[linux-audio-dev] Generic audio file handling

Conrads announcement just reminded me, but a freind of mine sugessted this for Linux months ago (I think he nicked it off the way Amigas filetype-drivers did generic file i/o) and I've been meaning to post about it ever since. The Idea Have .so files (similar to LADSPA plugins, but different API) that contain functions that can be called to inspect audio files on disk to determine if they understand them or not. If they do a second function is called to stream the audio from the file (in a single standard format, eg. IEEE float or double). These .so files would just be thin wrappers over existing libraries, libsndfile, libfishsound, libFLAC etc. you could even wrap up General MIDI file renderers as an extreme example. The same could be done for writing, but thats probably less useful, and harder. The Motivation It means that if your app only cares about audio i/o a little it can have a single, simple .so file handling routine that scans the installed plugins offfering them all the file, pick one that recognises the filetype and asks it for the data. If you need anything complex - like extracting loop points from metadata streams then you will still have to use a library directly, but for simple "get me this audio" tasks this should be fine. Users can just install new plugins in a global directory and all the apps that use this scheme will then be able to read the new filetype. The API The functions would be something like: "get metadata" -- pass a fd, FILE * or filename and get back a struct describing the contents of the file (length, no. of channels etc.), or NULL if its not recognised by this plugin "get data" -- pass the file reference, some preallocated float buffers, and a frame count and they will be filled. returns how many frames were actually written I havent actually written any code, so you might need more API cruft than that, but this is the basic idea. - Steve