Blurring is the act of altering images so they appear out of focus. My Maestro[2] and other illustriuos people already talked extensively about this topic here on planet KDE. However you don’t improve the world until a technology becomes available to everybody, right?

Then some days ago I read on notmart‘s blog that plasma and kwin supported blurring the background on windows.. so I rebuilt KDE4 svn, restarted it, and… boy.. does it make a difference! Look at Marco’s blog or at this picture I stole from him! All the popups become instantly more readable, even krunner does, and you immediately get the feeling that the plasma panel and all the systray applets really are above your desktop and they look as they were made of some glass/plastic material.

So: why limiting this just to some plasma applets? Well, I think that the usage of this effect must be wise, however I wanted that for my app too…

…’cause I like eye-candy!

After a quick tour through the beautiful KDE svn I grabbed all the needed source code and loaded Fotowall in creator. Since I already had the structure for the blur-behind on windows in place, adding the KDE4 effect was really easy. Here are some samples of what you can get:

To activate the back blur you need:

• obviously an application with a transparent background. Qt easily allows to work with transparency on top-level windows.
• you have to enable the “blur” effect in kwin’s “desktop effects” configuration.
• a couple of lines of code to ask kwin to activate blurring to your window.

Here is the code that does the magic:

#if defined(Q_WS_X11)
/**
  Blur behind windows (on KDE4.5+)
  Uses a feature done for Plasma 4.5+ for hinting the window manager to draw
  blur behind the window.
*/
#include <X11/Xlib.h>
#include <X11/Xatom.h>
#include <QX11Info>
static bool kde4EnableBlurBehindWindow(WId window, bool enable, const QRegion &region = QRegion())
{
    Display *dpy = QX11Info::display();
    Atom atom = XInternAtom(dpy, "_KDE_NET_WM_BLUR_BEHIND_REGION", False);

    if (enable) {
        QVector<QRect> rects = region.rects();
        QVector<quint32> data;
        for (int i = 0; i < rects.count(); i++) {
            const QRect r = rects[i];
            data << r.x() << r.y() << r.width() << r.height();
        }

        XChangeProperty(dpy, window, atom, XA_CARDINAL, 32, PropModeReplace,
                        reinterpret_cast<const unsigned char *>(data.constData()), data.size());
    } else {
        XDeleteProperty(dpy, window, atom);
    }
}
#endif

And then you can activate the blur on the caller widget with:

    kde4EnableBlurBehindWindow(winId(), true);

You could add the back blur to your application too, however I advise you that the interface is not standardized, so don’t blame either me or the kwin authors if something changes. Here is another picture, because everybody loves them

And now, as a side note, a picture from my new work place: the beautiful city of San Diego, blessed by the sun and by a ton of great programmers that by some chance happen to be co-workers of me With all the blue sky and long days, I can’t imagine a more inspiring place for making new blog posts and code experiments!

Notes:
[1] obviously the name is completely made up
[2] Italian word meaning “teacher” as in school or life teacher

KDM Ideas

1. QML in KDM
   This is all about animations, user interaction and eye candy. See this for example.
2. Plasma in KDM
   This is about consistence and configurability.
3. Webcam Login (face recognition) in KDM
   Show your face (or Mona Lisa‘s one) and you’re in! No password typing needed. Integrating with the pam-face-authentication may be the smart way to go.
4. Fingerprint Login in KDM
   This is KDE’s most requested feature… you can’t be wrong implementing it

KDE Ideas

1. Unified webcam library
   Webcam’s are being utilized a lot lately: they’re cheap, widespread, and they’re working good under Linux. Projects come out with webcam support, like Fotowall, Kamoso, Kopete and there are efforts to get it working in Phonon.
   Note that a camera is a source of video, images but also can be used for authentication and may even be used for user input – can you imagine manipulating stuff with your fingers? Note that webcams must usually be read in single instance, so the approach here might be 1 Qt-Only userspace daemon supporting multiple connections.
2. Unified Theme Manager
   One theme manager to rule them all: styles, colors, plasma theme, windeco & all. Of course the themes must be packed and shared in a binary-compatible manner with any other KDE system (yes, the challenges are there..
3. Unified Configuration Panel
   Making systemsettings really the one and only configuration tool and provide backends to the distributions for adding/changing modules so KDE systems will have the same feeling for configuring the system. Some modules like ‘partition disk’, ‘edit boot menu’ and ‘manage startup programs’ (daemons) should be added and left to the distribution for implementation – I think they will. Note that current systemsettings lets you configure the joystick even if you don’t have one (!).
4. KWIN: draggable windows
   And for windows I mean (actually J. Janz means) the whole non-interactive panel. Blank space as well as inactive labels should be draggable. From the user POV there is no window-deco + window contents. It’s a panel, let’s use it, move it, resize it, context-menu-it as a whole.
5. Raytraced UI
   I’ll blog more about this, but this is a little thing I’d really like to do but I don’t know if I will find time to do that. Do you think that iPhone gui is cool? Raytrace the .ui high-level gui definition format and you’ll get something better. Glassy too. Mail me if you want to know more about this

Conclusion

I’m sorry I missed some interesting suggestions but I don’t have space here to detail them; just look to my previous blog entry’s comments to find more about BasKet note pads integration into Plasma, file browsing enhancements (context dependent actions, menus, warnings and kparts) qt4 gui for vi, etc.

As someone pointed out there is quite a bit of overlap and dispersion between the ways to collect ideas. I can count:

• KDE Bugzilla’s wishlist
• KDE Forum’s brainstorm
• distribution specific bug trackers
• blog posts (ehm….:-)
• GSOC ideas
• kde-look’s brainstorm
• developers mailing lists
• other bugtrackers
• techbase or userbase
• direct mails and irc chats

Directing users to provide their feedback in a low-dispersive, constructive, modern, layered (from generic directions to specific actions) and focused way is probably one of the challenges that should be added to Aaron’s list for 2010.

Collecting Ideas

I want to blog about some cool things to do for the KDE project on week-1 2010. I’ve got just a couple of good ideas, so I’m asking you to add your ideas so we can make a good collection of nice things to do. You can either reply to this post (don’t be shy, just post it!) or send it to me at <enrico.ros@gmail.com>.

Sharing your idea with the community raises the chance of seeing it implemented!

The KDE login Manager

The login manager is the “insert password” dialog that gets in the way when you want to use your computer. Of course you may not be the only user of the machine, so it’s somewhat needed when you share the seat. Some facts:

Kdm Top to Bottom

Qml to the rescue!

What if we throw away (or keep for “retro-compatibility”) the classic and themer frontends and just add a frontend using QML? This is the name of the language of the Qt Declarative module, that has just seen the light and will be merged into Qt for the 4.7 release.

Imagine this:

• theme creators will have 100X or more expressive power: make snow! login via a tetris puzzle! bounce my webcam picture!
• there will be animated themes (for login, user change, logout, etc..) or simple ones for more ‘classic’ kind of people
• kdm only needs to add a couple of bindings to the qml runtime, like functions for logging in and shutting down / suspend the system
• there is no need for compiled binaries, packages will be interpreted by the qml runtime – finally a painless get-hot-new-stuff (compared to plasma widgets, at least).

Conclusion

MultiTouch can be Fun !

I couldn’t resist, so for 99€ (120$) I bought a Bamboo Fun Pen & Touclh tablet. I find this really affordable, since you get: A. the precision of a graphics tablet (with 1024 pressure levels, back eraser, etc..) and B. a real 2-Pointers input device!

I tried the tablet on vista, with the latest drivers and the bundled applications bot I got disappointed because:
1. the scroll gesture is remapped to the mouse wheel.. so you lose all the ‘analog’ fun and it feels unconfortable.
2. the zoom gesture is remapped to CTRL+wheel, awkward fixed-steps again.
3. the rotate gesture is not supported by any app even the bundled ones i tried.

In the end I thought this was not “multi-touch”, but some silly gesture-detection broken hardware. I was wrong.

Linux to the rescue

Step 1. After loading the wacom.ko module (updated with this patch) the Linux kernel prints out stuff like this:

[default   ] data: 02 00  00 00 00 00 00  00 00 00 00  00 00 00 00 00  00 00 00 00
[finger 1  ] data: 02 00  7d 80 fe 00 8a  80 fc 00 87  00 00 00 00 00  81 11 00 00
[finger 1+2] data: 02 00  ae 80 b6 00 72  81 08 00 70  87 81 5b 00 72  81 21 00 00
[finger   2] data: 02 00  00 00 00 00 00  81 59 00 b4  6d 81 59 00 ba  80 11 00 00

Can you see it ? This is a 2-finger device!

Step 2. The Kernel understands the data and splits it into 2 input devices that Udev maps to /dev/input/wacom and /dev/input/wacom-touch. From here you can read higher level information for each finger such as the decoded position, pressure, click type (tap, doubletab), and more.

I WANT QTouchEvents AND QGestures !!

I do too And I wanted to hack something about that. But I stopped here:

We get the input from evdev and the Application needs QTouchEvents and QGestures, so the easy way would be to hack up a touchpoint feeder to the Qt framework that directly reads from evdev (evdev -> qt). This would work fine for 1 application, but what about two? And what about moving the X cursor too? I think evdev->Xdriver->XInput2->qt is the solution, but we must be sure that no valuable information is lost in the chain, that the event filtering is done in the right place and obviously, that the X driver does a good job. Unfortunately xf86-input-wacom (X input driver for wacom’s kernel driver) doesn’t support my Bamboo Fun Touch & Pen yet and moreover I don’t see MPX support in the sources, so I’m wondering how far away is the Qt multi-touch backend from seeing the light.

Conclusion

I don’t know which way the multi-touch will happen on linux, but I know for sure that we want to play with gestures and we want it soon

For more information:

• Peter Hutterer’s blog. Peter is the MPX (Multi Pointer X) author and plays a big role in the X.org community.
• X Input Extension 2.0 features.
• The Linux Wacom project. A community-driven effort to develop kernel and X drivers.

I confess, I’m addiQted. You obviously think it’s normal (you’re not reading this blog by accident, aren’t you?) but it wasn’t always like this: I’ve liked some previous versions, but the upcoming Qt 4.6 really rocks!

The Fotowall case

As I shown previosly, Fotowall already uses some features of Qt 4.6. How much?
Qt 4.6 is referenced in 32 places, for example in snippets like this:

    // center the path
    QPointF pathCenter = path.boundingRect().center();
    #if QT_VERSION >= 0x040600
        path.translate(-pathCenter);
    #elif QT_VERSION >= 0x040500
        path = QTransform::fromTranslate(-pathCenter.x(), -pathCenter.y()).map(path);
    #else
        QTransform tx;
        tx.translate(-pathCenter.x(), -pathCenter.y());
        path = tx.map(path);
    #endif

From simple path translations to more complex property manipulations, everything got better. See this code snipped about extending QGraphicsObject in 4.6 versus 4.5 and 4.4.

    #if QT_VERSION >= 0x040600
    class AbstractContent : public QGraphicsObject
    #else
    class AbstractContent : public QObject, public QGraphicsItem
    #endif
    {
        Q_OBJECT
    #if QT_VERSION < 0x040600
        Q_PROPERTY(QPointF pos READ pos WRITE setPos)
        Q_PROPERTY(qreal rotation READ rotation WRITE setRotation NOTIFY rotationChanged)
    #endif
    #if QT_VERSION < 0x040600 && QT_VERSION >= 0x040500
        Q_PROPERTY(qreal opacity READ opacity WRITE setOpacity)
    #endif

Do you feel the pain?

I Need it

I need a world with Qt 4.6 only, because:

• It’s too much of a pain to support older versions in the same codebase
• It’s not even fair, because non-4.6 users won’t enjoy the program as it’s meant to be
• I can’t use A TON of new features because they’ll either simplify too much the code, or just because they don’t have a pre-4.6 equivalent
• and because I’d like to finally use
  • QGraphicsEffect for Fotowall’s content
  • QVector3D to say goodbye to enricomath.h
  • QGesture for multi-finger interaction
  • QPropertyAnimation to its full potential!

More 4.6 TEASING

1. As you seem to like videos, here is a preview of something we’re working on for Amarok Qt 4.6 only!!
   

   Click here to view the embedded video.

2. Also, since I’m soooo much a giver, see the new Fotowall’s Word Cloud feature, coming in 0.9

See you @ Munich

I’m packing for the DevDays 2009 Munich. See you there

Indirect Benefits, or ‘what you get for free’

• Improved quality and better speed in the OpenGL QPainter
• Better speed in the GraphicsView framework

Direct Benefits

• Animation Framework: this simply r0cks! I don’t know about the whole Kinetic stuff, I just mean the ability to create simple property animations.
  I don’t have a video to show you but just take a look at the Fotowall’s Top-Bar, the tabs appear and disappear with a nice and smooth grow-shrink animation, that make the program look soo 2009-ish! finally!
• QNetworkReply::isFinished(): thanks to this method I could implement a smart time-saver: when the user starts to drag a flickr picture, Fotowall starts downloading it. It can happen that, upon drop, the content has already been completely downloaded (so it can be placed in the canvas) or partially downloaded (so the used only needs to wait for the missing percent)

Drawbacks

• There are some graphics glitches when animating the opacity property of a QGaphicsProxyWidget (in OpenGL)
• Even by setting the QGraphicsItem::ItemSendsGeometryChanges flag, the previous behavior is not perfectly restored (we used itemChange, and the new behavior doesn’t allow Fotowall’s mirrors to sync properly to the objects)

Movie Time!

I’ve tried hard to record a screencast about the Cuteness & Performance of Qt 4.6, but in the end I had to use my camera.

Click here to view the embedded video.

Thanks Nokia for this great release!

[1] Fotowall is a creative program to mix local and online graphics content and build original work.

For those who don’t know me, I’m a long time KDE user (since 2.2), I worked on several KDE projects like okular, kdm, amarok, konqueror, as well as releasing other software like Fotowall and other stuff on various repositories.

This planet is the collector of many of my favorite blogs and it’s an honor for me to be here.

See you soon (with some developer tool hacks…stay tuned) !