views:

1005

answers:

8

hi,

I need to implement a class, using Swing, which can obtain the mouse coordinates when the user clicks anywhere on the screen. if I wanted to obtain the mouse coordinates inside my own window, I'd use a MouseListener, but I want it to work even when the user clicks outside my program.

I want my class to behave just like KColorChooser: the user clicks on the drop button and he can click anywhere on the screen to obtain the color of that spot. but I don't know if that's possible using pure Java.

A: 

Look at java.awt.Robot.getPixelColor(int x, int y)

Thomas
He's asking how to get the coordinates, not what to do with them.
Chris
He's asking about the same functionality as KColorChooser.
Thomas
I don't need to get some pixel color, KColorChooser was just an example of the mouse behavior I need. and even if that was my question, how would I obtain the two parameters (x, y) of this method? if I have, let's say, the pixel (100, 200), I'd use that method to obtain its color. but my question is exactly how I can obtain the mouse coordinates of somewhere the user clicks (to later obtain its color, or anything else I need with these numbers).
cd1
@Chris,CD1 - Right, my bad. I guess, unless you do some glass pane tricks like described above, it's actually a *feature*, not a bug that your application won't receive mouse-clicked events from outside your window. That makes sense from a window manager point of view: why should your WM send mouse events to (potentially) all runnings apps whenever the user clicks on arbitrary points on the screen?
Thomas
+1  A: 

I haven't tried this myself, but maybe you could create a full-screen, transparent panel/frame/etc, and add a MouseListener to that.

Chris
@Chris: it is possible but there are gotchas (see my answer) because not every version of Windows support fully transparent JFrame etc. Then the other issue is that on OS X a really "fully transparent" JFrame doesn't intercept the click.
Webinator
+2  A: 

I don't know if that's possible using pure Java.

Its not possible using pure Java, since Java is only aware of MouseEvents on Windows belonging to Java.

camickr
+3  A: 

It is possible though limited:

Add an AWTEventListener for focus events. As long as your app has focus before the button is clicked you'll receive a focus lost event. Then query for the pointer position.

The limitation is that, of course, your app loses focus. So depending on what you are ultimately trying to achieve this might not be useful.

If you don't want to lose focus then you will have to temporarily take a screenshot of the whole screen and display that in a screen filling window which listens for a mouse click as usual.

Proof of first method:

import java.awt.AWTEvent;
import java.awt.MouseInfo;
import java.awt.Toolkit;
import java.awt.event.AWTEventListener;

import javax.swing.JFrame;

public class Application1 {
    public static void main(String[] args) {
        Toolkit.getDefaultToolkit().addAWTEventListener(
          new Listener(), AWTEvent.MOUSE_EVENT_MASK | AWTEvent.FOCUS_EVENT_MASK);
        JFrame frame = new JFrame();
        frame.setDefaultCloseOperation(JFrame.EXIT_ON_CLOSE);
        frame.setVisible(true);
    }

    private static class Listener implements AWTEventListener {
        public void eventDispatched(AWTEvent event) {
            System.out.print(MouseInfo.getPointerInfo().getLocation() + " | ");
            System.out.println(event);
        }
    }
}

Clicking outside of the app produced:

java.awt.Point[x=198,y=59] | java.awt.event.MouseEvent[MOUSE_EXITED, ...
java.awt.Point[x=976,y=503] | java.awt.FocusEvent[FOCUS_LOST, ...

The second point is outside of the app.

Keilly
This is actually pretty clever, but it will of course only report the first click outside the application, the one that actually causes the loss of focus. After that, no other click will be reported unless the application regains focus again. Now I wonder if it's possible to react on a FOCUS_LOST event with focus request...?!
Thomas
I don't think so, the focus request or window.toFront() requests have no guarantee of affecting things outside of the VM. Testing on OSX I see this doesn't work.
Keilly
Another downside is that you won't be able to distinguish focus losses that are due to clicks outside the window and other types, such as e.g., CTRL-Tabbing to another running application.
Thomas
I have adapted this solution to my self and now it does what I need :-)
cd1
A: 

(answer deleted)

Thomas
+3  A: 

These events are directed to the window which has the focus, from all events on the desktop you can only get the mouse position.

As already shown by Keilly it's only possible to get the mouse postion.

You need to include a native lib

stacker
+2  A: 

Forget about GlassPane, there's another 100% native Java way to do it that works both on OS X and on Windows.

Java has always supported translucency for its windows on OS X and Java now supports translucency for its windows on Windows too (since Java 1.6.0_10 or so, needs to be checked).

So the trick is: upon clicking on the "pick a color" tool, you create a nearly transparent borderless Java window covering the entire screen. You set its alpha to 10 (alpha goes from 0 to 255). That alpha is so low the user won't notice that there's a very thin "nearly transparent but only very very very translucent" borderless window covering the entire screen.

Now when the user clicks on your "alpha set to 10 translucent borderless window" covering the entire screen, you get your (x,y).

Discard the borderless Java window.

Use Robot's getRgb(x,y) and you're done.

Why set the alpha to 10 and not 0? Because otherwise clicks aren't intercepted by Java but go directly to the OS (at least that's how it works for a fact on OS X). There's a treshold and I know it's not set at '1', nor '2', it's around 10 or so.

EDIT I just realized you know need to pick several colors, this is trickier but can still be done using 100% Java. Either you can live with "slightly off" colors (affected by the "nearly transparent" 'invisible' layer) or upon getting a click you must remove the layer, get the correct pixel color, and put again a "nearly transparent" layer. Now of course that is one heck of a hack but it can be done in 100% Java.

Webinator
+1 for pointing out the corner cases
Chris
it's very desirable that the class can be executed on Linux too.
cd1
A: 

I don't have enough rep yet to leave comments, but here are my comments on the other techniques:

  • Use a native lib: will work, but has obvious distribution limitations

  • Use GlassPane to fill entire screen: GlassPanes must be contained within a Window.

  • Create a Window containing a picture of the desktop and fill the entire screen: Will work, but it will suddenly make the desktop static. The cursor will no longer change, any animations or video in other windows or desktop will become eerily static.

Alternative solution: A refinement of the screen filling window, if you are using Java 6u10 or later is to make the window completely transparent. Put this window in front of all others and listen for mouse clicks. It still has shortcomings, such as no cursor changes, but it depends on what you want to do.

Keilly
Looks like WizardOfOdds has the same idea about translucent windows.Just remember that the implementation of them varies by platform each with various limitations, as he demonstrates on OSX, and in fact is not guaranteed to work on all platforms. Still, looks like we've found a few different techniques to try.
Keilly