From: Todd A. P. Ph.D. <tp...@ac...> - 2001-11-19 13:57:20
|
Thanks for all of your work. Things seem to be shaping up nicely. I just wanted to second some of the concerns below: > Complex Types: > ============== > > 1) I don't like the idea of complex types being a separate subclass of > ndarray. This makes them "different." Unless this "difference" can be > completely hidden (which I doubt), I would prefer complex types to be on the > same level as other numeric types. > > 2) Also, in your C-API, you have a different pointer to the imaginary data. > I much prefer the way it is done currently to have complex numbers > represented as an 8-byte, or 16-byte chunk of contiguous memory. > The second comment above is really critical for accessing utility available in a very large number of numerical libraries. In my view this would "break" the utility of numpy severely -- recopying arrays both on the way out and the way in would be extremely cumbersome. -Todd Alan Pitts |