1
\$\begingroup\$

This question has a second part here:

Comparing columns from two CSV files - follow-up

I'm currently trying to compare two CSV files column by column but only when their indexes match.

Here is what I've done:

import sys

def get_column(columns, name):
    count = 0
    columns = columns.split(';')
    for column in columns:
        array = list(column)
        if(array[0] == '"'):
            array[0] = ''
        if(array[len(column) - 1] == '"'):
            array[len(column) - 1] = ''
        column = ''.join(array)
        if column != name:
            count += 1
        else:
            return(count)

def set_up_file(file, variable):
    columns = next(file)
    siren_pos = get_column(columns, 'INDEX1')
    nic_pos = get_column(columns, 'INDEX2')
    variable_pos = get_column(columns, variable)
    return(siren_pos, nic_pos, variable_pos)

def test_variable(variable):
    with open('source.csv', 'r') as source:
        sir_s, nic_s, comp_s = set_up_file(source, variable)
        with open('tested.csv', 'r') as tested:
            sir_t, nic_t, comp_t = set_up_file(tested, variable)
            correct = 0
            memory = 0
            for row_s in source:
                row_s = row_s.split(';')
                tested.seek(0, 0)
                count = 0
                for row_t in tested:
                    count += 1
                    if(count >= memory):
                        row_t = row_t.split(';')
                        if(row_s[sir_s] == row_t[sir_t] and row_s[nic_s] == row_t[nic_t]):
                            if(row_s[comp_s] == row_t[comp_t]):
                                correct += 1
                                memory = count
                            break
            tested.seek(0, 0)
            print(correct / sum(1 for line in tested) * 100)

def main():
    test_variable('VARIABLE_TO_COMPARE')


if(__name__ == "__main__"):
    main()

This script is running into some performance issue and I would like to know how to make it run faster / make it more readable.

\$\endgroup\$
2
\$\begingroup\$

First I'd like to call out the good things that you've done:

  • Writing functions, including a standard main().
  • Having more or less sensible function and variable names.
  • Proper use of with.

An improvement here is to stop parsing the file yourself, and to start parsing it with Python's native csv library. Even though your format is not technically CSV (it's separated by semicolon), you can still configure csv to use a different delimiter.

I recommend the use of the DictReader class. Given that you only pay attention to one variable, just use csv.reader. During initialization, get the index of the column you want, and then use that on every record that the reader gives back.

The critical performance issue here is that you have nested loops for row comparison. Given that you are comparing two series expected to be in the same order, but with edits (insertions and deletions), effectively you're doing a diff. Read this for a nice walkthrough.

http://www.xmailserver.org/diff2.pdf

\$\endgroup\$
  • \$\begingroup\$ Indeed, changing the separator allowed me to use the csv module. However I can't use zip because the files are not containing exactly the same lines so it won't line up correctly \$\endgroup\$ – Comte_Zero Nov 21 '18 at 14:38
  • \$\begingroup\$ @Comte_Zero Are both in the same order, with some random insertions and deletions? Or are the files the same size, but out of order with respect to each other? \$\endgroup\$ – Reinderien Nov 21 '18 at 14:41
  • \$\begingroup\$ Same order, random insertions and deletions \$\endgroup\$ – Comte_Zero Nov 21 '18 at 14:44
  • \$\begingroup\$ And are you checking all of the columns in both files, or only some of the columns? \$\endgroup\$ – Reinderien Nov 21 '18 at 14:45
  • \$\begingroup\$ This may vary, that is why the main function contains only one, potentially copiable line \$\endgroup\$ – Comte_Zero Nov 21 '18 at 14:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.