Pandas: save to excel encoding issue

Question

I have a similar problem to the one mentioned here but none of the suggested methods work for me.

I have a medium size utf-8 .csv file with a lot of non-ascii characters. I am splitting the file by a particular value from one of the columns, and then I'd like to save each of the obtained dataframes as an .xlsx file with the characters preserved.

This doesn't work, as I am getting an error:

UnicodeDecodeError: 'ascii' codec can't decode byte 0xff in position 7: ordinal not in range(128)

Here is what I tried:

1. Using xlsxwriter engine explicitly. This doesn't seem to change anything.

2. Defining a function (below) to change encoding and throw away bad characters. This also doesn't change anything.

   def changeencode(data):
   cols = data.columns
   for col in cols:
   if data[col].dtype == 'O':
       data[col] = data[col].str.decode('utf-8').str.encode('ascii', 'ignore')
   return data   
   

3. Changing by hand all the offensive chars to some others. Still no effect (the quoted error was obtained after this change).

4. Encoding the file as utf-16 (which, I believe, is the correct encoding since I want to be able to manipulate the file from within the excel afterwards) doesn't help either.

I believe that the problem is in the file itself (because of 2 and 3) but I have no idea how to get around it. I'd appreciate any help. The beginning of the file is pasted below.

"Submitted","your-name","youremail","phone","miasto","cityCF","innemiasto","languagesCF","morelanguages","wiek","partnerCF","messageCF","acceptance-795","Submitted Login","Submitted From","2015-12-25 14:07:58 +00:00","Zózia kryś","[email protected]","4444444","Wrocław","","testujemy polskie znaki","Polski","testujemy polskie znaki","44","test","test","1","Justyna","99.111.155.132",

EDIT

Some code (one of the versions, without the splitting part):

import pandas as pd
import string
import xlsxwriter

df = pd.read_csv('path-to-file.csv')

with pd.ExcelWriter ('test.xlsx') as writer:
                df.to_excel(writer, sheet_name = 'sheet1',engine='xlsxwriter')

Solution 1

Supposedly this was a bug in the version of pandas which I was using back then. Right now, in pandas ver. 0.19.2, the code below saves the csv from the question without any trouble (and with correct encoding).
NB: openpyxl module have to be installed on your system.

import pandas as pd
df = pd.read_csv('Desktop/test.csv')
df.to_excel('Desktop/test.xlsx', encoding='utf8')

Solution 2

Try encoding the columns with non-ascii characters as

df['col'] = df['col'].apply(lambda x: unicode(x))

and then save the file to xlsx format with encoding 'utf8'

Comments

[jjj]

Thanks for your suggestion, unfortunately this doesn't work. The same error is returned but now it is triggered by the .apply line.

[Siva Arasu]

Can you attach snippet of the csv file here?

[jjj]

Sorry for the late reply. A part of the file which should be enough to trigger the error is posted in my question. Do you want something more?

[jjj]

@greghor hmm, weird. I installed the same version just now, and it works for me. Do you have openpyxl installed?

[greg hor]

thanks for your reply, I do have openpyxl 2-5-0 installed. After strugling for while, I noticed that if I specify the encoding when loading the data df=pd.read_csv("test.csv", encoding="utf-8"), then it works fine.