一种可能的解决方案是使用递归

import numpy as np

def multiplier (first_vector, second_vector, size, index, total):
    if index < size:
        addendum = first_vector[index]*second_vector[index]
        total = total + addendum
        index = index + 1
        # ongoing job
        if index < size:
            multiplier(first_vector, second_vector, size, index, total)
        # job done
        else:
            print("dot product = " + str(total))
    
def main():
    a = np.array([1.5, 2, 3.7])
    b = np.array([3, 4.3, 5])
    print(a, b)

    i = 0
    total_sum = 0

    # check needed if the arrays are not hardcoded
    if a.size == b.size:
        multiplier(a, b, a.size, i, total_sum)

    else:
        print("impossible dot product for arrays with different size")

if __name__== "__main__":
    main()

可能被认为是欺骗，但是Python 3.5添加了一个矩阵乘法运算符，numpy使用它来计算点积，而实际上并没有调用np.dot：

>>> arr1 = np.array([1,2,3])
>>> arr2 = np.array([3,4,5])
>>> arr1 @ arr2
26

问题解决了！