Paul Winkler wrote:
>>
What about storing the length of input in an
environment variable?

let's use the hypothetical scripts "flanger" and "echo".

You might call them like this:
flanger -i input.wav | echo -o output.wav

Here's what happens:
1. "flanger" sees that it has the "-i input.wav" arguments and decides
that it must call sndinfo, parse the result to find lenght of input.wav.
2. Length in sec. is stored in environment as $WAVLENGTH
3. echo sees that it does not have a "-i" argument and thus assumes its
input is from stdin.  So it looks to $WAVLENGTH for the duration to use.
It also modifies $WAVLENGTH if it will be changing the length. This is
so subsequent filters in the pipe get the new length correct.
>>

Another approach is to concatenate the processes directly into a new
Csound orc, thus obviating the need to pass the samples through a pipe.
Then you can mix and match your processes at will and have documentation
in synthetic orcs which you could also use for live processing. (See http://arcana.dartmouth.edu/cgi-bin/eric/CPROC/index.cgi
for an example.)

I don't understand why Csound would require a duration parameter if it is reading
samples from stdin. At most, duration should be an optional parameter to limit how
many samples are processed.