Using %lprun: fix the bottleneck

In the previous exercise, you profiled the convert_units() function and saw that the new_hts list comprehension could be a potential bottleneck. Did you notice that the new_wts list comprehension also accounted for a similar percentage of the runtime? This is an indication that you may want to create the new_hts and new_wts objects using a different technique.

Since the height and weight of each hero is stored in a numpy array, you can use array broadcasting rather than list comprehension to convert the heights and weights. This has been implemented in the below function:

def convert_units_broadcast(heroes, heights, weights):

    # Array broadcasting instead of list comprehension
    new_hts = heights * 0.39370
    new_wts = weights * 2.20462

    hero_data = {}

    for i,hero in enumerate(heroes):
        hero_data[hero] = (new_hts[i], new_wts[i])

    return hero_data

Load the line_profiler package into your IPython session. Then, use %lprun to profile the convert_units_broadcast() function acting on your superheroes data. The convert_units_broadcast() function, heroes list, hts array, and wts array have been loaded into your session. After you've finished coding, answer the following question:

What percentage of time is spent on the new_hts array broadcasting line of code relative to the total amount of time spent in the convert_units_broadcast() function?