With this new API, it was possible to fix all the flickering in the included examples! I’ve also added a few new example scripts. Check out the Polygon example, as well as all the text-related ones.

Changelog

1.1 Build 287:

• Added manual drawing mode with suspendDrawing/resumeDrawing methods
  (this should allow users to work around the flickering)
• Fixed text position being off baseline when drawing on a path
• Fixed a rare crash when drawing text on a path
• Activated Pixman MMX&SSE2 fast paths
• Updated Pango to fix font leak (rev 2746)
• Updated Pixman (0.13.2)

The main thing not fixed in this version is the flickering when drawing, but I think I know what the root of the problem is and I might be able to fix it for the final 1.1 build.

You can find a beta build here.

Changelog

1.1 Build 269:

• Added textPath, textAlongPath (with stroke or fill option)
• Added support for shadows
• arcTo is incorrect when the three points are collinear
• arcTo should draw a straight line from P0 to P1
• restore() with an empty stack has no effect
• Corrected order of data returned by getImageData (BRGA -> RGBA)
• ImageData is now converted to non-premultiplied color
• Preserve the current path when clipping
• rgba colors were not converted to string properly (green had the wrong value)
• Some composite operations are now handled correctly
• Fixed Debug build crashing when expired
• Fixed stack overflow crash on exit
• Added check for font validity on assignment and fallback to default font (will raise an error in debug mode if font is incorrect)
• Moved state management to its own class
• Updated Cairo(1.8.4), Pixman(0.12.0), Pango(1.22.3), Glib(2.18.3) and libpng(1.2.33)

You can set the window size (default is 800×600), chose an output file and FourCC code for the output video.

It currently only runs unprotected dreams: the dreams that come with DeskScapes will work, but any dynamic dream that you bought won’t.

See the readme file for command line switches.

Download test build here.

Update: Added preliminary support for protected Dreams (you will have to activate them first through DeskScapes).

Enter DXCanvas, a DesktopX drawing plugin that implements the Canvas spec (plus some DesktopX-specific additions).

How does it work

DXCanvas is a DesktopX drawing plugin, which means it takes over drawing for the object it is associated to.

In the configuration, you can set the initial size for the canvas. By default, the drawing surface size will be 300×150. You can also specify whether you want the surface to be transparent and show the other objects or windows under it, or opaque.

A canvas object is made available to scripting. To be able to draw on the canvas, you need to request a context. This is done by calling getContext(type) on the canvas object. The only supported type right now is "2d".

See the list of functions in the implementation section. Note that some are slightly different from the Canvas specification, or are completely new. This is to adjust to differences in what is possible in a browser and in DesktopX.

Canvas Controller Widget

I’ve made a little widget to make testing easier. It loads a list of scripts from a user-defined folder and allows you to switch between them. It support scripts written in both JScript and VBScript.

It basically associate the script to an object having the DXCanvas plugin as a capability.

Two functions are required for the scripts to work: Object_OnScriptEnter() and Object_OnScriptExit().

The canvas size is reset its default size of 300×150 when switching between scripts, so make sure to set it to the required size if needed by your script.

To add a new script to the list, create a new .js or .vbs file in a folder and select this folder in the widget preferences. A few test scripts are included in the Script folder.

It also support subfolders (only 1 folder deep), so you can organize your scripts in subfolders.

How to help

Download the Canvas Controller widget and start creating scripts!

What to look for

• strange behavior: colors are not right, context is not saved or restored correctly, etc.
• the output from a script is different in Firefox or Safari
• crashes or memory leaks (there are a least a few :P)

What is broken/not working properly

• Drawing still flicker (not as badly as before)
• This is a DEBUG build, so it’s going to be slower than normal

Links

• HTML5 Spec on Canvas
• Canvas Tutorial
• Yahoo! Widgets Canvas Documentation (Look in Core DOM Reference -> Canvas / CanvasRenderingContext2D)

Changelog

1.0 Build 225:

• Fixed large canvas objects taking a lot of CPU
• Log is always enabled

1.0 Build 217

• Fixed drawing being incomplete in some cases
• Added workaround for pink color drawing transparent. DesktopX always use pink to draw transparent surfaces, even in per-pixel mode. rgb(255, 0, 255) will be changed to rgb(255, 1, 255)

1.0 Build 214

• Added per-object log files
• Added more log info when creating a new context object

1.0 Build 211

• Added support for % in rgb/rgba colors
• Added "transparent" color (equivalent to "rgba(0,0,0,0)")
• Added SVG colors "darkgrey", "darkslategrey", "dimgrey", "grey", "lightgray", "lightslategrey" and "slategrey"
• Zero size canvas is allowed
• Properly share canvas state (context style stack was not properly shared before)
• DrawImage and DrawImageRegion check for negative width and height and adjust the coordinates
• Only match full color strings
• Accept rgb values outside [0;255] and clamp the value properly
• Clamp alpha and use premultiplied color values when painting with alpha (still somewhat wrong)
• Fixed leak and crash with CanvasImageData & CanvasPixelArray
• Fixed crash with invalid parameters in createImageData and getImageData
• Zero-length line gradients should paint nothing
• Radial gradients with the same start and end circle should paint nothing
• Image data now accepts negative width and height (you get the untransformed data for the resulting rectangle)
• arc() with zero radius draws a line to the start point.
• arcTo() has no effect if the context has no subpaths or if P0 = P1
• arcTo() draws a straight line to P1 if P1 = P2 or if radius = 0
• bezierCurveTo() has no effect if the context has no subpaths
• lineTo() has no effect if the context has no subpaths
• quadraticCurveTo() has no effect if the context has no subpaths
• Updated cairo, pixman and pango libraries

1.0 Build 201

• Added hsl/hsla color parsing
• Added some logging (DXCanvas.log in object directory)
• Fixed text stroking always drawing at position (0,0)
• Fixed text baseline being off by the text height

1.0 Build 191

• Added redraw calls buffering (less flickering)
• Added debugMode attribute to canvas. In debug mode, passing invalid parameters will result in an error being raised instead of the invalid value being silently ignored.
• No longer returns errors when passing invalid parameters to a lot of functions (as specified in Canvas specification)
• CanvasPixelArray.XXX6(index, value) now takes an int instead of a char and will clamp that value to [0;255]

1.0 Build 180

• Implemented toImage
• Implemented createImageData, getImageData and putImageData
• Added canvas support to createPattern, drawImage and drawImageRegion
• Fixed createPattern, drawImage and drawImageRegion not working in VBScript
• Fixed createPattern, drawImage and drawImageRegion not making a copy of the source image or canvas
• Fixed drawImageRegion using wrong default values for dw and dh
• Fixed drawImageRegion not creating a new path, resulting in image "corruption"

1.0 Build 168

• Implemented font attribute setter/getter
• Implemented fillText, strokeText and measureText (using Pango)
• Implemented arcTo (using patch from Behdad Esfahbod)
• Fixed image loading (SD_LOAD_IMAGE seems to be corrupting some images…)

1.0 Build 159

• Implemented textAlign and textBaseline attributes setters/getters  (not used yet)
• Implemented createPattern, loadImage, drawImage and drawImageRegion
• Update to cairo 1.7.4 (better support for text)
• Statically linked to cairo lib (no need for cairo.dll anymore)

1.0 Build 149

• First test version

Download

You can download a test version here. It only includes the Canvas Controller widget for now. Please do not use the DXCanvas plugin in your own objects and widgets yet. This version of the plugin will expire on the 10/01/2008.

Implementation

Under the hood, it’s using Cairo, a 2D vector graphics library that also powers the Mozilla and Yahoo Widgets implementations. Right now, the Cairo library is dynamically loaded at runtime, but I hope to have it statically linked into the plugin for the final version.

Here is a list classes with their attributes and functions and the state of their implementation

Canvas

CanvasRenderingContext2D

Gradient

ImageData

PixelArray

TextMetrics

Still some bugs with the Bouncing Box sample, but it’s getting there. Now let’s proxy some dlls

I’ve also added an UI for assigning per-monitor flurries. You can now have a different flurry on each of your monitors.

It has an updated configuration UI that exposes some existing options: an FPS indicator, the possibility to use a "Block Mode" or a "Whiteout Mode" and a way to shrink the flurry window to use only a percentage of the fullscreen window.

Under the hood, there has been quite a few changes with some heavy refactoring to prepare for the flurry editor.

Download from Wincustomize or Download from this site

Full changelog:

Version 1.3.1 Build 41 (07/21/08)
    - Added per-monitor flurries
        * Separated global preset and per-monitor preset
        * Simple UI for assigning flurries to monitors
    - Added flurry editor
        * Presets are saved directly after creation/modification. Deletion is effective when closing the main dialog.
    - Bug fixed
        * Preset names were not stored correctly in registry
        * Fixed deletion of preset (UI was not updated correctly)
        * Crash if Flurry wasn't installed before on the computer

Version 1.3.0 Build 26 (08/10/08)
   - UI Changes
     * Moved credits to About window
     * Added UI for shrink percentage & FPS indicator
       (only in single-buffer mode)
     * Added UI for Block & Whiteout modes
   - Editor
     * Added new/edit/delete buttons
     * Added Editor dialog (not yet functional)
   - Refactoring of code
     * separated settings in a class (preparing for per-monitor flurries)
     * Removed some global variables
     * cleanup/rename: all classes are now in the Flurry namespace
     * Updated Spec class (copy constructor, misc. updates) in preparation
       for editor
   - Bug fixed
     * Preview was not showing the selected preset correctly	 

Changes from the previous version:

• New “Heart” particle texture
• Blur is disabled by default (only for first time install)
• Toned down the blurring   
• Fixed configuration window (no more reloading of the Dream on configuration change)
• Fixed reloading of textures
• Fixed behavior on higher quality settings
• Fixed crash in update when the dream was not starting correctly

Since the last update, most of the work has been directed towards the configuration UI. Loading and saving settings is now fully implemented. I also added a slider to modify the size of the particles.

What’s not working correctly yet:

• Applying a new configuration is causing the dream to reset.
• The colorize shader still needs some work to make it use the particle color (the color is hard-coded right now).
• More particle textures are needed
• The colors need to be tweaked
• The "scaling" code (to adapt to the different quality settings) is not working properly.
• A few potential bugs in the config code

You can download a zip file containing the dll files and the resources, as well as a test program named TestDream that will load the dream inside it’s own window. Simply open a command prompt and launch it with "TestDream.exe Valentine32.dll".

There is also a .Dream file inside, but remember that it hasn’t be tested extensively, so use it with caution (make sure to take a look at the readme file).

Download: Valentine Dream Test Build

The configuration is loaded from the registry at startup (and initialized if the dream is started for the first time). You can tweak it by changing the registry keys as the UI is not functional at the moment.

Once I have the config UI working I will make a test build available (probably not a .Dream at first) so that you can play with it a little

(Background picture is Valentine Favorites by craftilyeverafter - Particle tutorial by Almar Joling was a great help)

It will use some particle system to create the heart itself. You will be able to change the color and texture used for the particles. The background will be chosen among a few different pictures, with a shader applied for blur & coloration.

Things to do:

• load/save configuration
• configuration UI
• Particle "engine"
• Background, with pixel shader for color & blur

The plan is to get a first test build as soon as possible to get some feedback, with a pre-release build on February 8 for final tweaking. I expect to make a final release on February 12.

I need to find a few nice backgrounds (flick is my friend here :D). Color selection for the heart will probably be limited to magenta and blue, with 2/3 different particle textures.

As you can see from the two pictures, I’m not really good at sketching I will try to get some screenshots of the real UI soon, so you can get a better idea of how it will look like.

BTW, for people wanting to create dynamic dreams, I can’t really release any source code right now, but you can start by downloading the December 2006 Microsoft DirectX SDK (the version is important!). Creating dynamic dreams is a lot like creating screensavers: get the CD3DScreensaver class that is part of DXUtil and start hacking on it. Converting that work to a dynamic dream should be relatively straightforward once the Dream SDK is released.