For some econometric work.
I often need to derive multiple parallel arrays of calculated variables given a (potentially) large number of parallel data arrays.
In the following example, I have two input arrays and two output arrays, but imagine in the real world there could by anywhere from 5-10 input and output arrays.
w, x are inputs
y, z are outputs
Method A:
w = [1, -2, 5]
x = [0, 3, 2]
N = len(w)
I = range(N)
y = map(lambda i: w[i] + x[i], I)
z = map(lambda i: w[i] - x[i], I)
Method B:
w = [1, -2, 5]
x = [0, 3, 2]
N = len(w)
I = range(N)
y, z = [], []
for i in I:
y.append(w[i] + x[i])
z.append(w[i] - x[i])
Method C:
w = [1, -2, 5]
x = [0, 3, 2]
y, z = [], []
for w_i, x_i in zip(w, x):
y.append(w_i + x_i)
z.append(w_i - x_i)
Method D:
w = [1, -2, 5]
x = [0, 3, 2]
N = len(w)
I = range(N)
(y, z) = transpose(map(lambda i: [w[i] + x[i], w[i] - x[i]], I))
D seems to be the most concise, extendable, and efficient. But it's also the most difficult to read, especially with many variables with complicated formulae.
A is my favorite, with a little duplication, but is it less efficient to construct a loop per vairable? Will this not scale with large data?
B vs. C: I know C is more pythonic but B seems more convenient and concise, and scales better with more variables. In both cases, I hate the extra line where I have to declare the variables up-front.
Overall, I am not perfectly satisfied with any of the above approaches. Is there something missing from my reasoning or is there a better method out there?
use numpy ... that performs the operations in C++ so its much faster ... (especially if we assume your arrays are much bigger than 3 items)
w = numpy.array([1, -2, 5])
x = numpy.array([0, 3, 2])
y = w+x
z = w-x
i think @Beasley's suggestion works well, and i suggest using multiprocessing
on top of it so that the output generation is in parallel. your computation seems perfectly parallelizable!
what i can offer can't beat the tips discussed on here: Does python support multiprocessor/multicore programming?
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.