[issue23041] csv needs more quoting rules

[tegdev]

tegdev <tegdev at dragon.escape.de> added the comment:

The correct handling of None values belongs to the csv module.

There is a use case to migrate a DB2 database to PostgreSQL.
DB2 has a command line tool "db2 export ..." which produces csv-files.
A row 
  ['Hello', null, 'world'] 
is exported to
  "Hello,,"world".

I would like to read in these exports with python and put it to PostgreSQL.

But with the csv library I can't read it in correctly. The input is converted to:
  ['Hello', '', 'world']
It should read as:
  ['Hello', None, 'world']

It is pretty easy to write a correct CSV reader with ANTLR but it's terribly slow.

And last but not least: if someone writes a list the reading should the identity.
Thats not True for the csv libraray.

Example:

import csv
hello_out_lst = ['Hello', None, 'world']
with open('hello.csv', 'w') as ofh:
    writer = csv.writer(ofh, delimiter=',')
    writer.writerow(hello_out_lst)

with open('hello.csv', 'r') as ifh:
    reader = csv.reader(ifh, delimiter=',')
    for row in reader:
        hello_in_lst = row

is_equal = hello_out_lst == hello_in_lst
print(f'{hello_out_lst} is equal {hello_in_lst} ?  {is_equal}')

The result is:
['Hello', None, 'world'] is equal ['Hello', '', 'world'] ?  False

----------
nosy: +tegdev

_______________________________________
Python tracker <report at bugs.python.org>
<https://bugs.python.org/issue23041>
_______________________________________