1# mode: run
2# tag: pythran, numpy, cpp
3# cython: np_pythran=True
4
5import numpy as np
6cimport numpy as cnp
7
8def diffuse():
9    """
10    >>> u = diffuse()
11    >>> count_non_zero = np.sum(u > 0)
12    >>> 850 < count_non_zero < (2**5) * (2**5) or count_non_zero
13    True
14    """
15    lx, ly = (2**5, 2**5)
16    u = np.zeros([lx, ly], dtype=np.double)
17    u[lx // 2, ly // 2] = 1000.0
18    _diffuse_numpy(u, 50)
19    return u
20
21
22def _diffuse_numpy(cnp.ndarray[double, ndim=2] u, int N):
23    """
24    Apply Numpy matrix for the Forward-Euler Approximation
25    """
26    cdef cnp.ndarray[double, ndim=2] temp = np.zeros_like(u)
27    mu = 0.1
28
29    for n in range(N):
30        temp[1:-1, 1:-1] = u[1:-1, 1:-1] + mu * (
31            u[2:, 1:-1] - 2 * u[1:-1, 1:-1] + u[0:-2, 1:-1] +
32            u[1:-1, 2:] - 2 * u[1:-1, 1:-1] + u[1:-1, 0:-2])
33        u[:, :] = temp[:, :]
34        temp[:, :] = 0.0
35
36
37def calculate_tax(cnp.ndarray[double, ndim=1] d):
38    """
39    >>> mu, sigma = 10.64, .35
40    >>> np.random.seed(1234)
41    >>> d = np.random.lognormal(mu, sigma, 10000)
42    >>> avg = calculate_tax(d)
43    >>> 0.243 < avg < 0.244 or avg  # 0.24342652180085891
44    True
45    """
46    tax_seg1 = d[(d > 256303)] * 0.45 - 16164.53
47    tax_seg2 = d[(d > 54057) & (d <= 256303)] * 0.42 - 8475.44
48    seg3 = d[(d > 13769) & (d <= 54057)] - 13769
49    seg4 = d[(d > 8820) & (d <= 13769)] - 8820
50    prog_seg3 = seg3 * 0.0000022376 + 0.2397
51    prog_seg4 = seg4 * 0.0000100727 + 0.14
52    return (
53        np.sum(tax_seg1) +
54        np.sum(tax_seg2) +
55        np.sum(seg3 * prog_seg3 + 939.57) +
56        np.sum(seg4 * prog_seg4)
57    ) / np.sum(d)
58