToView() in the constructor (where I add the pixmap) and a resizeEvent is generated. Once I then start interactive window resizing it jumps to correct, full size, and maintains that from then on correctly.

[Also your way I did not have to put in any "magic" - 2 on sizes, it never caused scrollbars to appear (even when I still had them allowed), so no need for "small pixel adjustment, which is an added plus point :) ]

I tried "everything" (I'm pretty thorough!) but nothing remedied this. So on a hunch I put in a showEvent() override to do the fitToView() there, and sure enough that, and only that, made it also be correct when first shown.

The essentials of my now-working code are:

void MapWindow::setupUi()
{
    // as per https://doc.qt.io/qt-5/qgraphicsview.html#fitInView
    // since we intend to always call `QGraphicsView::fitToView()` in `resizeEvent()`
    // we explicitly disable scrollbars to *ensure* no infinite recursion
    view->setHorizontalScrollBarPolicy(Qt::ScrollBarAlwaysOff);
    view->setVerticalScrollBarPolicy(Qt::ScrollBarAlwaysOff);
    // create the map as a `QGraphicsPixmapItem`
    // `pixmapMap` will be fitted tightly in view in `showEvent()`/`resizeEvent()`
    this->pixmapMap = new QGraphicsPixmapItem(*pixmapOriginalMap);
    scene->addItem(pixmapMap);
}

/*virtual*/ void MapWindow::showEvent(QShowEvent *showEvent) /*override*/
{
    // call the base method
    QWidget::showEvent(showEvent);

    view->fitInView(pixmapMap, Qt::KeepAspectRatio);
}

/*virtual*/ void MapWindow::resizeEvent(QResizeEvent *event) /*override*/
{
    // call the base method
    QWidget::resizeEvent(event);

    view->fitInView(pixmapMap, Qt::KeepAspectRatio);
}

[I am unsure about the need to call QWidget::showEvent/resizeEvent() base class, and where, in my overrides. Some examples do, some don't. With/without, or moving/before after the fitInView() call, all make no difference.]

I don't know whether you have any comment/alternative suggestion about this need to do fitView() in showEvent(), but I couldn't get it to work first time without?

ollarch

@JonB said in Exact correct size needed:

When the window is first shown the pixmap is "small", does not fit whole area

This is because when the widget constructor is executed, the widget size is not the "final" widget size. After the constructor is called you call "show" and then it calculates the size of the ui elements.

[I am unsure about the need to call QWidget::showEvent/resizeEvent() base class

You don't need to call it.

Can you check the values on the debugger of "event->s" and "event->olds" ? The first one is the new size and the last is the old size.

A workarround for this could be to use a "QTimer::singleshot(100,this,&YourWindowClass::fitMyPixmapToView)" into constructor.
, but I have not need to use it.

JonB

@ollarch said in Exact correct size needed:

This is because when the widget constructor is executed, the widget size is not the "final" widget size.

Yep, I do realize this :)

After the constructor is called you call "show" and then it calculates the size of the ui elements.

It doesn't matter when I call show(), I find I have to do the "fit" in showEvent(), i.e. when it's actually shown.

Can you check the values on the debugger of "event->s" and "event->olds" ? The first one is the new size and the last is the old size.

I did that right from the start. In the very first resizeEvent() the old size is always (-1, -1), while the (new) size is always the correct, big one. This is still the case even after I have fitted in showEvent(). So, it looks like I get the first resizeEvent() before the showEvent(), and at that point fitting doesn't work in resizeEvent() but it does in showEvent().

12:33:20: Debugging starts
resizeEvent QSize(-1, -1) QSize(1492, 772)
showEvent
// now start dragging
resizeEvent QSize(1492, 772) QSize(1491, 772)
resizeEvent QSize(1491, 772) QSize(1490, 772)
resizeEvent QSize(1490, 772) QSize(1490, 771)
...

A workarround for this could be to use a "QTimer::singleshot(100,this,&YourWindowClass::fitMyPixmapToView)" into constructor.

Indeed I could have done that. But since it works via showEvent() I'm happy.

What else can I say? :)

ollarch

Hi, I just found that I use to call "showMaximized" on "main.cpp" instead of "show" to make the application full screen.
I've tested with calling "show" and yes, it does not fit the item to the view the first time.

I don't like too much this solution but you can use this

QTimer::singleShot(50, this, SLOT(fit()));

Define "fit" method as slot.

JonB

@ollarch
Thanks, it sounds like (if I understand correctly) you are saying you find the same issue as me first time? Though I haven't seen this commented on elsewhere.

As I said earlier, I have no problem with rectifying via your QTimer, but since I have rectified via showEvent() (as per my code) why are you not satisfied with that approach. any reason?

I have moved on now to mouseworks, and am happy with my coordinates :)

ollarch

As I said earlier, I have no problem with rectifying via your QTimer, but since I have rectified via showEvent() (as per my code) why are you not satisfied with that approach. any reason?

No, I have any problem and I'm satisfied that it works.

I have moved on now to mouseworks, and am happy with my coordinates :)

Now you can try to rotate the view 45 degrees and you will see that when you map the view coordinates to scene coordinates you still have the correct pixel coordinate.

JonB

@ollarch
Indeed, though rotation is not something I will want to do, but I take your point.

Overall I am very happy with the approach you stuck with and made me adopt for my situation. As I said, I thought I was going to have to do my own coordinate transformation arithmetic taking into account my current scale/zoom setting etc. I am much happier that all I have to do is call Qt's already-written mapToScene() to do it for me, and everything makes logical, "purist" sense to me.

So thank you for sticking to your guns and recommending the approach I was really looking for!

sainivedant41

Thanks for the clarification. mobdro

ollarch

Overall I am very happy with the approach you stuck with and made me adopt for my situation. As I said, I thought I was going to have to do my own coordinate transformation arithmetic taking into account my current scale/zoom setting etc. I am much happier that all I have to do is call Qt's already-written mapToScene() to do it for me, and everything makes logical, "purist" sense to me.

I'd rather not reinvent the wheel.

So thank you for sticking to your guns and recommending the approach I was really looking for!

I'm happy. ;)

JonB

@ollarch
Hi @ollarch , hope you're around and see this question (or anyone else who knows!)....

I am moving onto my requirement to draw a shape on the view to follow the mouse. It will be see-through but with a frame, so that user can see where on map the mouse is.

For now I will do that via a QAbstractGraphicsShapeItem on the QGrapcicsScene, because that's the only way I know how to do it; plus it will use scene coordinates, which is easy.

However, I am aware this is not "correct". There should be no permanent object on the scene for this. And, for example, although I do not have multiple views onto the scene, if I did it would show up in all views. Which is not correct. This shape is a temporary object which belongs/should be shown only in the view where the mouse is moved, not on the scene shared across views.

The "correct" thing to is to draw this on the GraphicsView only (with necessary coordinate calculations, but that's not the point). I came across QRubberBand which looks like the sort of thing which should be used (and I think that can be put on the view only). But unfortunately for me that only allows a line or a rectangle, and I will want a circle or a hexagon for my usage.

So.... what is the right way to do this, please? :)

EDIT
I have now come across, say, https://www.qtcentre.org/threads/4479-Adding-Rectangular-overlay-on-graphics-view [from 2006!], which is asking just my kind of question:

To simplify my problem, lets assume that we have two different graphics views. Both of these views have the same scene applied to it. When the user clicks on a specific point in the first graphics view I want to draw a rectangle related to the point she clicked, but I want it to only be visible in that graphics view.

I think this ends up saying to use QGraphicsView::drawForeground(QPainter *painter, const QRectF &rect) to achieve this. Is that the route I should be following? [And if it is I shall have further questions! Because the principle there is not good enough as-is for following the mouse and redrawing when it moves.]

UPDATE
Now that I am beginning to understand what facilities the graphics scene/view offer. I am now sailing along, using the foreground layer of the view via QGraphicsView::drawForeground() to draw my mouse-follow-shape, calling scene()->invalidate(QRectF(), QGraphicsScene::ForegroundLayer); to cause it to redraw from mouseMoveEvent(). I think I have got this right!

Which now makes me think: maybe I should be drawing the map on which all this happens via QGraphicsScene::drawBackground() instead of in a QGraphicsPixmapItem() on the scene? For one thing, I believe the scene's background layer is cached for speed by default, which sounds good for an unchanging map background?