A more efficient code

[titanif]

I just started with Python. I want to add the information of the array ‘toc’ to the array ‘toc_n.’ The first two columns of both ones are ‘year’ and ‘day of the year,’ but the array ‘toc’ lacks some days (i.e., fewer rows). I wrote the following code. Could you please suggest something more efficient/clever?

for i in range(2196):
for j in range(2173):
dd = np.logical_and(toc_n[i,0] == toc[j,0] , toc_n[i,1] == toc[j,1])
if dd == True:
toc_n[i,2]=toc[j,2]

buran write Oct-16-2023, 11:06 AM:
Please, use proper tags when post code, traceback, output, etc.
See BBcode help for more info.

[Larz60+]

would numpy.concatenate work here?

[deanhystad]

I don't think concatenate will work. An odd combination of matching and replacing makes me think that there is not a vectorized solution. In pandas you could make a date intex and merge, but that adds a column instead of replacing values in an existing column. The code below does he matching and replacement in a single pass. It does require that the arrays are sorted by date.

import numpy as np

# make two Nx3 arrays to work with.
toc_n = np.array([[a, b, 0] for a in range(1492, 2023) for b in range(1, 366)])
toc = np.array([[a, b, 1] for a in range(1492, 2023) for b in range(1, 366, 3)])

# Combine year and day into a date value.  If not sorted, sort arrays by
# this value.  Sorting is an expensive operation, but the numpy sort is
# efficient compared to looping in python.
toc_n = np.hstack((toc_n, (toc_n[:, 0]*365+toc_n[:, 1]).reshape(-1, 1)))
toc = np.hstack((toc, (toc[:, 0]*365+toc[:, 1]).reshape(-1, 1)))

# Loop through toc_n and toc, copying toc[2] to toc_n[2] when
# the dates match.  Since both arrays are sorted this can be
# done in a single pass.
try:
    toc_iter = iter(toc)
    t = next(toc_iter)
    for t_n in toc_n:
        while t[-1] < t_n[-1]:
            t = next(toc_iter)
        if t[-1] == t_n[-1]:
            t_n[2] = t[2]
            t = next(toc_iter)
except StopIteration:
    pass