Pythonic way to iterate through multidimensional space?
gelonida at gmail.com
Wed Oct 8 12:48:23 CEST 2014
On 10/7/2014 1:01 PM, Ned Batchelder wrote:
> On 10/7/14 2:10 AM, Gelonida N wrote:
>> Disadvantage of itertools.product() is, that it makes a copy in memory.
>> Reason ist, that itertools also makes products of generators (meaning of
>> objects, that one can't iterate several times through)
>> There are two use cases, that I occasionaly stumble over:
>> One is making the product over lists(),
>> product( list_of_lists )
>> product( [ [1,2,3], ['A','B'], ['a', 'b', 'c'] ] )
>> the other one making a product over a list of functions, which will
>> create generators
>> product( [ lambda: [ 'A', 'B' ], lambda: xrange(3) ] )
>> I personally would also be interested in a fast generic solution that
>> can iterate through an N-dimensional array and which does not duplicate
>> the memory or iterate through a list of generator-factories or however
>> this would be called.
> itertools.product makes a copy of the sequences passed in, but it is a
> shallow copy. It doesn't copy the objects in the sequences. It also
> doesn't store the entire product.
> If you are calling product(j, k, l, m, n), where len(j)==J, the extra
> memory is J+K+L+M+N, which is much smaller than the number of iterations
> product will produce. Are you sure that much extra memory use is a
> problem? How large are your lists that you are product'ing together?
Thanks for the clarification.
You are right. Consumption of a shallow copy of each iterator, should
not be a problem in the cases that I encountered so far.
> I don't understand your point about a list of functions that create
> generators? What is the problem there?
The idea was to even avoid the creation of a shallow copy, by having a
function, that will return the same generator over and over again (thus
no need for shallow copy)
More information about the Python-list