[SciPy-dev] scipy.io.wav, data format convention

[David Cournapeau]

Hi,

    I wondered what would be the best data representation for audio
data-related functionalities in scipy. Generally, since numpy uses
"applying operation on last axis first", I thought it would make sense
to use the representation of one channel / row. OTOH, I've just found
out that scipy.io.wav does the contrary, and there is also the problem
on how to deal with mono signals (column vector vs array) - as well as
an efficiency question: most audio signals are interleaved (N
consecutive samples in memory for signal with N channels), so the format
of one column/channel is C-contiguous.
    I would like to keep audiolab as close as possible to whatever scipy
will use, but I wondered if anyone thought about the issue at all ?

cheers,

David