0

I am processing a waveform stored in a vector. The audio channels are interleaved. I have to process the channels separately and then provide the final result as interleaved (again). The functions that operate on every single channel accept a range delimited by random access iterators, and assume the range is continuous.

As these operations have to be performed real-time (or sort of), I'd like to fake the deinterleave phase: in other terms, I am looking for a way to make the single-channel functions operate on a particular channel without actually deinterleaving anything and without affecting their current code which assumes that the ++ (or --) operator of the iterators moves to the next (or previous) element.

What do you suggest? I am currently thinking about writing a custom random access iterator. Is there any other viable solution? I'd prefer not to use boost.

gd1
  • 11,300
  • 7
  • 49
  • 88

1 Answers1

1

You should provide a custom iterator which implements operator++() and operator--() w/r/t the number of channels to keep the interleaving. You can do this without using boost.

However, you'll also have to adapt the vector class to use the new iterators. So the current code has to change by means of using the new vector class.

gd1
  • 11,300
  • 7
  • 49
  • 88
Sebastian
  • 8,046
  • 2
  • 34
  • 58
  • The current code does not assume it is operating on `std::vector` instances. Fortunately, it is made of STL-like functions with accept random access iterator ranges. – gd1 Jan 01 '14 at 18:28
  • 1
    I think having a custom iterator is the best option. It hides implementation details from the user and does what he wants. – Sebastian Jan 01 '14 at 18:43