Python script to turn input csv columns into output csv row values

Question

I have an input csv that look like

email,trait1,trait2,trait3
foo@gmail,biz,baz,buzz
bar@gmail,bizzy,bazzy,buzzy
foobars@gmail,bizziest,bazziest,buzziest

and I need the output format to look like

Indv,AttrName,AttrValue,Start,End
foo@gmail,"trait1",biz,,,
foo@gmail,"trait2",baz,baz,,
foo@gmail,"trait3",buzz,,,

For each row in my input file I need to write a row for the N-1 columns in the input csv. The Start and End fields in the output file can be empty in some cases.

I'm trying to read in the data using a DictReader . So for i've been able to read in the data with

import unicodecsv
import os
import codecs

with open('test.csv') as csvfile:
    reader = unicodecsv.csv.DictReader(csvfile)
    outfile = codecs.open("test-write", "w", "utf-8")
    outfile.write("Indv", "ATTR", "Value", "Start","End\n")
    for row in reader:
        outfile.write([row['email'],"trait1",row['trait1'],'',''])
        outfile.write([row['email'],"trait2",row['trait2'],row['trait2'],''])
        outfile.write([row['email'],"trait3",row['trait3'],'','')

Which doesn't work. (I think I need to cast the list to a string), and is also very brittle as I'm hardcoding the column names for each row. The bigger issue is that the data within the for loop isn't written to "test-write" . Only the line outfile.write("Indv", "ATTR", "Value", "Start","End\\n") actually write out to the file. Is DictReader the appropriate class to use in my case?

Answer 1

This uses a unicodecsv.DictWriter and the zip() function to do what you want, and the code is fairly readable in my opinion.

import unicodecsv
import os
import codecs

with open('test.csv') as infile, \
     codecs.open('test-write.csv', 'w', 'utf-8') as outfile:

    reader = unicodecsv.DictReader(infile)
    fieldnames = 'Indv,AttrName,AttrValue,Start,End'.split(',')
    writer = unicodecsv.DictWriter(outfile, fieldnames)
    writer.writeheader()
    for row in reader:
        email = row['email']
        trait1, trait2, trait3 = row['trait1'], row['trait2'], row['trait3']
        writer.writerows([  # writes three rows of output from each row of input
            dict(zip(fieldnames, [email, 'trait1', trait1])),
            dict(zip(fieldnames, [email, 'trait2', trait2, trait2])),
            dict(zip(fieldnames, [email, 'trait3', trait3]))])

Here's the contents of the test-write.csv file it produced from your example input csv file:

Indv,AttrName,AttrValue,Start,End
foo@gmail,trait1,biz,,
foo@gmail,trait2,baz,baz,
foo@gmail,trait3,buzz,,
bar@gmail,trait1,bizzy,,
bar@gmail,trait2,bazzy,bazzy,
bar@gmail,trait3,buzzy,,
foobars@gmail,trait1,bizziest,,
foobars@gmail,trait2,bazziest,bazziest,
foobars@gmail,trait3,buzziest,,

Answer 2

I may be completely off since I don't do a lot of work with unicode, but it seems to me that the following should work:

import csv

with open('test.csv', 'ur') as csvin, open('test-write', 'uw') as csvout:
    reader = csv.DictReader(csvin)
    writer = csv.DictWriter(csvout, fieldnames=['Indv', 'AttrName', 
                                                'AttrValue', 'Start', 'End'])
    for row in reader:
        for traitnum in range(1, 4):
            key = "trait{}".format(traitnum)
            writer.writerow({'Indv': row['email'], 'AttrName': key,
                             'AttrValue': row[key]})

Answer 3

import pandas as pd
pd1 = pd.read_csv('input_csv.csv')
pd2 = pd.melt(pd1, id_vars=['email'], value_vars=['trait1','trait2','trait3'], var_name='AttrName', value_name='AttrValue').rename(columns={'email': 'Indv'}).sort(columns=['Indv','AttrName']).reset_index(drop=True)
pd2.to_csv('output_csv.csv', index=False)

Unclear on what the Start and End fields represent, but this gets you everything else.