I'm programming a small PoC in QML. In a couple of places in my code I need to bind to/query global mouse position (say, mouse position in a scene or game window). Even in cases where mouse is outside of MouseAreas that I've defined so far.
Looking around, the only way to do it seems to be having whole screen covered with another MouseArea, most likely with hovering enabled. Then I also need to deal with semi-manually propagating (hover) events to underlying mouseAreas..
Am I missing something here? This seems like a pretty common case - is there a simpler/more elegant way to achieve it?
EDIT:
The most problematic case seems to be while dragging outside a MouseArea. Below is a minimalistic example (it's using V-Play components and a mouse event spy from derM's answer). When I click the image and drag outside the MouseArea, mouse events are not coming anymore so the position cannot be updated unless there is a DropArea below.
The MouseEventSpy is taken from here in response to one of the answers. It is only modified to include the position as parameters to the signal.
import VPlay 2.0
import QtQuick 2.0
import MouseEventSpy 1.0
GameWindow {
id: gameWindow
activeScene: scene
screenWidth: 960
screenHeight: 640
Scene {
id: scene
anchors.fill: parent
Connections {
target: MouseEventSpy
onMouseEventDetected: {
console.log(x)
console.log(y)
}
}
Image {
id: tile
x: 118
y: 190
width: 200
height: 200
source: "../assets/vplay-logo.png"
anchors.centerIn: parent
Drag.active: mausA.drag.active
Drag.dragType: Drag.Automatic
MouseArea {
id: mausA
anchors.fill: parent
drag.target: parent
}
}
}
}
You can install a eventFilter on the QGuiApplication, where all mouse events will pass through.
How to do this is described here
In the linked solution, I drop the information about the mouse position when emitting the signal. You can however easily retrieve the information by casting the QEvent that is passed to the eventFilter(...)-method into a QMouseEvent and add it as parameters to the signal.
In the linked answer I register it as singleton available in QML and C++ so you can connect to the signal where ever needed.
As it is provided in the linked answer, the MouseEventSpy will only handle QMouseEvents of various types. Once you start dragging something, there won't be QMouseEvents but QDragMoveEvents e.t.c. Therefore you need to extend the filter method, to also handle those.
bool MouseEventSpy::eventFilter(QObject* watched, QEvent* event)
{
QEvent::Type t = event->type();
if (t == QEvent::MouseButtonDblClick
|| t == QEvent::MouseButtonPress
|| t == QEvent::MouseButtonRelease
|| t == QEvent::MouseMove) {
QMouseEvent* e = static_cast<QMouseEvent*>(event);
emit mouseEventDetected(e->x(), e->y());
}
if (t == QEvent::DragMove) {
QDragMoveEvent* e = static_cast<QDragMoveEvent*>(event);
emit mouseEventDetected(e->pos().x(), e->pos().y());
}
return QObject::eventFilter(watched, event);
}
You can then translate the coordinates to what ever you need to (Screen, Window, ...)
As you have only a couple of places where you need to query global mouse position, I would suggest you to use mapToGlobal or mapToItem methods.
I believe you can get cursor's coordinates from C++ side. Take a look on answer on this question. The question doesn't related to your problem but the solution works as well.
On my side I managed to get global coordinates by directly calling mousePosProvider.cursorPos() without any MouseArea.
Related
I have a composable representing list of results:
#Composable
fun ResultsList(results: List<Pair<Square, Boolean>>) {
val coroutineScope = rememberCoroutineScope()
val listState = rememberLazyListState()
LazyRow(state = listState) {
items(results) { result ->
ResultsItem(result.first, result.second)
coroutineScope.launch {
listState.animateScrollToItem(results.size)
}
}
}
}
Expected behaviour: The list smoothly scrolls to the last item whenever a new item is added
Actual behaviour: All is good, but whenever I manually scroll fast through the list, it is also automatically put on the bottom. Also, the scrolling is not smooth.
Your code gives the following error:
Calls to launch should happen inside a LaunchedEffect and not composition
You should not ignore it by calling the side effect directly from the Composable function. See side effects documentation for more details.
You can use LaunchedEffect instead (as this error suggests). By providing results.size as a key, you guarantee that this will be called only once when your list size changes:
#Composable
fun ResultsList(results: List<Pair<Square, Boolean>>) {
val listState = rememberLazyListState()
LaunchedEffect(results.size) {
listState.animateScrollToItem(results.size)
}
LazyRow(state = listState) {
items(results) { result ->
ResultsItem(result.first, result.second)
}
}
}
Philip's solution will work for you. However, I'm posting this to ensure that you understand why
A.) The scroll was not smooth
B.) The list gets scrolled to the bottom when you scroll through it fast enough.
Explanation for A.)
It is because you are using animateScollTo. I've experienced issues with this method if called too often,
Explanation for this lies in how Lazy scrollers handle their children internally. You see, Lazy scrollers, as you might know, are meant to display only a small window of a large dataset to the user. To achieve this, it uses internal caching. So, the items on the screen, AND a couple of items to the top and bottom of the current window are cached.
Now, since in your code, you are making a call to animateScrollTo(size) inside the Composable's body (the items scope), the code will essentially be executed upon every composition.
Hence, on every recomposition, there is an active clash between the animateScrollTo method, and the users touch input. When the user scrolls past in a not-so-fast manner, this is what happens - user presses down, gently scrolls, then lifts up the finger. Now, remember this, for as long as the finger is actually pressed down, they animateScrollTo will seem to have no effect (because the user is actively holding a position on the scroller, so it won't be scrolled past it by the system). Hence, while the user is scrolling, some items ahead of the list are cached, but the animateScrollTo does not work. Then, because the motion is slow enough, the distance the scroller travels because of inertia is not a problem, since the list already has enough cached items to show for the distance. That also explains the second problem.
B.)
When you are scrolling through the list FAST enough, the exact same thing as the above case (the slow-scroll) happens. Only, this time the inertia carries the list too forward for the scroller to be handled based on the internal cache, and hence there is active recomposition. However, now since there is no active user input (they have lifted their finger off the screen), it actually does animate to the bottom, since their is no clash here for the animateScrollTo method.
For as long as your finger is pressed, no matter how fast you scroll, it won't scroll to the bottom (test that!)
Now to the solution of the actual problem. Philip your answer is brilliant. The only thing is that it might not work if the developer has an item remove implementation as well. Since only the size of the list is monitored, it will scroll to end when an item is added OR deleted. To counteract that, we would actually need some sort of reference value. So, either you can implement something of your own to provide you with a Boolean variable that actually confirms whether an item has been ADDED, or you could just use something like this.
#Composable
fun ResultsList(results: List<Pair<Square, Boolean>>) {
//Right here, create a variable to remember the current size
val currentSize by rememberSaveable { mutableStateOf (results.size) }
//Now, extract a Boolean to be used as a key for LaunchedEffect
var isItemAdded by mutableStateO(results.size > currentSize)
LaunchedEffect (isItemAdded){ //Won't be called upon item deletion
if(isItemAdded){
listState.animateScrollToItem(results.size)
currentSize = results.size
}
}
val listState = rememberLazyListState()
LazyRow(state = listState) {
items(results) { result ->
ResultsItem(result.first, result.second)
}
}
}
This should ensure the proper behaviour. Of course, let me know if there is anything else, happy to help.
Pretty obvious. Why are you calling:
listState.animateScrollToItem(results.size) inside your LazyList? Of course you're going to get extremely bad performance. You shouldn't be messing around with scrolling when items are being rendered. Get rid of this line of code.
hoping someone can help. I am creating an app whereby the user will touch a series of images to rotate them. What I am trying to do. Is highlight the image once the user has rotated to a particular position.
Is this possible? If, so any tips greatly appreciated.
edit - ok here's an example instead!
First, the simplest way, based off the code example you just posted:
r1c1.setOnClickListener {
r1c1.animate().apply{ duration = 100 rotationBy(270f) }.start()
}
So the issue here is that you want to highlight the view when it's rotated to, say 90 degrees, right? But it has an animation to complete first. You have three options really
do something like if (r1c1.rotation + 270f == 90) and highlight now, as the animation starts, which might look weird
do that check now, but use withEndAction to run the highlighting code if necessary
use withEndAction to do the checking and highlighting, after the anim has finished
the latter probably makes the most sense - after the animation finishes, check if its display state needs to change. That would be something like this:
r1c1.animate().setDuration(100).rotationBy(270f).withEndAction {
// need to do modulo so 720 == 360 == 0 etc
if (r1c1.rotation % 360 == TARGET_ROTATION) highlight(r1c1)
}.start()
I'm assuming you have some way of highlighting the ImageViews and you weren't asking for ways to do that!
Unfortunately, the problem here is that if the user taps the view in the middle of animating, it will cancel that animation and start a new one, including the rotationBy(270) from whatever rotation the view currently happens to be at. Double tap and you'll end up with a view at an angle, and it will almost never match a 90-degree value now! That's why it's easier to just hold the state, change it by fixed, valid amounts, and just tell the view what it should look like.
So instead, you'd have a value for the current rotation, update that, and use that for your highlighting checks:
# var stored outside the click listener - this is your source of truth
viewRotation += 270f
# using rotation instead of rotationBy - we're setting a specific value, not an offset
r1c1.animate().setDuration(100).rotation(viewRotation).withEndAction {
// checking our internal rotation state, not the view!
if (viewRotation % 360 == TARGET_ROTATION) highlight(r1c1)
}.start()
I'm not saying have a single rotation var hanging around like that - you could, but see the next bit - it's gonna get messy real quick if you have a lot of ImageViews to wrangle. But this is just to demonstrate the basic idea - you hold your own state value, you're in control of what it can be set to, and the View just reflects that state, not the other way around.
Ok, so organisation - I'm guessing from r1c1 that you have a grid of cells, all with the same general behaviour. That means a lot of repeat code, unless you try and generalise it and stick it in one place - like one click listener, that does the same thing, just on whichever view it was clicked on
(I know you said youre a beginner, and I don't like loading too many concepts on someone at once, but from what it sounds like you're doing this could get incredibly bloated and hard to work with real fast, so this is important!)
Basically, View.onClickListener's onClick function passes in the view that was clicked, as a parameter - basically so you can do what I've been saying, reuse the same click listener and just do different things depending on what was passed in. Instead of a lambda (the code in { }, basically a quick and dirty function you're using in one place) you could make a general click listener function that you set on all your ImageViews
fun spin(view: View) {
// we need to store and look up a rotation for each view, like in a Map
rotations[view] = rotations[view] + 270f
// no explicit references like r1c1 now, it's "whatever view was passed in"
view.animate().setDuration(100).rotation(rotations[view]).withEndAction {
// Probably need a different target rotation for each view too?
if (rotations[view] % 360 == targetRotations[view]) highlight(view)
}.start()
}
then your click listener setup would be like
r1c1.setOnClickListener { spin(it) }
or you can pass it as a function reference (this is already too long to explain, but this works in this situation, so you can use it if you want)
r1c1.setOnClickListener(::spin)
I'd recommend generating a list of all your ImageView cells when you look them up (there are a few ways to handle this kind of thing) but having a collection lets you do things like
allCells.forEach { it.setOnClickListener(::spin) }
and now that's all your click listeners set to the same function, and that function will handle whichever view was clicked and the state associated with it. Get the idea?
So your basic structure is something like
// maybe not vals depending on how you initialise things!
val rotations: MutableMap<View, Float>
val targetRotations: Map<View, Float>
val allCells: List<ImageView>
// or onCreateView or whatever
fun onCreate() {
...
allCells.forEach { it.setOnClickListener(::spin) }
}
fun spin(view: View) {
rotations[view] = rotations[view] + 270f
view.animate().setDuration(100).rotation(rotations[view]).withEndAction {
val highlightActive = rotations[view] % 360 == targetRotations[view]
highlight(view, highlightActive)
}.start()
}
fun highlight(view: View, enable: Boolean) {
// do highlighting on view if enable is true, otherwise turn it off
}
I didn't get into the whole "wrapper class for an ImageView holding all its state" thing, which would probably be a better way to go, but I didn't want to go too far and complicate things. This is already a silly length. I might do a quick answer on it just as a demonstration or whatever
The other answer is long enough as it is, but here's what I meant about encapsulating things
class RotatableImageView(val view: ImageView, startRotation: Rotation, val targetRotation: Rotation) {
private var rotation = startRotation.degrees
init {
view.rotation = rotation
view.setOnClickListener { spin() }
updateHighlight()
}
private fun spin() {
rotation += ROTATION_AMOUNT
view.animate().setDuration(100).rotation(rotation)
.withEndAction(::updateHighlight).start()
}
private fun updateHighlight() {
val highlightEnabled = (rotation % 360f) == targetRotation.degrees
// TODO: highlighting!
}
companion object {
const val ROTATION_AMOUNT = 90f
}
}
enum class Rotation(var degrees: Float) {
ROT_0(0f), ROT_90(90f), ROT_180(180f), ROT_270(270f);
companion object {
// just avoids creating a new array each time we call random()
private val rotations = values()
fun random() = rotations.random()
}
}
Basically instead of having a map of Views to current rotation values, a map of Views to target values etc, all that state for each View is just bundled up into an object instead. Everything's handled internally, all you need to do from the outside is find your ImageViews in the layout, and pass them into the RotatableImageView constructor. That sets up a click listener and handles highlighting its ImageView if necessary, you don't need to do anything else!
The enum is just an example of creating a type to represent valid values - when you create a RotatableImageView, you have to pass one of these in, and the only possible values are valid rotation amounts. You could give them default values too (which could be Rotation.random() if you wanted) so the constructor call can just be RotatableImageView(imageView)
(you could make more use of this kind of thing, like using it for the internal rotation amounts too, but in this case it's awkward because 0 is not the same as 360 when animating the view, and it might spin the wrong way - so you pretty much have to keep track of the actual rotation value you're setting on the view)
Just as a quick FYI (and this is why I was saying what you're doing could get unwieldy enough that it's worth learning some tricks), instead of doing findViewById on a ton of IDs, it can be easier to just find all the ImageViews - wrapping them in a layout with an ID (like maybe a GridLayout?) can make it easier to find the things you want
val cells = findViewById<ViewGroup>(R.id.grid).children.filterIsInstance<ImageView>()
then you can do things like
rotatables = cells.map { RotatableImageView(it) }
depends what you need to do, but that's one possible way. Basically if you find yourself repeating the same thing with minor changes, like the infomercials say, There Has To Be A Better Way!
I am trying to find out why flicking is not working with a TreeView example on my Raspberry Pi3 with touch screen.
Looking at the qml code of TreeView.qml, e.g.
https://github.com/RSATom/Qt/blob/master/qtquickcontrols/src/controls/TreeView.qml:
BasicTableView {
...
__mouseArea: MouseArea {
id: mouseArea
parent: __listView
width: __listView.width
height: __listView.height
z: -1
propagateComposedEvents: true
focus: true
// If there is not a touchscreen, keep the flickable from eating our mouse drags.
// If there is a touchscreen, flicking is possible, but selection can be done only by tapping, not by dragging.
preventStealing: !Settings.hasTouchScreen
...
}
}
By similarly looking at the qml code for BasicTableView.qml, it seems that behavior is controlled by Settings.hasTouchScreen.
According to:
https://code.woboq.org/qt5/qtquickcontrols/src/controls/Private/qquickcontrolsettings.cpp.html
it corresponds to the following method:
bool QQuickControlSettings1::hasTouchScreen() const
{
const auto devices = QTouchDevice::devices();
for (const QTouchDevice *dev : devices)
if (dev->type() == QTouchDevice::TouchScreen)
return true;
return false;
}
However, in my case, Settings.hasTouchScreen returns false; i.e. the touch screen (although working for the rest), is not
correctly detected by the QML environment, which probably explains why the flicking does not work.
According to https://doc.qt.io/qt-5/qtouchdevice.html, my touch device should have been registered somehow by the private QWindowSystemInterface::registerTouchDevice() method, but wasn't.
How can I get this to work?
Thanks!
It seems not to work correctly with tslib, but works by using the evdevtouch plugin which is enabled by adding the following command line arguments when launching the program:
-plugin evdevtouch:/dev/input/eventX
where eventX is the event assigned to the touch input.
With that, QTouchDevice::devices() is no longer empty, and the flicking of the TreeView works.
I have two shapes in my canvas using CreateJS. In each shape I included a hit area with the own shape with a mouseover listener. Two shapes are one above the other. When I click into the shape, I received the two callbacks. It's possible to get only the callback to the visible shapes?
enter image description here
Similar to the DOM, the way mouse interaction works is to bubble up the display list, which excludes elements that are not part of the hierarchy chain of the event target.
This means siblings, or elements of other display lists that are underneath will not receive event handlers (which is what you described), and you will not receive mouse events for elements that are not the target of the mouse event.
However, you can wire up your own interaction fairly easily using getObjectsUnderPoint, which tells you what is under the mouse.
stage.on("click", handleClick);
function handleClick(event) {
var list = stage.getObjectsUnderPoint(event.localX, event.localY);
for (var i=0, l=list.length; i<l; i++) {
console.log(list[i]);
}
}
Here is a quick sample: http://jsfiddle.net/y8jhb26x/
Note that you can add the mouse event to any container you want to constrain what objects will trigger this check (I just used stage), but when you call getObjectsUnderPoint, it will return anything under the mouse. If you want to only check items in that container, you can use the contains method to filter out unwanted children:
for (var i=0, l=list.length; i<l; i++) {
if (someContainer.contains(list[i])) {
console.log(list[i]);
}
}
You can also use arguments on getObjectsUnderPoint to filter out items with mouse handlers, or respect the mouseChildren/mouseEnabled property, which is how actual mouse interaction works.
getObjectsUnderPoint method
mouse interaction code
Hope that helps!
In my app, I have a couple of UISlider instances to change various values. The values are displayed right next to the slider, as well as rendered in a 3d space in another visible part of the app.
The 3d part includes some rather heavy calculations, and right now it doesn't seem possible to update it live as the slider changes. That would imply that I'd have to set the slider's continuous property to NO, therefore only getting updates when the slider has finished changing.
I'd prefer to have the displayed value update live, however. Is there a way to have a slider that is continuous (so I can update my value-label in real time) and still sends some kind of message once the user has finished interacting with it? My gut feeling right now is to subclass UISlider and override the touchesEnded: method. Is that feasible?
You can do this with simple target/actions.
Set a target and action for the UIControlEventValueChanged event, and then another target and action for the UIControlEventTouchUpInside event. With the continuous property set to YES, the value changed event will fire as the slider changes value, while the touch up inside event will only fire when the user releases the control.
I just had to do this, so I looked up touch properties, and used the full IBAction header.
This should be a viable alternative for people who want some extra control, though Jas's is definitely easier on the code side.
- (IBAction)itemSlider:(UISlider *)itemSlider withEvent:(UIEvent*)e;
{
UITouch * touch = [e.allTouches anyObject];
if( touch.phase != UITouchPhaseMoved && touch.phase != UITouchPhaseBegan)
{
//The user hasn't ended using the slider yet.
}
}
:D
Also note you should connect the UIControlEventTouchUpOutside event as well in case the user drags his finger out of the control before lifting it.
In Swift 3:
#IBAction func sliderValueChanged(_ slider: UISlider, _ event: UIEvent) {
guard let touch = event.allTouches?.first, touch.phase != .ended else {
// ended
return
}
// not ended yet
}