[Tutor] initialising all elements of a matrix

[Peter Otten]

Sivaram Neelakantan wrote:

> 
> I was wondering whether there is a faster/better/cleaner way of
> element wise operations of arbitrarily nested list.  I wrote something
> like this for 1 level nested lists and am wondering whether there are
> any good idioms in python.  I did this after a ridiculous amount of
> bad thinking/missteps in python for the simplest of cases.
> 
> def init_p (arr):
>     # input is always 2D matrix; init to uniform probability dist.
>     q = []
>     row = len(arr)
>     col = len(arr[0])
>     uni_dist = 1.0/(row *col)
>     q = [ [uni_dist] * col for i in range(row)]
>     return q
> 
> Of course, without using external packages like numpy or any other
>  scientific packages.

Here's how I would break up the problem:

# untested
def nested_list(shape, value):
    if len(shape) == 1:
        return [value] * shape[0]
    return [nested_list(shape[1:], value) for _ in range(shape[0])]

def get_shape(arr):
    shape = []
    while isinstance(arr, list):
        shape.append(len(arr))
        arr = arr[0]
    return shape

def product(factors, product=1):
    for factor in factors:
        product *= factor
    return product

def init_p(arr):
    shape = get_shape(arr)
    value = 1.0 / product(shape)
    return nested_list(shape, value)

Now you can proceed to improve the parts independently...