Pandas - Writing an excel file containing unicode

Question

I have the following code:

import pandas as pd

x = [u'string with some unicode: \x16']
df = pd.DataFrame(x)

If I try to write this dataframe as an excel file:

df.to_excel("test.xlsx")

Or, if I try to write this dataframe as an excel file, with utf-8 encoding:

ew = pd.ExcelWriter('test.xlsx',options={'encoding':'utf-8'})
df.to_excel(ew)

I get the following error:

IllegalCharacterError                     Traceback (most recent call last)
<ipython-input-4-62adec25ae8d> in <module>()
      1 ew = pd.ExcelWriter('test.xlsx',options={'encoding':'utf-8'})
      2 #df.to_excel("test.xlsx")
----> 3 df.to_excel(ew)

/usr/local/lib/python2.7/dist-packages/pandas/util/decorators.pyc in wrapper(*args, **kwargs)
     86                 else:
     87                     kwargs[new_arg_name] = new_arg_value
---> 88             return func(*args, **kwargs)
     89         return wrapper
     90     return _deprecate_kwarg

/usr/local/lib/python2.7/dist-packages/pandas/core/frame.pyc in to_excel(self, excel_writer, sheet_name, na_rep, float_format, columns, header, index, index_label, startrow, startcol, engine, merge_cells, encoding, inf_rep)
   1258         formatted_cells = formatter.get_formatted_cells()
   1259         excel_writer.write_cells(formatted_cells, sheet_name,
-> 1260                                  startrow=startrow, startcol=startcol)
   1261         if need_save:
   1262             excel_writer.save()

/usr/local/lib/python2.7/dist-packages/pandas/io/excel.pyc in write_cells(self, cells, sheet_name, startrow, startcol)
    679             colletter = get_column_letter(startcol + cell.col + 1)
    680             xcell = wks.cell("%s%s" % (colletter, startrow + cell.row + 1))
--> 681             xcell.value = _conv_value(cell.val)
    682             style_kwargs = {}
    683 

/usr/local/lib/python2.7/dist-packages/openpyxl/cell/cell.pyc in value(self, value)
    360     def value(self, value):
    361         """Set the value and infer type and display options."""
--> 362         self._bind_value(value)
    363 
    364     @property

/usr/local/lib/python2.7/dist-packages/openpyxl/cell/cell.pyc in _bind_value(self, value)
    269             elif self.guess_types:
    270                 value = self._infer_value(value)
--> 271         self.set_explicit_value(value, self.data_type)
    272 
    273 

/usr/local/lib/python2.7/dist-packages/openpyxl/cell/cell.pyc in set_explicit_value(self, value, data_type)
    235             raise ValueError('Invalid data type: %s' % data_type)
    236         if isinstance(value, STRING_TYPES):
--> 237             value = self.check_string(value)
    238         self._value = value
    239         self.data_type = data_type

/usr/local/lib/python2.7/dist-packages/openpyxl/cell/cell.pyc in check_string(self, value)
    220         value = value[:32767]
    221         if next(ILLEGAL_CHARACTERS_RE.finditer(value), None):
--> 222             raise IllegalCharacterError
    223         return value
    224 

IllegalCharacterError: 

How can I write a pandas dataframe containing unicode to an excel file?

Answer 1

Use this to remove any error that you might be getting. You can save to excel post this.

df = df.applymap(lambda x: x.encode('unicode_escape').
                 decode('utf-8') if isinstance(x, str) else x)

Answer 2

Not a Unicode issue as such... \x16 (or in Unicode strings \u0016 refers to the same character) is ASCII control code 22 (SYN). Pandas says it's invalid to have control codes (other than tab and newlines) in an Excel file, and though I don't know much about Excel files it would certainly be impossible to include them in an XML 1.0 file, which is what's inside a xlsx.

So most likely there is no way to include arbitrary character sequences (with control codes) in an Excel. You should filter them out before writing, or if you really need to preserve the original data use some form of ad hoc encoding recognised only by your application.

Answer 3

When I encounter this error, I usually go around it by writing the file to a '.csv instead of '.xlsx' files. So instead of

yourdataframe.to_excel('Your workbook name.xlsx')

I would do:

yourdataframe.to_csv('Your workbook name.csv')

It appears the way pandas decodes .csv files by default is:

encoding : string, optional
A string representing the encoding to use in the output file,
defaults to 'ascii' on Python 2 and 'utf-8' on Python 3.

On the other hand default encoding of .xlsx files is:

encoding: string, default None
encoding of the resulting excel file. Only necessary for xlwt,
other writers support unicode natively.

This difference is responsible for that error. You will also get the error when you write data with strings that start with - or + to a .xlsx file.

Answer 4

I don't know this particular language, but generally there is an error with excel and UTF8. If you just open a file of UTF8 characters with excel programatically, it will corrupt them (it doesn't seem to handle all the bits in the character, but truncates it to effectively the first 2 and last 2 hex numbers of the 8 present in extended characters).

A work around, to load a utf file correctly into excel, is to get the program insert a macro into your excel sheet after you have loaded it which imports the data. I have some code to do this in C#, if that's any help?

does your input contain any extended characters (i.e. àâäçæèëéêìïîñòöôœûüùÿÀÂÄÇÆÈËÉÊÌÏÎÑÒÖÔŒÛÜÙŸ) and if you take them out, does it work?

Answer 5

The same problem happened to me. I solved it as follows:

First, install python package xlsxwriter:

pip install xlsxwriter

Second, replace the default engine 'openpyxl' with 'xlsxwriter':

df.to_excel("test.xlsx", engine='xlsxwriter')