Channel selection API

[LastExceed]

Currently, the stream creation API allows specifying how many channels to use, but not which channels to use. n channels always means 0..n, so if e.g. you wanted a mono stream on the 3rd channel, you would have to

• create a stream with n = 3
• hope that this configuration is even supported
• manually (de-)interleave the data
• (if input) silence the first 2 channels

To solve this, there would need to be some way in the stream builder functions to specify which channels the stream should occupy. A (probably naive) solution would be to change StreamConfig.channels to Vec<usize>, indicating the channel indices.

[Marco-Farruggio]

A related feature I was wondering about, I work primarily with WASAPI, and thus don't know too much about the low level specifics of the other hosts and what they present under the hood, but I know that WASAPI has a channel mask which shows what the physical position of the speaker which that output represents, some drivers don't present this, but it is part of the official WAVEFORMATEXTENSIBLE spec (Win docs on channel mask).

If cpal were to add some api for selecting channels inside the range, (eg instead of just choosing channels [0..y], you could choose [x..y]) then I think implementing this at the same time would make this feel a lot richer.

if not many other hosts than windows implement this, maybe it could be part of the explicit windows trait (like the plan for raw mode) mentioned in the v0.19 goals.

(one last thing did pop to mind, if the driver doesnt support just [x..y], maybe cpal could automatically strip [0..x] from the callback, though idk if that escapes the scope of cpal)