As I understand, the list
type in Python is a dynamic pointer array, which will increase it's capacity when items are appended to it. And an array in NumPy uses a continuous memory area to hold all the data of the array.
Are there any types that dynamically increase its capacity as a list, and stores the value as a NumPy array? Something like List in C#. And it's great if the type has the same interface as a NumPy array.
I can create a class which wraps a NumPy array inside, and resize this array when it's full, such as:
class DynamicArray(object):
def __init__(self):
self._data = np.zeros(100)
self._size = 0
def get_data(self):
return self._data[:self._size]
def append(self, value):
if len(self._data) == self._size:
self._data = np.resize(self._data, int(len(self._data)*1.25))
self._data[self._size] = value
self._size += 1
but DynamicArray can't be used as a NumPy array, and I think all the views returned by get_data() before np.resize() will hold the old array.
Edit: array type in array module is dynamic array. The following program test the increase factor of list and array:
from array import array
import time
import numpy as np
import pylab as pl
def test_time(func):
arrs = [func() for i in xrange(2000)]
t = []
for i in xrange(2000):
start = time.clock()
for a in arrs:
a.append(i)
t.append(time.clock()-start)
return np.array(t)
t_list = test_time(lambda:[])
t_array = test_time(lambda:array("d"))
pl.subplot(211)
pl.plot(t_list, label="list")
pl.plot(t_array, label="array")
pl.legend()
pl.subplot(212)
pl.plot(np.where(t_list>2*np.median(t_list))[0])
pl.plot(np.where(t_array>2*np.median(t_array))[0])
pl.show()
from the graph: the increase factor of list is bigger than array.
See Question&Answers more detail:
os 与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…