I have a problem about list to array conversion. I have a list from a csv file, like
a=[['1','a'],['2','b']]
Now I only want the first column, the number '1' and '2', and convert them to a numpy array. How do I accomplish this? Using b = np.array(a) put all items as string into the array.
You can use numpy.fromiter with operator.itemgetter. Note a standard NumPy array is not a good choice for mixed types (dtype object), as this will cause all data to be stored in pointers.
a = [['1', 'a'], ['2', 'b']]
from operator import itemgetter
res = np.fromiter(map(itemgetter(0), a), dtype=int)
print(res)
array([1, 2])
Some performance benchmarking:
a = [['1', 'a'], ['2', 'b']] * 10000
%timeit np.fromiter(map(itemgetter(0), a), dtype=int) # 4.31 ms per loop
%timeit np.array(a)[:, 0].astype(int) # 15.1 ms per loop
%timeit np.array([i[0] for i in a]).astype(int) # 8.3 ms per loop
If you need a structured array of mixed types:
x = np.array([(int(i[0]), i[1]) for i in a],
dtype=[('val', 'i4'), ('text', 'S10')])
print(x)
array([(1, b'a'), (2, b'b')],
dtype=[('val', '<i4'), ('text', 'S10')])
You'd first need to create a new list`, that only contains the first values of the lists in a. For example
c = []
for row in a:
c.append(row[0])
b = np.array(c)
More Pythonic would probably be a list comprehension:
c = [x[0] for x in a]
b = np.array(c)
import numpy as np
a = [['1', 'a'], ['2', 'b']]
print(np.array(a)[:, 0].astype(int))
try this:
a=array([int(i[0]) for i in a])
