Edit function to work for arrays instead of lists

Question

I have a function that does the following:

Takes uses the list as an input
Computes the sum of the list
Appends the output (the sum) to the input list
Runs itself again (this time with a list of N + 1 data).

I want to to do this exact thing for an input of a numpy array, but I don’t get a complete solution.

This is what I currently have:

data = [15.5, 19.2, 27.8, 44.6, 71.0, 54.1, 60.2]


def AppendOutput(data, steps):
    
    for i in range(steps):
        s = sum(data)
        emp.append(s)
        data.append(emp)
    return data


s = AppendOutput(data, 3)

This outputs the following, which is what I want:

[15.5, 19.2, 27.8, 44.6, 71.0, 54.1, 60.2, 292.4, 584.8, 1169.6]

My current issue is that I’m trying to accomplish the same thing for an input that is a numpy array.

data_array = np.array([data])

def AppendOutputNP(data, steps):
    for j in range(steps):
        temp = np.sum(data_array).reshape((-1,1))
        new = np.append(data_array, temp, axis=1)
    return new

AppendOutputNP(data_array, 3)

I get a partial solution in comparison to what I would get for the original function; this time only one element gets appended instead of three:

array([[ 15.5,  19.2,  27.8,  44.6,  71. ,  54.1,  60.2, 292.4]])

Could anyone direct me as to what I’m doing wrong here?

Asked By: Volti

||

Source

Answer 1

It’s probably best to form the result array at the start of the function. numpy.append is quite slow.

def AppendSumNP(data, steps, dtype = np.float64 ):
    result = np.zeros( len(data) + steps, dtype = dtype ) 
    # The result array is big enough for the whole result
    result[ :len(data) ] = data
    for j in range( len(data), len(data) + steps ):
        result[ j ] = result[:j].sum()
    return result

Also the second sum is twice the first and the third twice the second, etc. so only sum once, then multiply.

def AppendMultNP(data, steps, dtype = np.float64 ):
    result = np.zeros( len(data) + steps, dtype = dtype )
    result[ :len(data)] = data
    total = result[:len(data)].sum()
    for j in range( len(data), len(data) + steps ):
        result[ j ] = total
        total *= 2
    return result

It depends how long the data list is but the Mult approach can be several times as fast.

Answered By: Tls Chris

Edit function to work for arrays instead of lists

Question:

Answers: