I have a function f([i_0, i_1, ..., i_k-1]) which takes as input a k-dimensional array of integers and returns some object.
Given the range for each index (as a two-dimensional array ranges=[i_0_range, i_1_range, ...]), how do I generate a k-dimensional list / array containing objects evaluated for each value of indices?
If k was fixed, I'd simply do k nested loops. But I would like to have a solution working for any k. How can I do this in Python?
You can use itertools.product to generate all the different combinations of indexes from the ranges. You can then iterate over the tuples produced by that iterator, calling f for each tuple. For example, if f is defined to return a string of the input indexes:
import itertools
def f(indexes):
return ','.join(map(str, indexes))
ranges = [range(0, 2), range(1, 3), range(2, 4)]
objs = [f(list(t)) for t in itertools.product(*ranges)]
print(objs)
Output:
['0,1,2', '0,1,3', '0,2,2', '0,2,3', '1,1,2', '1,1,3', '1,2,2', '1,2,3']
Note that dependent on your implementation of f, it might not be necessary to convert the returned tuple from itertools.product to a list and you could just use f(t) instead of f(list(t)).
itertools.product(*ranges) does, and why you called f with [*t] argument? (Sorry, apparently I'm not familiar with these Python basics.)
*ranges flattens the ranges array, so that each element is passed separately to itertools.product, which then creates the cartesian product of each of the ranges. Note that I changed the code to be f(list(t)) for t in itertools.product(*ranges) so that it would correctly deal with k greater than 2; itertools.product returns a tuple of values (for i_0, i_1, ..., i_k), list(t) then converts that into a list as described in your question. Dependent on your implementation of f, it might not be necessary to convert to a list and you could just use f(t)