UI design

Saw this very interesting UI-video today called Away with Applications: The Death of the Desktop by Aza Raskin (part of the google tech talks). It shows an interesting application enso, very similar to quicksilver. The interesting part is of course the abstract (what it is a solution to) and the conclusion (and why). Especially things like application silos and how enso marries the CLI with the UI gives some food for thought.

Finally skip-free audio playback in Linux

Stumpled upon this story the other day. Finally a tool where you can actually decide which things get prioritized instead of the kernel relying on heuristiscs. I did a test with alsaplayer playing back using through jackd (very sensitive to not getting enough time slices). To stress the system I did a emerge sync together with a fetchmail process sucking down the days mail. I don’t really care if it takes 5 seconds longer to fetch my mail or if it takes 2 minutes long to emerge sync as long as I can be sure that my music doesn’t skip. And guess what? With DeskOpt you can :-)

Boost part 2

In the quest for improving the quality of the MMS code in 1.1.0 I’ve done several things, one of which was to convert all the source to the boost foreach statement. Since I liked the result of the foreach cleanup so much, I decided to see what other libraries could be of use. I quickly found the function and bind library which allowed me to remove the internal function pointers framework we have currently been using. While converted I was surprised to see how cleaner the code became. The bind function is truly easy to use and read and made it possible to remove several helper classes which only purpose was to capture state to create a uniform interface.

So instead of:

class FP
{
  std::list numbers;

  FP(std::list n)
    : numbers(n)
  {}

  call(int t)
  {
    int total;
    foreach (int i, numbers)
      total += i*t;
    return total;
  }
};

(I saved the part of actually creating the function pointer object). One gets:

call(int t, std::list numbers)
{
  int total;
  foreach (int i, numbers)
    total += i*t;
  return total;
}

boost::bind(&call, _1, list_of_numbers);

Voila. Now if only the wordpress code formatting didn’t suck so much ;-)

The future of C++

Recently I’ve been looking a lot into boost and it’s really a great set of libraries. Although the syntax of some of the libraries could use a helping hand (assign library). Luckily I was watching a presentation by Bjarne Stroustrup on the next C++ standard and it appears that they will finally add a way to construct containers such as vectors with elements as construction time. I think it was called initializer lists and the syntax was the following:

      vector v {1,2,3}

So you can now initialize them just like regular arrays :-) Furthermore it appears that we’ll get threads and perhaps a common filesystem + network abstraction. Now if only they could be a little quicker at bringing forward these new standards ;-)

It’s all the little things

Having been coding in c# for some time now, what is really nice about the language compared to c++ is the foreach statement. The syntax is really natural and easy to read:

     foreach (string s in strings)

Compare this to the standard c++ notation:

     for (vector::const_iterator i = strings.begin(),
             iend = strings.end(); i != iend; ++i)

So I decided there must be something easier and vague remembered that boost had something for this. And low and behold they had. So now instead of writing that horrible for line, one can just write:

     foreach (const string& s, strings)

It’s not quite as nice as i c# since one can’t have a pair on the leftside of the comma, so one needs to be using typedefs to fix that. And one must also remember the & as to not create unneeded copies of the string.

And the best part is that it’s written as a bunch of templates and macro’s so one can just include a header and of it goes. Doesn’t even have to change the Makefile :-)

I’ve converted the whole MMS 1.1.0 code base to the new syntax and it’s really so much easier to read. Especially because I tend to use better names for the loop variable than i :-)