csvstat: UnicodeError in Python 2 when --csv option enabled

[binarytooth]

csvstat 1.0.2 Python 2.7.12 Ubuntu 16.04.3 LTS xenial

When I run csvstat on a file containing a header and a single euro sign € (ISO-8859-15 character A4), it exits with an error, but only if the --csv option is included. If the option is not included csvstat generates output correctly. A copy of the test file

euro-sign-iso-8859-15.txt

is attached to this ticket.

This happens with all characters from 0x80 (128) through 0xFF (255) with the exception of 0x85 (Next line character) All produce the same error when csvstat is run with the --csv option

When csvstat is run on the test file without the csv option, it produces the following correct output:

$ csvstat -e ISO-8859-15 euro-sign-iso-8859-15.txt

1. "Contents"

Type of data:          Text
Contains null values:  False
Unique values:         1
Longest value:         1 characters
Most common values:    € (1x)

Row count: 1

When I run the same command but with the --csv option enabled it blows up.

$ csvstat -v -e ISO-8859-15 euro-sign-iso-8859-15.txt --csv column_id,column_name,type,nulls,unique,min,max,sum,mean,median,stdev,len,freq Traceback (most recent call last): File "/usr/local/bin/csvstat", line 9, in load_entry_point('csvkit==1.0.2', 'console_scripts', 'csvstat')() File "/usr/local/lib/python2.7/dist-packages/csvkit/utilities/csvstat.py", line 335, in launch_new_instance utility.run() File "/usr/local/lib/python2.7/dist-packages/csvkit/cli.py", line 114, in run self.main() File "/usr/local/lib/python2.7/dist-packages/csvkit/utilities/csvstat.py", line 166, in main self.print_csv(table, column_ids, stats) File "/usr/local/lib/python2.7/dist-packages/csvkit/utilities/csvstat.py", line 318, in print_csv writer.writerow(output_row) File "/usr/local/lib/python2.7/dist-packages/agate/csv_py2.py", line 190, in writerow UnicodeWriter.writerow(self, row) File "/usr/local/lib/python2.7/dist-packages/agate/csv_py2.py", line 103, in writerow self.writer.writerow([six.text_type(s if s is not None else '').encode(self.encoding) for s in row]) File "/usr/lib/python2.7/codecs.py", line 369, in write data, consumed = self.encode(object, self.errors) UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 32: ordinal not in range(128)

This problem also occurs for files encoded using WINDOWS-1252. This bug prevents the use of many useful characters in csv files such as smart quotes, dagger, double dagger, many accented letters, the euro sign, etc.

[jpmckinney]

It's an issue in Python 2 only. I tried to identify the issue, but running the same code in a Python shell doesn't reproduce the error – the error only occurs when run from the command-line.

[aborruso]

The same for me.

You have this error with python 2.7 when you have accented characters.

Try with console and this csv file

fieldA,fieldB
aa,bb
aa,cc
hello,world
à,b

[TomekZet]

I get similar problem with csvlook, only when redirecting output to a file:

$ csvlook -
fieldA,fieldB
aa,bb
aa,cc
hello,world
à,b
| fieldA | fieldB |
| ------ | ------ |
| aa     | bb     |
| aa     | cc     |
| hello  | world  |
| à      | b      |

When printing directly to the console it is ok, because I have UTF-8 locale set:

$ locale
LANG=en_US.UTF-8
LANGUAGE=en_US:en
LC_CTYPE="en_US.UTF-8"
[...]
$ csvlook - > test.tmp
fieldA,fieldB
aa,bb
aa,cc
hello,world
à,b
'ascii' codec can't encode character u'\xe0' in position 2: ordinal not in range(128)

Looks as if you don't encode unicode output properly. I am not sure but probably the missing encode should be addede to https://github.com/wireservice/agate/blob/master/agate/table/print_table.py

The problem is not present in versions <= 1.0.0 The earliest version in which I can reproduce it is 1.0.1

[jpmckinney]

Closing as Python 2 no longer supported.