It’s really not as exciting as it sounds unfortunately 🙂 I’ve been quiet on the subject for some time, though, and wanted to put out some information.
For a long time I debated whether I should fully refactor Paint.NET, or do it from scratch. I finally decided on the latter, and have finally been able to make some progress. Bear in mind that "from scratch" means that I started with a brand new, empty solution file in Visual Studio. It does not mean I’m throwing out all the current Paint.NET code. There are many places where I will be able to reuse vast amounts of it — rendering kernels, effects, math, algorithms, etc. The entire structure of the program will be different, however. If you demolish your old house to build a new one, you don’t throw out the grand piano in the living room too. It still has plenty of value.
The screenshot above is from a build that so far contains about 12,000 lines of code. This sounds like an exorbitant amount for a few menus and an About dialog, but hear me out. I’m building it using Inversion-of-Control (IoC) and related patterns, which requires a lot of new decisions to be made regarding assembly layout (layering, essentially). I’ve also got a new system for asynchronous programming, separation of UI and logic, separation of "contract" and "implementation" assemblies (this becomes very important when plugins come into play), command routing and handling, etc. Along the way I’m learning more about .NET 3.5 and C# 3.0, and have come across some wickedly cool things that are now possible.
As far as possible, everything is lazy initialized or loaded in Paint.NET v4. To aid this, I have a new asynchronous programming model and exception handling pattern, which I could probably talk about for hours. For example, in the screenshot above, the icons for the menu items are not loaded until you click on the menu — in fact, they are loaded using background threads (the threadpool). The graphics and rich text that populate the About dialog are also loaded from the resource manager service using a background thread. For these specific cases, there are no major performance requirements that necessitate this; however, it’s much easier to solve these simple cases and to establish good patterns for later, than the reverse. I’ve also got an old Pentium 4 that I now have available for baseline performance testing, and it will be interesting to see if I can keep the program as responsive on that system as on my Core 2 Quad — even if rendering kernels take longer to complete, it should still always immediately respond to input. *crosses fingers* (Anyone else get sick of Safari on iPhone not being responsive while loading some web pages? Yeah.)
Also, in the screenshot you may notice that the mouse cursor is resting in the File menu even though the About dialog is open. In Paint.NET 3, all dialogs are modal and you can’t do anything with the rest of the program while one is open. In Paint.NET 4, I’m trying to remove this as much as possible and as far as it makes sense to. For example, why should an update download prevent you from using the rest of the program? Why can’t you switch from image 1 to image 2 while a compute-intensive (aka, "takes a long time") effect is rendering on image 1? Why can’t you interact with the canvas while an effect dialog is open? Or change the selected colors?
Anyway, that’s all for now. I estimate that Paint.NET v4 will be 200,000 lines-of-code when it is done, which places me at about 6% complete.