This is a follow-up to How to make mouse events propagate to widgets in `scroll` containers?
TL;DR:
The :layout method I implemented makes the widget draw all over the other widgets (see pictures below). How can I constrain where the :layout method will draw my widgets and where it will allow me to interact with the children?
Longer version:
So I ended up patching the scroll container and what I basically did was that I implemented a :layout method, based on the offset calculations that were already made in the original scroll container code.
This is basically what I did (I'll only put the relevant parts here):
-- this function emits signals `self._private.fps` times a second
local _need_scroll_redraw = function(self)
if not self._private.paused and not self._private.scroll_timer then
self._private.scroll_timer = timer.start_new(1 / self._private.fps, function()
self._private.scroll_timer = nil
self:emit_signal("widget::redraw_needed")
self:emit_signal("widget::layout_changed") -- this is the only
-- line that I added
-- to this function
end)
end
end
local function calculate_info(self, context, width, height)
-- this is a longer function, but in summary here we calculate the
-- ideal size that the child would like to be, we see if the child
-- is bigger than the space we have for drawing, and if it is,
-- we calculate offsets (and we call `_need_scroll_redraw` here)
-- so we later know where and how often to `:draw` and `:fit` it
end
function scroll:fit(context, width, height)
local info = calculate_info(self, context, width, height)
return info.fit_width, info.fit_height
end
function scroll:layout(context, width, height)
local result = {}
local i = calculate_info(self, context, width, height)
table.insert(result, base.place_widget_at(
self._private.widget, i.first_x, i.first_y, i.surface_width, i.surface_height
))
return result
end
-- Also, there was a `:draw` method, but I took it out entirely since
-- if I add the `:layout` method, things get drawn just fine
-- P.S: I also tried to implement what was in the `:draw` method, inside
-- the `:layout` method, so that it'll clip properly. I also tried that idea
-- with the `:before_draw_children` and `:after_draw_children` methods
-- but since I don't know how to use cairo, god knows what I wrote there,
-- but it didn't work
With the default scroll widget, my widget looks like this, but nothing I click on works:
But with the changes I made above, the widget with the rows does scroll, and I can click on each child just fine and have it react, it's just that it draws everything outside of its boundaries, and I can also click on stuff that's outside of the boundaries:
So my question is: how would I go about restricting what the :layout method displays, to have it work the way the default scroll layout works, but still be able to interact with the children?
Untested, but I guess:
function scroll:before_draw_children(context, cr, width, height)
cr:rectangle(0, 0, width, height)
cr:clip()
end
This takes the rectangle describing the size of the scroll widget and clips all children to it, meaning that they cannot draw outside of the scroll widget.
Note that this is only prevents the drawing of the child widgets. It does not make awesomeWM "forget about them". What I mean with that is: You can still click on the where the widget would be if it were visible. The click will still be handled by the widget. Only the actual drawing of the child widget is prevented.
The above is why I asked in your original question whether you want just this one widget to be scrolled, without anything next to it. Because if the scroll widget covers the whole wibox, this whole class of issues just goes away.
And no, as far as I know / see, there is no way to prevent the not-drawn widgets from receiving button clicks.
Related
I'm working on a simple widget system, and I'm implementing some containers right now.
Here's the situation I find myself in:
I have a Widget base class, a Container class, which is a widget that can contain other widgets, and several widget sub classes like Button.
I have two types of container: Container itself, which positions children absolutely, and Box, which will stack widgets next to each other, either horizontally or vertically.
Each widget draws itself at x=0, y=0. Therefore, containers need to add an offset to the drawing context before the widgets are told to draw themselves.
Each widget does its own hit testing based on its x/y position.
So far, it works fine. But it falls apart now that I'm implementing Box: What I do is that I overwrite the drawfunction inherited from Container to draw them all in next to each other, instead of based on their x/y position. Quite simple.
But event handling is totally off now, as the widget's x/y position has become meaningless.
I think I have two options:
Have the widget do hit testing at position x=0, y=0, like drawing. Then recalculate the mouse position to match that in Container.
Make each layout set x/y for its children, and make children draw themselves at their x/y position again. No more offsets for the drawing context
The first one is a bit ugly, I think. The second one is pretty complicated to implement, since I need to react to position changes in widgets.
How to other widget systems like Qt, Gtk and wxWidgets generally tackle this? I've looked at the source of some of these, but can't quite figure that out, it's too sophisticated. I don't have any resizing or packing issues to consider.
You are trying to implement your own layout system. You should expect it to be difficult.
I would advise against the first method. The x,y coordinates of a widget are not only used by the widgets themselves, but by anyone outside of the container who wants to do something with the widget.
The second solution is what I've chosen to implement custom widgets made of several smaller widgets and it's not that hard to put together if you don't want too many features.
Just get the widgets when they are added to your container, set their position to the current free spot, and move on to the next.
Basically I've got a QGridLayout with a few widgets in it. The important ones are 2 labels, which I use for drawing images to the screen. Well, if the user wants, he can change the resolution of the incoming images, thus, forcing the Labels to resize.
Let's assume the initial size of the label is 320x240. The user changes the VideoMode to 640x480, the label and the entire GUI resizes perfectly. But when the user switches back to 320x240, the label shrinks, but the Layout/Window does NOT.
I've played around with sizePolicies and sizeHints, and resize(0,0), but nothing did the trick. Could somebody help me with this?
Here some screenshots to clarify the problem:
You need to set the size constraint of the layout holding all your widgets to "SetFixedSize". Although the name doesn't sound like it will work, it ensures that your layout will only use the space it needs. You will not have the problem like you do in your second screenshot.
Example:
mainLayout.setSizeConstraint(QLayout::SetFixedSize);
QLayout::setSizeConstraint(QLayout::SetFixedSize) solves this problem well when you prefer keeping your widget's size fixed at all times--that is, if you'd like it to always be fixed to its "packed" size (which may still vary as the child widgets change size). That is what the "fixed" means there: "fixed" to the correct size, even as the latter varies. (In Qt terms, what I'm calling the "packed" size is simply the widget's sizeHint.)
But a constraint may be too strong a solution in some instances. In particular, if you apply it to a top-level window, then the user will not be free to resize the window. If you don't like that, you can instead perform the "set size to sizeHint" operation instantaneously each time it's needed, rather than imposing it as an unrelenting constraint. The way to do that is to call QWidget::adjustSize().
http://doc.qt.io/qt-5/qwidget.html#adjustSize
Note that if the container whose children are changing size is not the top-level window, then adjustSize() may have to be called recursively on the container and its parents. (In my case I had to do that, anyway. I also tried the size-constraint scheme, and found that applying the constraint at only the topmost level was successful in compacting all levels. I haven't enough knowledge of Qt to comment usefully on these observations, so I merely share them.)
You need to store the original size of your widget parent window before applying any changes to the layout and restore it when the user switches back to the original.
Notice that you need to work with the widget parent window size and not the widget parent size.
in your widget before applying the layout changes:
minimumWindowSize = this->window().size();
when you finished reorganizing the widget to the compact size
this->window().resize(minimumWindowSize);
So that is exactly what i'm doing in mu project.
Resolution os doesn't matter. I have only to have a widget for rendering video, or image in your case.
void MainWindow::resizeEvent(QResizeEvent* event)
{
QMainWindow::resizeEvent(event);
if ((player != 0) && ((player->isPlaying()) || player->isLoaded() || player>isLoaded())){
renderer->resize(ui->mainVideoWidget->width(),ui->mainVideoWidget->height());
resizeFilter();
}
}
I have a program that has several QDialogs. Most are a fixed size, but one is a resizable editor. The editor dialog gets its geometry restored from a QByteArray.
The problem I'm having is the Editor is not being centered correctly. (only the window's x/y dimensions) The Editor usually appears in the upper left corner when it is instantiated.
The Editor class gets instantiated like this in MainWindow:
Editor e(this);
e.exec();
For some reason, restoring the window geometry breaks the Editor's automatic centering relative to its parent. Before I implemented the SaveGeometry and RestoreGeomentry code today, the editor worked fine.
Here's the SaveGeometry code in Editor.cpp: (it gets executed right before the current Editor object is destroyed)
// Remember current window size
QByteArray geo=saveGeometry();
SettingsManager s;
s.SaveEditorSize(geo);
Here's the restore code in Editor.cpp: (this is part of a function that sets up the UI, assigns a layout, etc.)
if(!Buffer::editor_geometry.isEmpty()){
restoreGeometry(Buffer::editor_geometry);
}
Does anyone know what the problem is?
Edit: Actually, it does remember where it was last time, it just won't center.
Actually, I think I've solved this.
Instead of storing the entire window geometry, I tried storing just the size as a QSize. The size is then reapplied through QDialog::Resize() when the Editor is initialized. The window remembers its previous size and remains centered to parent like it is supposed to.
I am trying to use a QTextEdit widget inside of a form containing several QT widgets. The form itself sits inside a QScrollArea that is the central widget for a window. My intent is that any necessary scrolling will take place in the main QScrollArea (rather than inside any widgets), and any widgets inside will automatically resize their height to hold their contents.
I have tried to implement the automatic resizing of height with a QTextEdit, but have run into an odd issue. I created a sub-class of QTextEdit and reimplemented sizeHint() like this:
QSize OperationEditor::sizeHint() const {
QSize sizehint = QTextBrowser::sizeHint();
sizehint.setHeight(this->fitted_height);
return sizehint;
}
this->fitted_height is kept up-to-date via this slot that is wired to the QTextEdit's "contentsChanged()" signal:
void OperationEditor::fitHeightToDocument() {
this->document()->setTextWidth(this->viewport()->width());
QSize document_size(this->document()->size().toSize());
this->fitted_height = document_size.height();
this->updateGeometry();
}
The size policy of the QTextEdit sub-class is:
this->setSizePolicy(QSizePolicy::MinimumExpanding, QSizePolicy::Preferred);
I took this approach after reading this post.
Here is my problem:
As the QTextEdit gradually resizes to fill the window, it stops getting larger and starts scrolling within the QTextEdit, no matter what height is returned from sizeHint(). If I initially have sizeHint() return some large constant number, then the QTextEdit is very big and is contained nicely within the outer QScrollArea, as one would expect. However, if sizeHint gradually adjusts the size of the QTextEdit rather than just making it really big to start, then it tops out when it fills the current window and starts scrolling instead of growing.
I have traced this problem to be that, no matter what my sizeHint() returns, it will never resize the QTextEdit larger than the value returned from maximumViewportSize(), which is inherited from QAbstractScrollArea. Note that this is not the same number as viewport()->maximumSize(). I am unable to figure out how to set that value.
Looking at QT's source code, maximumViewportSize() is returning "the size of the viewport as if the scroll bars had no valid scrolling range." This value is basically computed as the current size of the widget minus (2 * frameWidth + margins) plus any scrollbar widths/heights. This does not make a lot of sense to me, and it's not clear to me why that number would be used anywhere in a way that supercede's the sub-class's sizeHint() implementation. Also, it does seem odd that the single "frameWidth" integer is used in computing both the width and the height.
Can anyone please shed some light on this? I suspect that my poor understanding of QT's layout engine is to blame here.
Edit: after initially posting this, I had the idea to reimplement maximumViewportSize() to return the same thing as sizeHint(). Unfortunately, this did not work as I still have the same problem.
I have solved this issue. There were 2 things that I had to do to get it to work:
Walk up the widget hierarchy and make sure all the size policies made sense to ensure that if any child widget wanted to be big/small, then the parent widget would want to be the same thing.
This is the main source of the fix. It turns out that since the QTextEdit is inside a QFrame that is the main widget in a QScrollArea, the QScrollArea has a constraint that it will not resize the internal widget unless the "widgetResizable" property is true. The documentation for that is here: http://doc.qt.io/qt-4.8/qscrollarea.html#widgetResizable-prop. The documentation was not clear to me until I played around with this setting and got it to work. From the docs, it seems that this property only deals with times where the main scroll area wants to resize a widget (i.e. from parent to child). It actually means that if the main widget in the scroll area wants to ever resize (i.e. child to parent), then this setting has to be set to true.
So, the moral of the story is that the QTextEdit code was correct in overriding sizeHint, but the QScrollArea was ignoring the value returned from the main frame's sizeHint.
Yay! It Works!
You may try setting minimumSize property of the QTextEdit to see if that force the layout to grow.
I don't understand most of Qt's layout scheme but setting minimum and maximum size pretty much does what I want it to do. Well, most of the time anyways.
I've noticed that the default behaviour for a DataGrid's vertical scroll bar is to scroll one row at a time. This is all well and good when the rows are all uniform and small (e.g. displaying a single line of text), but gets really ugly as soon as you have rows with variable heights.
I'm curious, is there a way to make DataGrid scrolling "smooth"? For instance, is there a way to have the DataGrid scroll by a set number of pixels, lines of text, etc. rather than scrolling one row at a time?
So far, the only solution I've managed to come up with is to place the DataGrid in a Canvas and have the Canvas do the scrolling instead of the DataGrid. The issue with this approach, though, is that as soon as the Canvas scrolls far enough, the DataGrid headers scroll off-screen. Ideally, I'd like to get the smooth-scrolling nature of the Canvas, but also keep the DataGrid headers visible. Is that possible?
The way that ItemRenderer's work in Flex 3 makes smooth scrolling difficult to achieve. Basically Flex recycles item renderers scrolled off of the top of the list as the display objects used for new data at the bottom of the list. Adobe's implementation of most list components in Flex 3 creates and adds these items as they come on to the screen rather than just off the screen, so they "pop in" and smooth scrolling isn't available. I'm not sure why they couldn't have done it in a similar manner for items +/- one position above or below the current scroll pane, but they didn't, and we're stuck with sticky scrolling by default.
Work-arounds do exist, though the one you've noted (dropping the datagrid into a canvas) negates the display-object saving intention of item renderers and incurs a performance cost. This will be fixed for most list-based Flex components in Flex 4, though it won't be fixed immediately for DataGrid. The DataGrid / AdvancedDataGrid component is maintained by a separate team based in India, last time I heard, and so it tends to be a bit behind the rest of the SDK.
I'd recommend trying something similar to this implementation of a smooth-scrolling list by Alex Harui. I'm not sure exactly how well it'd work for DataGrid or AdvancedDataGrid, but this is the most intuitive technique I can think of for making the list scroll correctly.
Try this... It's still based on Alex's code that was mentioned above. His should still be a great start for removing the snap-to-row behavior. Original source:
http://blogs.adobe.com/aharui/2008/03/smooth_scrolling_list.html
Alex's original some code for smooth vertical scrolling but that was not an issue I had with the DataGrid. It was smooth scrolling horizontally that I needed. I am using the DataGrid in an unorthodox manner for analyzing plain text reports output by our database (great way of providing visual feedback on a document). The code below allows content to go off screen and the user can scroll without that snap-to-column behavior.
You can adapt this to use the same math routines for vertical scrolling and then it will make scrolling possible and ignore the snap to row behavior. In particular switch the usage of the listContent.move method to move the contents vertically and use a inverse of the rounded pixel value you calculate from the vertical scroll bar (as opposed to my using the horizontal).
This method is bit simpler than Alex's method from the link above - a lot less code so try adapting and see how it works.
override protected function scrollHandler(event:Event):void
{
// Override the default scroll behavior to provide smooth horizontal scrolling and not the usual "snap-to-column" behavior
var scrEvt:ScrollEvent = event as ScrollEvent;
if(scrEvt.direction == ScrollEventDirection.HORIZONTAL) {
// Get individual components of a scroll bar for measuring and get a horizontal position to use
var scrDownArrow:DisplayObject = horizontalScrollBar.getChildAt(3);
var sctThumb:DisplayObject = horizontalScrollBar.getChildAt(2);
// I replaced maxHorizontalScrollPosition in Alex's code with "1300" to fix my exact application. In other situations you may finding using some property or different value is more appropriate. Don't rely on my choice.
var hPos:Number = Math.round((sctThumb.y - scrDownArrow.height) / (scrDownArrow.y - sctThumb.height - scrDownArrow.height) * 1300);
// Inverse the position to scroll the content to the left for large reports
listContent.move(hPos * -1, listContent.y);
}
// Go ahead and use the default handler for vertical scrolling
else {
super.scrollHandler(event);
}
}