node-spotify development started with vim. At the time I didn’t really know vim and wanted to learn it, actually doing a project with it seemed as a good way. I even tried out a auto completion plugin (I don’t remember the name). At some point node-spotify became too large for me to handle it with vim. Maybe there are a lot of plugins out there which would have helped but I didn’t want to put too much time into it.

That was when I switched to Sublime Text. It was hip at the time, fast and didn’t get in may way. For more than a year now I was very happy with it - only actual code completion was missing (again, there probably is a plugin somewhere that tries to do it). Coming from a Java background with excellent tooling like IntelliJ I certainly missed that convenience.

This week a colleague introduced me to QtCreator. I had searched for a C++ IDE and tried out Eclipse and Code::Blocks, but somehow QtCreator got past me. I heard of Qt before but somehow I remembered it being a paid solution which I didn’t like to do for a hobby project. Turns out I was wrong, it’s free. And it is really great. It doesn’t have nearly as many features as IntelliJ, but certainly enough to satisfy me. So now node-spotify is developed in QtCreator - but of course that should not impact the project.

Progress in the audio code

After a long week of thinking I made some progress in the audio code. Context: I want to provide a way to register the music_delivery callback in Javascript. This means the user must have a way of switching between native audio and Javascript. I started out without thinking too much about this switch and just implemented the Javascript audio part. But I noticed I would have to add many ifs here and there and it was getting really messy.

Some people nowadays think object oriented design is outdated and rather use functional or procedural styles. Again, coming from Java I think it still has its place. There isn’t much object oriented design in node-spotify, just objects. When I noticed I couldn’t switch session callbacks in libspotify while a session is active it hit me - why not actually use objects for this.

class Audiohandler {
  audio_fifo_t* audioFifo;
  static int musicDelivery(...);
  virtual void afterMusicDelivery() = 0;

class NativeAudioHandler : public Audiohandler {};
class NodeAudioHandler : public AudioHandler {};

class Spotify {
  std::unique_ptr<AudioHandler> audioHandler;

Given correct implementations of AudioHandler I can now switch how audio is handled easily at runtime and follow RAII for everything that needs to be done (initializing the fifo, creating a thread for native audio and so on). Again I am very happy with these changes. They provide a much clearer separation of concerns.