csv.Error: iterator should return strings, not bytes

Question

You open the file in text mode.

More specifically:

ifile  = open('sample.csv', "rt", encoding=<theencodingofthefile>)

Good guesses for encoding is "ascii" and "utf8". You can also leave the encoding off, and it will use the system default encoding, which tends to be UTF8, but may be something else.

The reason it is throwing that exception is because you have the argument rb, which opens the file in binary mode. Change that to r, which will by default open the file in text mode.

Your code:

import csv
ifile  = open('sample.csv', "rb")
read = csv.reader(ifile)
for row in read :
    print (row) 

New code:

import csv
ifile  = open('sample.csv', "r")
read = csv.reader(ifile)
for row in read :
    print (row)

In Python3, csv.reader expects, that passed iterable returns strings, not bytes. Here is one more solution to this problem, that uses codecs module:

import csv
import codecs
ifile  = open('sample.csv', "rb")
read = csv.reader(codecs.iterdecode(ifile, 'utf-8'))
for row in read :
    print (row) 

Your problem is you have the b in the open flag. The flag rt (read, text) is the default, so, using the context manager, simply do this:

with open('sample.csv') as ifile:
    read = csv.reader(ifile) 
    for row in read:
        print (row)  

The context manager means you don't need generic error handling (without which you may get stuck with the file open, especially in an interpreter), because it will automatically close the file on an error, or on exiting the context.

The above is the same as:

with open('sample.csv', 'r') as ifile:
    ...

or

with open('sample.csv', 'rt') as ifile:
    ...

I had this error when running an old python script developped with Python 2.6.4

When updating to 3.6.2, I had to remove all 'rb' parameters from open calls in order to fix this csv reading error.