I have repeatedly heard that using a numpy array on large arrays has better performance over using a python list, but I cannot find a concise explanation to what circumstances it needs to see improvement. Here is what I want to know: What operations/functions actually compute faster than with a list How does size affect performance? Numpy is apparently faster with bigger sizes, but where is the cutoff? (also how does the cutoff change with multidimensional arrays) Some example code problems I am encountering def problem1(list): #list["string", size: 2.000.00] for i in list: # accessing is the biggest slowdown. # Best times so far: List < Numpy.array < Pandas.Series val1 = list if i != 0: val2 = list[i-1] if i >= 10: val3 = list[i-10] do_something(val1, val2, val3) def problem2(list) #list["string", size: 2.000.00] # List operations. # Best times so far: Pandas.Series < List < Numpy.array list.string.slice(x, y) list.shift(x) == list list.apply(func) def problem3(list1, list2) #list1["string", size: 2.000.00] #list2[special, size: 100.000] # A LOT of searches, and a lot of getting and setting. # Best times so far (for list2; might be split up between the search/index and set/get): # (dict, Numpy.array) < (Pandas.Index, dict) < (Pandas.Index, Numpy.array) < (list, Pandas.Series) < (Pandas.Series) for i in list1: seperate_function(i) def seperate_function if i in list2: val_new = list2.index(i) if list2 == val1: list2 = val2 else: # do something else: # do something Continue reading...
