[Audacity-nyquist] Output stream?

[Ryan H. <rr...@ho...>]

I've got an idea for a plug in. I have almost no idea what it will sound like, but I can describe what it will do.

It will run through each sample, (probably using snd-fetch) and find each point where the sample crosses zero. It will also have been keeping track of the peak. Each time it reaches a zero crossing, it will output a half-wave scaled to the dimensions specified by the peak and the zero points on either side of it.

Here's a picture. Blue is before, red is after.
http://horriblepain.com/quicks/graph.gif

But the trick is, everything I've done so far has involved transforming the sound as a whole. I know I can treat snd-fetch kind of like an input stream, taking in samples of sound sequentially, but I'm not clear on how I would output the results, one segment at a time, as it is calculated.

Or is this a sign of imperative language bias, and I should be approaching this in an entirely different way?