Friday, December 22, 2006

Goals of wxWidgets

A comment attached to the last post asked the question about what wxWidgets goals are. I was initially quite surprized because the answer was so obvious to me, but thinking a bit more about this I realized that we don't clearly answer this question anywhere in the public documentation and so for someone not following wx-dev mailing list the direction of wxWidgets development may indeed seem mysterious. To help with understanding it, I'd like to briefly explain where wxWidgets is coming from, what are the guiding principles for its development today and what I hope it is going to be tomorrow. Please note that, as anyone who does follow wx-dev would know, we don't always agree about everything and so what follows is just my personal point of view and while it hopefully does overlap with that of the others -- at least for the past and present parts -- I don't pretend to speak for everybody.

This disclaimer aside, let's start by looking at the current state of wxWidgets and how did we get there. On the negative, and often mentioned, side, it's an accretion of almost 15 years of development, by different people, some of which have stopped contributing to the project since a long time. So it is not surprising that there are some inconsistencies in the API (although we work hard on ironing them out) and some components (notably those whose maintainers have left) are not as good as one'd expect them to be. It is also true that compatibility with MFC seemed much more important in the nineties than now and this explains why wxWidgets has some superficial similarities to it, although it has never been nor designed to be, an MFC clone. So it's true that this long history resulted in some problems and we had to change quite a few things to fix them -- even the original name of the library turned out to be a bad idea.

However, on the bright side, not everything we did during these 15 years was a complete waste of time and, in addition to the accumulated historical luggage, everything that we did right is in the wxWidgets codebase too. And this is quite enough to allow people to create real life, professional complicated graphical cross-platform applications -- which is, to answer the initial question, what wxWidgets is about. To be more precise, the main goal is to allow the developers to do everything they need to do without resorting to writing platform-specific code: the "Write once, compile anywhere" principle. Of course, this goal is a moving target which can never be quite achieved because developers discover more and more things that they need to do with time as the new graphical environments introduce new standards (use of gradients and animations in the GUI was not exactly common when wxWidgets was started) but, still, I believe we're pretty close to it -- and getting closer with every new release.

So wxWidgets primary goal is to allow creating sophisticated portable applications without writing platform-specific code. This is not the only goal, of course, but it's the most important one and so trumps all the others. For example, wxWidgets could be more size efficient but this would be mostly important for toy applications and not for the programs which are themselves of reasonably important size. And so, while we'd be as glad to make wxWidgets smaller as anybody else, this currently has a relatively low priority because it's not very important from the main goal perspective. On a perhaps more constructive note, this also explains why do we consider the differences in behaviour of wxWidgets on different platforms important bugs: it's useless to have the code which compiles on all platforms if it doesn't behave the same everywhere during run-time.

But, to repeat, as important as this goal is, I think we're quite close to it and have been there since a long time. However other things could really be improved. Chief among them is the ease of use of wxWidgets: things have changed a lot since its creation and, in particular, modern C++ doesn't have much to do with C++ of the nineties. So it would be great to make wxWidgets evolve with the language and other standard C++ libraries. This would mean providing a more modern API, removing the need for memory management completely (this is very rarely a problem in wxWidgets even now, as most of the objects are owned by the library anyhow, but it would be nice to make this more explicit) and improve error handling by using exceptions. We should also rely more on the standard C++ library and Boost instead of implementing all various useful but not directly GUI-related classes on our own as we had to do in the past (as not implementing them would have conflicted with our primary goal whereas now, when such classes are available from elsewhere, it doesn't any more). This, and much more, is in our plans for wxWidgets 3.

There are certainly a lot of other tactical goals. The fact that I didn't mention at all fixing the bugs or improving the documentation doesn't mean that these goals are not important but just that they are obvious and I don't want to make this article even longer than it already is by stating the obvious. The point of this post was to just say that strategically, our main goal always was and remains making wxWidgets suitable for real-life portable applications development and, in addition, creating a new, modern, API for wxWidgets 3 in (hopefully near) future.
Post a Comment