[Numpy-discussion] Arithmetic on arrays of pointers
Christopher Barker
Chris.Barker at noaa.gov
Wed Apr 29 13:31:22 EDT 2009
Dan Goodman wrote:
> (because I want to construct the 1D arrays X, Y above from values picked
> from a variety of possible data structures, including dense and sparse
> 2D arrays, but in such a way that the code that uses these values
> doesn't have to know where the values came from in these data
> structures).
I wonder if you could define a new datatype (or set of datatypes) for
numpy: pointer_to_double, etc. What I'm wondering is how you would
construct an array of such types in python, though I suppose you could
do something like:
a = np.array(data, dtype=np.float)
pa = np.empty(size, dytpe=np.p_float)
pa[i] = a[i]
and have it do "the right thing". You might also be able to grab
pointers with ctypes or cython.
> code which involves scheduling events in the future (simple events like
> add a value to another value), but the library is supposed to be easily
> extensible for people who don't know C so I want to make it as easy to
> use as possible.
>
> Well, that's a slightly underdefined problem, basically I just have a
> vague idea that something like the above would be useful to me and I'm
> looking for any ideas that might seem relevant.
Overall, if you want easy to use, and not a lot of segfalts, I suspect
you are better off getting away from pointers in user code, and instead
creating objects for your users to manipulate that grab copies of data
from where they need it, and update the original data when it changes.
-Chris
--
Christopher Barker, Ph.D.
Oceanographer
Emergency Response Division
NOAA/NOS/OR&R (206) 526-6959 voice
7600 Sand Point Way NE (206) 526-6329 fax
Seattle, WA 98115 (206) 526-6317 main reception
Chris.Barker at noaa.gov
More information about the NumPy-Discussion
mailing list