CODEWITHSUNDEEP
pythonpython-3.xcsv
TG Gowda
TG Gowda

Reputation: 11947

python3's csv.DictReader is not splitting records by delimiter as expected

I am trying to read a TSV file using csv.DictReader, however, the reader is reporting wrong number of fields. When I looked inside, I found that it is not splitting delimiter properly (I confirmed with str.split(delimiter) and awk -F 'delim'. 

from pprint import pprint
import csv
import sys, os
keys = ['id', 'src', 'src_len', 'sys1_o', 'sys1_b', 'sys1_l',
        'sys2_o', 'sys2_b', 'sys2_l', 'sys3_o', 'sys3_b', 'sys3_l', 'x']
# get this data from https://gist.github.com/thammegowda/95613b203a442fbe72fc5b51af491367 
my_data = """segment-22 Kture . 27  Thutaalchisu.   -   33  Koture. -   27  Th jump.    -   33  3
segment-23  ‘Yunker .   7   "said.  -   8   ‘Yunker.    -   7   "said.  -   8   1"""

tmp_file = "tmp.tsv"
with open(tmp_file, 'w', encoding='utf-8') as f:
    f.write(my_data)

def read_recs_csv(path):
    with open(path, 'rt', encoding='utf-8') as f:
        rdr = csv.DictReader(f, fieldnames=keys, delimiter='\t')
        for rec in rdr:
            yield(dict(rec))

def read_recs_raw(path):
    with open(path, 'rt', encoding='utf-8') as f:
        for line in f:
            rec = dict(zip(keys, line.strip().split('\t')))
            yield(rec)

print("Reading through CSV DictReader ")
pprint(list(read_recs_csv(tmp_file)))
print("Reading directly")
pprint(list(read_recs_raw(tmp_file)))
# Debug
print(sys.version_info)
print(os.environ['LANG'])

Output:

    Reading through CSV DictReader 
[{'id': 'segment-22',
  'src': 'Kture .',
  'src_len': '27',
  'sys1_b': '-',
  'sys1_l': '33',
  'sys1_o': 'Thutaalchisu.',
  'sys2_b': '-',
  'sys2_l': '27',
  'sys2_o': 'Koture.',
  'sys3_b': '-',
  'sys3_l': '33',
  'sys3_o': 'Th jump.',
  'x': '3'},
 {'id': 'segment-23',
  'src': '‘Yunker .',
  'src_len': '7',
  'sys1_b': '-',
  'sys1_l': '8',
  'sys1_o': 'said.\t-\t8\t‘Yunker.\t-\t7\tsaid.',
  'sys2_b': None,
  'sys2_l': None,
  'sys2_o': '1',
  'sys3_b': None,
  'sys3_l': None,
  'sys3_o': None,
  'x': None}]
Reading directly
[{'id': 'segment-22',
  'src': 'Kture .',
  'src_len': '27',
  'sys1_b': '-',
  'sys1_l': '33',
  'sys1_o': 'Thutaalchisu.',
  'sys2_b': '-',
  'sys2_l': '27',
  'sys2_o': 'Koture.',
  'sys3_b': '-',
  'sys3_l': '33',
  'sys3_o': 'Th jump.',
  'x': '3'},
 {'id': 'segment-23',
  'src': '‘Yunker .',
  'src_len': '7',
  'sys1_b': '-',
  'sys1_l': '8',
  'sys1_o': '"said.',
  'sys2_b': '-',
  'sys2_l': '7',
  'sys2_o': '‘Yunker.',
  'sys3_b': '-',
  'sys3_l': '8',
  'sys3_o': '"said.',
  'x': '1'}]
sys.version_info(major=3, minor=6, micro=1, releaselevel='final', serial=0)
en_US.UTF-8

Note: sample data is posted in a gist. Please download it if tabs are replaced by spaces here.

Upvotes: 2

Views: 1611

Answers (1)

Jean-François Fabre
Jean-François Fabre

Reputation: 140196

The difference between raw and csv parsing (and that's why I insisted to get the input data) is that csv module handles quoting by default.

There are quotes in your data, and csv consider quote-protected fields as a single field. awk and str.split don't care.

Just tell csv module not to consider quoting:

rdr = csv.DictReader(f, fieldnames=keys, delimiter='\t', quoting=csv.QUOTE_NONE)

doing this, I get all fields filled.

Upvotes: 1

Related Questions