Python, multiprocessing: How to optimize the code? Make the code faster?

Question

I use Python. I have 100 zip files. Each zipfile contains more than 100 xmlfiles. Using the xmlfiles I create csvfiles.

from xml.etree.ElementTree import fromstring
import zipfile
from multiprocessing import Process

def parse_xml_for_csv1(data, writer1):
    root = fromstring(data)
    for node in root.iter('name'):
        writer1.writerow(node.get('value'))

def create_csv1():
    with open('output1.csv', 'w') as f1:
        writer1 = csv.writer(f1)

        for i in range(1, 100):
            z = zipfile.ZipFile('xml' + str(i) + '.zip')
            # z.namelist() contains more than 100 xml files
            for finfo in z.namelist():
                data = z.read(finfo) 
                parse_xml_for_csv1(data, writer1)


def create_csv2():
    with open('output2.csv', 'w') as f2:
        writer2 = csv.writer(f2)

        for i in range(1, 100): 
            ...


if __name__ == "__main__":
    p1 = Process(target=create_csv1)
    p2 = Process(target=create_csv2)
    p1.start()
    p2.start()
    p1.join()
    p2.join()

Please tell me, how to optimize my code? Make the code faster?

Answer 1

You just need to define one method, with parameters. Split the processing of your 100 .zip files across a given number of threads or processes. The more processes you'll add, the more CPU you'll use, and maybe you can use more than 2 processes, it will be faster (there can be a bottleneck because of disk I/O at some point)

In the following code, I can change to 4 or 10 processes, no need to copy/paste code. And it processes different zip files.

Your code processes the same 100 files twice in parallel: it was even slower than if there were no multiprocessing!

def create_csv(start_index,step):
    with open('output{0}.csv'.format(start_index//step), 'w') as f1:
        writer1 = csv.writer(f1)

        for i in range(start_index, start_index+step):
            z = zipfile.ZipFile('xml' + str(i) + '.zip')
            # z.namelist() contains more than 100 xml files
            for finfo in z.namelist():
                data = z.read(finfo)
                parse_xml_for_csv1(data, writer1)



if __name__ == "__main__":
    nb_files = 100
    nb_processes = 2   # raise to 4 or 8 depending on your machine

    step = nb_files//nb_processes
    lp = []
    for start_index in range(1,nb_files,step):
        p = Process(target=create_csv,args=[start_index,step])
        p.start()
        lp.append(p)
    for p in lp:
        p.join()