[Qt] Consider the scale of the page view when scrolling QTouchWebView.

[Qt] Consider the scale of the page view when scrolling QTouchWebView.

Created an attachment (id=100829) [details]
Patch

(From update of attachment 100829 [details])
View in context: https://bugs.webkit.org/attachment.cgi?id=100829&action=review

> Source/WebKit2/UIProcess/API/qt/qtouchwebview.cpp:42
> +    pageView->moveBy(deltaX * pageView->scale(), deltaY * pageView->scale());

I think TouchViewInterface::panGestureRequestScroll() should use m_viewportView->mapFromItem() and send the coordinate in the right space to the viewport.

QTouchWebViewPrivate::scroll(qreal deltaX, qreal deltaY) should be in viewport coordinate.

Created an attachment (id=100993) [details]
Patch

Use item mapping functions instead.

(In reply to comment #2)
> I think TouchViewInterface::panGestureRequestScroll() should use m_viewportView->mapFromItem() and send the coordinate in the right space to the viewport.
> 
> QTouchWebViewPrivate::scroll(qreal deltaX, qreal deltaY) should be in viewport coordinate.

Humm didn't see your comment, since the method is on the viewport it would make sense to give it viewport coordinates. Cooking a new patch.

Created an attachment (id=101001) [details]
Patch

Move the mapping to TouchViewInterface

(From update of attachment 101001 [details])
View in context: https://bugs.webkit.org/attachment.cgi?id=101001&action=review

> Source/WebKit2/UIProcess/qt/TouchViewInterface.cpp:51
> +    // This assumes that m_viewportView's and m_pageView parent's coordinate systems match.
> +    QPointF destInViewportCoords = m_pageView->mapToParent(m_pageView->mapFromParent(m_pageView->pos()) + QPointF(deltaX, deltaY));
> +    QPointF offsetInViewportCoords = destInViewportCoords - m_pageView->pos();

I don't understand why you do all this.
Isn't it 
QPointF destInViewportCoords = m_viewportView->mapFromItem(m_pageView, QPointF(deltaX, deltaY)); ?

The 2 delta is like a vector, you can just multiply it by the combined transformation matrix of a subtree to get the transformed vector.  Or am I missing something?

(In reply to comment #6)
> I don't understand why you do all this.
> Isn't it 
> QPointF destInViewportCoords = m_viewportView->mapFromItem(m_pageView, QPointF(deltaX, deltaY)); ?
> 
> The 2 delta is like a vector, you can just multiply it by the combined transformation matrix of a subtree to get the transformed vector.  Or am I missing something?

I tried but it's not enough when there is also a translation transform. i.e. a (1,2) delta is mapped to (11,12) if there is a (10,10) translation on top.

(From update of attachment 101001 [details])
View in context: https://bugs.webkit.org/attachment.cgi?id=101001&action=review

Isn't m_pageView->mapFromParent(m_pageView->pos()) always gonna be zero though?
In page view coordinate, its own position would be the top left corner.

> Source/WebKit2/UIProcess/qt/TouchViewInterface.cpp:49
> +    // This assumes that m_viewportView's and m_pageView parent's coordinate systems match.

Personnally I would remove that comment and remove the assumption by using mapFromItem instead of mapFromParent :)

(In reply to comment #8)
> (From update of attachment 101001 [details] [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=101001&action=review
> 
> Isn't m_pageView->mapFromParent(m_pageView->pos()) always gonna be zero though?
> In page view coordinate, its own position would be the top left corner.
> 
That's a good point, and along that your code seems right in comment 6, but what you get is a vector from the viewport's origin toward the applied vector on the page view's origin. I didn't realize that this is the final destination in viewport coordinate and was trying to apply it as a vector again by sending it as the delta values.

> > Source/WebKit2/UIProcess/qt/TouchViewInterface.cpp:49
> > +    // This assumes that m_viewportView's and m_pageView parent's coordinate systems match.
> 
> Personnally I would remove that comment and remove the assumption by using mapFromItem instead of mapFromParent :)

That sounds like a good idea too, would also help fixing my assumptions instead of spreading them around :)

Created an attachment (id=101149) [details]
Patch

New patch using mapFromItem instead of mapToParent.
I tried assuming that mapFromParent(pos()) and remove the call, but the assumption wouldn't hold when scaling the item, event though the documentation says it should (it might not help that we set a different transform origin on that item.).

(From update of attachment 101149 [details])
Great.

Committed r91183: <http://trac.webkit.org/changeset/91183>