[Numpy-discussion] finding close together points.
Thu Nov 12 12:09:36 CST 2009
On Thu, Nov 12, 2009 at 11:39, Charles R Harris
> On Thu, Nov 12, 2009 at 10:32 AM, Christopher Barker <Chris.Barker@noaa.gov>
>> David Cournapeau wrote:
>> > I would love having a core C library of containers -
>> I'm all for that. However, I think that a very, very common need is
>> simply for a growable numpy array.
>> It seems this would actually be pretty darn easy (again, for someone
>> familiar with the code!).
>> IIUC, it would be exactly like a regular old ndarray, except:
>> 1) The data block could be larger than the current size of the array:
>> - this would require an additional attribute to carry that size
>> 2) There would be an "append" method, or some other name:
>> - this would increase the size of the array, manipulating
>> .dimensions, and, if needed, doing a reallocation of the data pointer.
>> - This is not very complex code, really, you simply allocate more
>> than required, choosing some (user-resettable?) scale: 25% more than
>> before, or whatever. re-allocating memory is actually pretty cheap these
>> days, so it's not all that critical how you do it.
>> 3) we'd probably want a .fit() method that would resize the data block
>> to fit the current size of the array.
>> In fact, as I write this, I wonder if these features could simply be
>> added to the regular old ndarray -- if not used, they wouldn't be much
> I'm rapidly losing interest here. Perhaps you could supply some code
> implementing this new array? Why not just a class using an array that
> doubles the array size when an index is out of bounds and copies over the
> old data. That is pretty much what realloc does. As to python lists, do you
> have any benchmarks showing how bad python lists are compared to arrays?
Didn't we already do this?
"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
-- Umberto Eco
More information about the NumPy-Discussion